User talk:Jimbo Wales/Personal Image Filter

This is a talk page!

Wikipedia's prime objective
(Edit conflict) Its clear that adult content is an issue, and if Wikipedia wants to live up to its Prime objective, its clear that something needs to be done. While I disagree with the idea of an image filter - I think we can and should impose our encyclopedic concerns on Commons - I think this is a good way to move the issue forward. Regards, -Stevertigo (t | c) 20:02, 15 July 2012 (UTC)

Protection against involuntary censorship
Prior to categorizing commons images with censorship categories, changes should be made to the image serving system to make it as hard as possible for any third party censorship system to use the censorship categories to censor anyone involuntarily. A third party censorship program could scrap commons category information, and then use that information to block those images on all projects, regardless of whether the project opted into the image filter, and regardless of whether the individual user wants the images censored. While a program could censor images now, the risk is greatly increased if we provide widespread censorship categories for such a program to use. To avoid such a program piggybacking off the censorship categories, we should make sure that image URLs are not the same on commons as they are on other projects. It must not be possible to predict the URL that an image link on a different project will generate by looking at the image URLs generated by commons. We should also be wary of any other ways a censorship program could piggyback on the censorship categories to involuntarily censor readers. Monty 845  20:00, 15 July 2012 (UTC)


 * To Monty: An image filter system would naturally use descriptive tags to identify adult content. It seems the only other way would be to use some kind of unreadable or hidden tags, which doesn't make sense in an openly editable project. Hence third party content systems can do what they want, even so far as to do what they are actually designed to do. We are not under any real charge to make their job any more difficult, by deliberately trying to circumvent them.
 * More importantly, censorship is only one issue here, and in fact I think its a red herring because we are some of the most open, free-sharing people on the planet. We are not inclined to censor, however we are inclined to be encyclopedic. There are other issues involved, namely the issue of Wikipedia's primary mission, and its overall design of being an encyclopedia. Much of the content we are talking about cannot be construed by any rational person as being encyclopedic - the proper description for these images is "exhibitionist," or simply "prurient." In our last Talk:Jimbo discussion about this issue, someone was claiming that the pic of a woman with a cucumber in her yoni was encyclopedic and educational. I was left with no choice but to completely destroy his argument. -Stevertigo (t | c) 20:18, 15 July 2012 (UTC)
 * I believe that censorship is fundamentally inconsistent with the spirit of Wikipedia, and we should do anything that promotes censorship. The spirit of the image filter as discussed here is that its not censorship if you only filter what you see yourself. Yet in supporting the image filter, we risk aiding the cause of those who would impose involuntary censorship on others. To the maximum extent possible we should avoid aiding such censors. I opposed the whole concept of the image filter for this very reason, but arguendo, if we are to have an image filter, it must not facilitate involuntary censorship. Monty  845  20:32, 15 July 2012 (UTC)

You may not remember this but the image filter idea first came up in 2004 after the murder of Nick Berg. Someone uploaded an image of Berg's corpse, namely his severed head, and its inclusion in the Berg article was the focus of a long and acrimonious debate over "censorship" and our apparent need to freely impose images of murder on everybody whether they want to see such or not. In your 'censorship free' Wikipedia would we all have to manually opt out of seeing such unnecessary imagery? - Stevertigo (t | c) 20:59, 15 July 2012 (UTC)


 * I think it is completely unnecessary for us to change the category system at Commons at all. Our existing category system is already perfectly usable for this purpose, and if that is problematic in terms of 3rd party censors using it, then that's already true.  In fact, I am unaware of any evidence one way or the other as to whether the Chinese government (for example) takes any note of our category system specifically.  I doubt it.--Jimbo Wales (talk) 21:24, 15 July 2012 (UTC)
 * If we are really going to avoid additional categorization for the purpose of the filter then your right, but many of the filter proposals I have seen have assumed such additional categorization. It will be important to have some sort of protection against categorization creep as a result of the filter. Monty  845  01:23, 16 July 2012 (UTC)

Discussion by GMaxwell

 * Allow me to suggest some old technical discussions: and my comments on . The idea here is to implement in mediawiki a simple general mechanism where tags (of any kind) on the image description page are passed through into the HTML of the pages where the images are used as CSS classes on the images.  Then user JS / gadgets/ etc. could be triggered based on this and do whatever the user wants.
 * So for example, the communities could tag images as containing human nudity— as pure metadata— and user could enable gadges/js hide those images, or make them only display on click/mouseover. Likewise, uses with different motivations could display the nudity 2x larger, or more likely people with poor eyesight could do the same for images containing small text. Alternatively, Photographers could trigger images with non-free licenses or cleanup tags to get highlighted so they'd find more things to fix. Editors could have watermarks display when an image is up for deletion. etc.
 * This is a general mechanism which doesn't presuppose a particular use and in particular doesn't presuppose any project or culturally specific definitions of appropriateness. It would allow the community to create and evolve novel uses organically, bottom up, and through consensus and cooperation, the way most of our successful initiatives happen, rather than handcuffing us with software that presupposes a particular way of doing something which we've not done before and are not of one mind about. I think if the original image filter discussion had focused on general mechanisms and not on imposing a specific policy favored by some they would have been a lot less contentious and more successful. Lets not repeat those mistakes.
 * Cheers --Gmaxwell (talk) 20:03, 15 July 2012 (UTC)


 * Why bother? [adblock-plus] can with the appropriate blacklists be used to block the images you don't like from any website including wikipedia. Requires no new code and no effort on our part.©Geni 20:05, 15 July 2012 (UTC)
 * Passing metadata along would make it far to easy for a third party censorship program to use the metadata to involuntarily censor images. No censorship enabling data should be passed on by the server unless a reader affirmatively opts in to the filter. Monty  845  20:13, 15 July 2012 (UTC)


 * We already pass _tons_ of data that censorship infrastructure could (and does in some cases) key on, including the file names, which can be automatically equated to categories by censorware spiders. I share your concern on enabling censorship, and perhaps communities would opt to not apply tags which had a poor ratio of personal empowerment to involuntary censorship risk, though that is a question for which tags get used— not a question for a general mechanism of making them available for inlined objects.
 * Ultimately, if you're concerned about censorship you should be pushing to get the entire site moved behind HTTPS by default, and getting the WMF to invest in the infrastructure to make that possible, and doing things like using control over its trademark to go after involuntary censors in places where the rule of law still functions, and supporting tools like Tor which can be used for circumvention in places where it doesn't. --Gmaxwell (talk) 20:35, 15 July 2012 (UTC)


 * Given that we really want such a system and that we want to enable it. Is tagging categories a viable option? From my experience i can tell that the existing categories aren't good or meant for this purpose. They were not invented to differentiate NSFW content from the usual content and it is a current bad practice to do so. We have many categories that contain some NSFW images but also many other images that are SWF. A good example would be the category Violence. How do you think should we handle this cases? (This question is directed at any reader, not only Jimbo himself) -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 20:18, 15 July 2012 (UTC)
 * That's pretty much what I wanted to point to, as well. I fear that this would rather require a new categorization system aside from the current, which is at least to a large extent rather based on theme than on the form of the image contents. Cheers, —Pill (talk) 20:21, 15 July 2012 (UTC)
 * I ask this because it was one argument inside the German poll, that creating categories for the purpose of filtering (for example: "Bananas" and "Bananas (NSFW)/Bananas in ..." ) isn't seen as an good idea. It disrupts the experience of users that don't want to use or support a filtering system and matches the view of the ALA which was widely shared in the latest polls. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 20:26, 15 July 2012 (UTC)
 * I just looked through Commons:Category:Bananas and didn't see anything NSFW. I think that, in general, cases where an image only belongs in an 'innocent' category like 'Bananas' without also belonging in a clearly NSFW category like Commons:Category:Sexual penetrative use of bananas but which we'd generally agree should be in an NSFW filter is vanishingly small.  NPOV is powerful!--Jimbo Wales (talk) 21:13, 15 July 2012 (UTC)
 * What Pill said— think it is because it adds another purpose to categories. The ideal "nsfw" tagging (as if that were just a function of the image and not the application of the image and the user...) can't be expected to agree with the ideal category usage, and so using the same thing for both purposes would inevitably result in compromises. I'd rather we have something where people could express preferences like "hide nudity except in articles which aren't primarily about human sexuality", something that would be possible with a metadata driven gadget but not possibly with simplistic blacklists. --Gmaxwell (talk) 20:35, 15 July 2012 (UTC)
 * I advocate not changing anything about categories at all. There should be zero effort on the part of the commons community.  Our existing categories are perfectly good for doing the work that needs doing.
 * I also don't advocate for solutions which would require user js edits. I want to be able to turn this on and off with a single click.  I am not in favor of adding any complexity to the system in version 1.0, so adding arcane options like "hide nudity except in articles which aren't primarily about human sexuality" should be put on hold or kept as thought experiments.  I actually think such options are undesirable, as compared to keeping it extremely lightweight for people to *with a single click* open any image which has been hidden.
 * Two reasons (not the only two) for an extreme-simplicity design principles: (1) it forces us to keep the system really easy to turn on and off, so that any categorization errors (which are inevitable) are no big deal (2) it makes it possible to explain to people who are fearful and don't want to read a long document.--Jimbo Wales (talk) 20:57, 15 July 2012 (UTC)
 * I understand your intention with respect to the simplicity of the solution (and I agree that this is something we should strive for), but I'm wondering how you come to the conclusion that "[o]ur existing categories are perfectly good for doing the work that needs doing." To elaborate on my point a bit more: I see a fundamental difference between the categorization scheme employed on Commons and the one required for this purpose. If, say, I am looking for pictures of male genitalia on Wikimedia Commons, I am expecting all images that display male genitalia. The criterion, therefore, is the images sujet; it is not the particular form of display. For the purpose of filtering for critical content, however, the question is precisely that. We would probably agree that a drawing as can be found in medical books does not constitute "critical content," while a photograph may very well do. But nevertheless both images are perfectly suitable for the same category on Commons. Your point is that it should simply be made easy to toggle the feature off; but the usefulness is fundamentally impaired if there is a high number of false positives. Cheers, —Pill (talk) 21:18, 15 July 2012 (UTC)
 * I just don't see this as a serious problem. Yes, there will be relatively rare categorization errors in some edge cases but that doesn't diminish the usefulness of the system.  Right now we have a system that is 100% useless for those desiring an NSFW filter.  A high number of false positive or negatives is a bad thing but (a) unlikely because the edge cases really are rare and (b) not a huge deal as long as it is super easy to turn it on or off.  If you are looking for male genitalia on commons, you probably should click the filter off.  If you don't s want male genitalia to show up on your screen without a single click 'unhide' then you should probably turn the filter on.  If you want something in the middle, then our filter just doesn't do everything.  But 'doesn't do everything' doesn't strike me as much of an objection! :-) --Jimbo Wales (talk) 21:31, 15 July 2012 (UTC)
 * To add some examples to it, think about commons:Category:Porn actresses (or the +1,000 file subcategory "unidentified porn actresses"), commons:Category:Film sets of pornographic films, commons:Category:Erotic engravings or similar. Cheers, —Pill (talk) 21:25, 15 July 2012 (UTC)
 * Answered in some detail, below. In addition to the first one, which I discuss there, I looked at the others and found the general principle to apply quite well.  It really is extremely rare to have an image that is NSFW that isn't categorized in a category that is exclusively or predominantly NSFW.  There are 'mixed' categories like these, but they don't pose a problem as long as there are NPOV subcategories to do the heavy lifting.--Jimbo Wales (talk) 21:42, 15 July 2012 (UTC)

Prior discussions
Relevant previous discussions that may help inform this effort are here:


 * http://meta.wikimedia.org/wiki/Controversial_content/Brainstorming
 * http://meta.wikimedia.org/wiki/Wikilegal/Age_Record_Requirement
 * http://meta.wikimedia.org/wiki/Controversial_content/Problems

Good luck! -- J N  466  21:15, 15 July 2012 (UTC)

Commons:Category:Violence
I was asked up above about this category. It's worth some commentary.

This category contains only 121 files. The vast majority of them would appear in any newspaper and are therefore not NSFW. I only see one that is problematic, and that one is. This category would therefore not belong in the image filter, and this image would slip through. Perfection is not a valid goal.

One might argue that this image should be placed into a more specific category as well, but if that isn't possible, then I don't mind one way or the other. Key to this design exercise is keeping everything very lightweight and being tolerant of some errors.--Jimbo Wales (talk) 21:08, 15 July 2012 (UTC)


 * You mentioned previously that the NSFW categorizing should not affect the current category system. But would it not somehow, even if not intended, enforce the creation of categories to improve filtering? I would expect it, that at some point categories are solely created to serve this single purpose or that images are moved in and out of such categories to enforce or avoid NSFW categorization. What should we do about this?
 * One other aspect would be the visibility to the public. Could we expect that public terminals (in broad sense, even including internet providers) would use our categorization data (the new tagging as NSFW) to exclude content against the will of the single user that isn't free to choose another way to access Wikipedia? A plausible scenario would be a medical article and a student that wants to look at the illustrations, but he can't look at them because the university network gurus decided to use the NSFW Filter to exclude pornography. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 21:54, 15 July 2012 (UTC)
 * I don't see any reason for any new categories to be created, no. It is possible that people may notice that our current NPOV category system could be improved in various ways.  (Let's say, for example, that we had a category for 'Porn actresses' but not one for 'Nude porn actresses' - people might well create one, but so what?  (In fact, in most cases, we do already deal with such things quite well.)
 * Second, the idea that external censors can use our category system to improve their filtering software strikes me as nothing more than FUD. What I mean by that is that if they wanted to, they already could do it quite easily.  The categories are already there.  That we create a new user preference for users to allow them to have some images displayed beyond a one-click javascript hide appears to me to have exactly zero impact on the ability of external providers to do whatever they want - it neither helps nor hinders them in the slightest.--Jimbo Wales (talk) 11:53, 16 July 2012 (UTC)
 * "Let's say, for example, that we had a category for 'Porn actresses' but not one for 'Nude porn actresses'"


 * Thats exactly what shouldn't be done. It clutters the existing category system, making it even more complicated to navigate and increases the efforts to maintain it. This pushes the additional work in the hands of the community that already suffers due to missing convenient tools to maintain it. Even editors with good experience have already problems to figure out the right categories and first time users struggle to find any appropriate categories. You say that it is already done well, but most editors on Commons would strongly disagree with your view. Saying that an issue isn't big doesn't make it smaller or eradicates it's existence.
 * No, they can't do this easily at the moment. We enable it right at the moment when we start to tag and create categories for filtering reasons. Thats one thing you never understood in the previous debates and you still ignore. If you really want to find a compromise, then you shouldn't ignore facts or describe elephants as ants. I have the strong impression that you say "The filter is dead. Long live the filter!" by playing down the issues that make your current approach not much different from the previous one. I hoped for some kind of genuine idea, but what i see is the same black sheep in a new bag. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 14:02, 16 July 2012 (UTC)


 * "Key to this design exercise is keeping everything very lightweight and being tolerant of some errors" - Noble thought Jimbo, but I think as hopeless the goal of perfection would be. The complexity of editing and administering Wikipedia has grown apace with the project itself, and I think it would be naive to hope that this process would remain lightweight in the long run. You are going to see people trying to catch these one-offs and recategorizing them, and you are going to see people arguing about it.  If the PIM is to go forward, I think this must be accepted as a consequence of doing so, and that both Commons and the local Wikipedias will endure a period of adjustment in the time that follows. Resolute 14:16, 16 July 2012 (UTC)

Commons:Category:Porn actresses
I was asked about this category, up above, and I wanted to work through my thinking on this one here separately so it wouldn't be lost in the discussion up above.

This category contains only 23 files. Only 5 of them contain nudity. Of those, 2 are in the category "Nude porn actresses" as well, and 3 are in the category "Topless porn actresses". Therefore including the latter two categories as NSFW and leaving the first one out, would work perfectly well. This is a good example of how excellent and detailed NPOV categories are very useful for our purposes. Errors are rare.--Jimbo Wales (talk) 21:37, 15 July 2012 (UTC)
 * Toplessness implies to me that you intend to draw the line at what I in the UK would consider soft porn. Some would consider it an error to draw the line so as to include quite a large part of western art in your filter. Others would be unhappy if you have a filter that allows bare arms, midriffs and thighs. Errors are partly a matter of cultural perspective, but even if you consider the female breast taboo you will have errors in a category based system simply because someone categorising a particular image may not have concerned themselves with an element of nudity that to them was minor and inconsequential.  Ϣere Spiel  Chequers  22:46, 15 July 2012 (UTC)


 * As I say in I believe this is a question best left to the community by allowing a variety of community generated filters rather than the foundation getting involved in such decisions. Dmcq (talk) 01:13, 16 July 2012 (UTC)
 * How the filters are implemented is an integral part of the filter, and could be done in ways objectionable to many. The fear of bad filters/categorization is likely to be a source of opposition to the filter proposal as a whole if not adequately addressed in the proposal. There are already factions on en.wikipedia agitating for a break from commons and if not handled well this would give them a big boost. Monty  845  01:18, 16 July 2012 (UTC)

I think that in a first pass implementation - remember my goal here is to create something broadly acceptable to nearly everyone in the community - we should strongly avoid complexity. Imagining a "perfect" system that becomes unusable and complex (for anyone) is a sure-fire way to create unpopularity. Coming up with something that solves the main problem with a minimum of fuss is a great first step, and yes, we might imagine more complex solutions in the future.

In particular, regarding the question of "where to draw the line" I am agnostic. I believe that the cultural differences that people imagine are a great deal smaller than usually supposed. I propose that we simply follow industry standards - reviewing the policies of Google, Flickr (which hosts vast quantities of porn which you can filter if you like or view if you like), and similar would be worthwhile.

And keep in mind my most important point: a system which simply uses a one-click javascript hide is highly resilient to categorization errors. If you have it turned on, and a particular nude image is hidden, you'll see "nsfw image hidden" and you can click to see it if the circumstances are right for you. Alternatively, if a particular nude image is shown, you can just click to turn it off if you like. Allowing user choice to be easy and straightforward is the best way forward, because it makes errors less of a pain in the neck for users to deal with.--Jimbo Wales (talk) 12:00, 16 July 2012 (UTC)
 * If you are proposing to follow "industry standards" then you are proposing to go with a system based on the financially dominant culture of the developed world - specifically skewed to North America. That is especially true if you want a simple one size fits all system, as at least one commercial operator has a "swimsuit" setting for certain middle eastern countries. As for where to draw the line you may think you are agnostic, but your proposal is for a simplistic one size fits all approach where clear lines have to be drawn between what is in the filter and what isn't. Your goal may be to create something broadly acceptable to nearly everyone in the community, but you are starting by going back to the proposal that was made late last year and which was opposed by a very large part of the community including some like me who support the principle of an image filter.  Ϣere Spiel  Chequers  17:35, 16 July 2012 (UTC)

Can a technical solution work for a non-technical problem?
While I understand that this page is set up under the premise that a technical solution is possible and that previous discussions have largely pointed to some kind of technical solution, I feel its necessary to make some elementary points about the nature of the problem and the nature of the proposed solution. Technical solutions are typically reserved for technical problems, and when technical solutions are presented for non-technical problems, there is often a mismatch. So for example when we needed to make recursive usage of text, our coders came up with the template and transclusion system. When we wanted to categorize pages, our coders came up with the category tag system.

This doesn't appear to be a technical problem, rather it appears to be a content and purpose problem. Hence a technical solution would aim to be a kludge at best. Its a content and purpose problem because its an matter of content, such that issues about its inclusion or disinclusion are editorial, and not a matter of taste or personal choice. The issue of whether or not to include an image about a woman with a cucumber in her womanhood is editorial, not an issue of preference. And its an issue about purpose because the use of such images goes to the issue of Wikipedia's overall purpose and mission. Is it our purpose to be a catch-all for any kind of topic, or any kind of material? Is it our purpose to cater to the prurient interests of certain readers? Is it "censorship" to simply stay constrained to our encyclopedic mission, and reject material which is not encyclopedic? I think its clear that we are well within our rights to refuse to be a catch-all. One of the earliest critiques of Wikipedia claimed that Wikipedia would eventually become just a wiki, and not an encyclopedia. We have to resist becoming just a wiki, and remain an encyclopedia, and this is one of those issues where the choice becomes clear.

I agree that the issue of Commons and its general autonomy makes this difficult. However even Commons should have boundaries. For example we can imagine there are some cultures with their own language wikis which are extremely libertine and accept any and all media content. Commons should not accommodate such broad usage. On the other hand, we know of certain cultures which reject freedom of expression and promote uniquely extreme limitations on personal freedom. Of course Commons should not accommodate such strict constraints. I think there is a balance to be found. -Stevertigo (t | c) 22:03, 15 July 2012 (UTC)

Contrast with Wikia
Hi Jimmy, with all respect to Wikia, I believe that there you can impose an image filter tuned to Western Culture, but we shouldn't do that here. Our aim is to serve the whole of humanity, some of our projects most notably Wikipedia have a policy of having a neutral point of view. That is incompatible with having an image filter tuned to one culture's mores. If we introduce an image filter we should do so in a way that works whatever your cultural background. Some time ago I detailed a proposal at m:Controversial content/Brainstorming/personal private filters. This proposal doesn't rely on our existing categorisation system, and doesn't try to impose a rigid clear black and white line where there are naturally many shades of grey. I can see the superficial attraction of a system that would be easier to code and that one could start by hiring a bunch of people to get almost there. But think how these two approaches would handle the following scenarios: I think we can and should introduce an image filter, but if we do so it is worth investing in a solution that is harder to game and fits with our global multicultural ethos.  Ϣere Spiel  Chequers  22:32, 15 July 2012 (UTC)
 * 1) Borderline cases. With a simple category based system you would have to draw a line somewhere, and sure as there are trolls on the Internet there will be be borderline cases to argue about.
 * 2) Editorial arguments. If people know whether an image is in or out of the filter then there will be arguments as to whether or not such an image is appropriate to use in any article, or conversely whether even considering such a US based categorisation is inappropriate for certain projects.
 * 3) Uploader responsibility. We have over 13 million files on Commons plus a loads more on various other projects. Many are essentially uncategorised, and if you rely on paid staff to patrol the uploads they are going to have a big job. I can't see a category based system being introduced without ultimately an obligation on the uploaders to flag their content as controversial. This puts your proposal on a collision course with our GLAM partnerships and especially the image releases involved, both for the practical reason that such uploads are often poorly categorised due to metadata incompatibilities, and also because some of our GLAM partners will have ethical concerns about censorship. Especially if any of the stuff being filtered starts to remind people of the time when museum creators in London vandalised statues by replacing genitals with fig leaves.
 * 4) Vandalism. A category based system is vulnerable to various forms of vandalism, not least people trying to sneak something past the censors.
 * 5) BLP problems. Just because an individual has agreed to an image being taken doesn't mean they are going to be happy that what they consider is not pornographic will be labelled pornographic by "Wikipedia".
 * My main response here is to this: "A category based system is vulnerable to various forms of vandalism, not least people trying to sneak something past the censors". I am proposing a single-click javascript hide/unhide, to be enabled by users at their own choice, and so characterizing it some "censors" one might try to sneak past is really not persuasive.  We already know that people try to sneak giant penis pictures into various articles all the time, and that's nothing new.  I estimate that only a tiny percentage of our audience (1%? 5%? no one really knows) will want to turn on the feature, and the chances that we'll have to deal with an enormous new wave of bad behavior strikes me as quite small - and the result of successful vandalism will be extremely minimal.  (If you manage to get a NSFW image to show for users with a 'please hide NSFW images' preference turned on, then so what?  Those users who would like such a filter do not expect it to be perfect.
 * That some people will be unhappy if their image is hidden to a minority of users who have chosen to hide NSFW images, perhaps in a category for good NPOV reasons, and perhaps due to an error in categorization, strikes me again as unnecessary fear. Such cases will be incredibly rare, and either the person is simply wrong to be unhappy, or we can solve their unhappiness by fixing the category error.--Jimbo Wales (talk) 12:07, 16 July 2012 (UTC)

Good intentions have positive karma
I would like to remind, that even if results are not perfect, showing some pro-active partial solutions also has a multiplier effect, as a good-will dividend, to invite other people to help solve the problem(s). Otherwise, people tend to wait and hope for leadership, but will "join a parade" in progress. -Wikid77 (talk) 23:12, 15 July 2012 (UTC)
 * I agree. This community can do this. Dennis Brown has mentioned making stone soup in other initiatives. I think his analogy works here, too. NewtonGeek (talk) 23:39, 15 July 2012 (UTC)
 * The problem I'm seeing is that it seems like there is no interest in any solution that doesn't involve devising some authoritative global rating of images as good or bad. If the true objective is not to provide users with options for display, but to become judgmental on an official, centralized basis about image content, then the outcome will not be good. Wnt (talk) 13:29, 17 July 2012 (UTC)
 * I wouldn't agree there was no interest in a solution that sort to address as many of the objections as practical. But it does seem that there is insufficient willingness for compromise for a filter to be implemented. However I wouldn't totally give up hope, I've had very few objections from people who are against any sort of filter even one that sets out to address as many of the objections as possible. So in my view consensus for a filter could be achieved. But only if those who want a filter were willing to compromise and try to encompass the concerns of those who've opposed the two proposals that were based on the existing Commons categories.  Ϣere Spiel  Chequers  18:00, 26 July 2012 (UTC)
 * Not only existing categories. Most people from the opposition (me included) also disagree with the introduction of additional categories/labels/tags/... that are based on or are someones POV. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 18:24, 26 July 2012 (UTC)
 * They are not. They base on INTERsubjective view of the community. This is a huge difference. Also, we make such "subjective" decisions about notability, AfD, etc. Pundit | utter  23:09, 26 July 2012 (UTC)
 * Would you please stop your POV pushing? There is no such a thing as an objective community view regarding subjective topics or subjective categorization. We as the community can agree on objective topics, some usual trouble aside, but agreeing on subjective categorization is a whole different thing. There isn't even a happy solution for a single community. There would be just a majority vote, ignoring the arguments of minorities. The best way to avoid such conflicts is to stay neutral. I have nothing against objective categories. But subjective, personal stuff should not be represented to the reader.
 * PS: Judging and agreeing on notability is based on references and sources. They are the neutral arguments. Additionally the inclusion or exclusion of a topic isn't comparable to stick a rating/advisory message onto a topic that we cover. We don't write the receptions of a movie by ourself, but ideally cite critics from various sources that show the whole range of criticism or views. If i would go along with your argumentation then we would write the receptions of a movie (image) by ourself after gaining consensus that the movie (image) was good or bad, ignoring any other opinion. But that wouldn't happen, because it would be a violation of NPOV. Why is tagging an image as god or bad (nice to look at or pure ugly) any different? -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 10:58, 27 July 2012 (UTC)
 * With all due respect, my expressing my honest opinions is no different than your expressing yours, and I do not run around calling it "POV pushing", even though I find your point of view very extreme, and quite incompatible with wiki-projects philosophy. I see no difference between the community's consensus-based decision about categories, and labels. Wait, there is one difference: while Commons categories are decided for all projects collectively and thus are more crude and less adjusted to projects' needs, with proper labeling/filtering, dependent on each project's communal choice, it is a much better fit. Also, I really don't understand why should you and several others dictate the projects that they should not be allowed to use filters, if they want to. Take your opinions to projects you're active in, and persuade your fellows there not to use filters, but don't dictate how should other projects proceed. This is exactly the kind of total oppression you rhetorically seem to criticize. Pundit | utter  16:38, 27 July 2012 (UTC)
 * You miss the point that Commons categories are viewpoint neutral aids for the user, while the discussed labels aren't viewpoint neutral. If they would be viewpoint neutral, then i would not have a problem and we would not have this discussion. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 17:32, 27 July 2012 (UTC)
 * I don't exactly miss any points, I just respectfully disagree. I consider a wide consensus decision to be as close to neutral aid as it can get anywhere. But I think our stances are quite clear now. Cheers Pundit | utter  17:35, 27 July 2012 (UTC)
 * Thats a point we disagree on. My argument is that the median of our community isn't the same as neutrality. Even our median is a point of view, our (median) point of view. Did you miss my movie argument? If not, could you please explain why we don't write our own receptions? -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 17:42, 27 July 2012 (UTC)
 * The thing is, there is no abstract objective neutrality anywhere, ever. There is, however, a communal consensus. We use it widely to make "arbitrary" decisions about excluding images from our repositories, and you seem to be fine with this (even though your argument could be very easily extended to "anything goes, all deletions are POV, we should keep all pics"). We don't comment on movies for a very simple reason: our readers don't expect us to pass judgments of this sort (they have different websites which do exactly that), but they very often stumble upon controversial images unexpectedly, feel hurt or offended, and consider it a breach of trust from Wikimedia projects' side. If you spent even some minimal time on OTRS, you'd know, and at least acknowledge this reality, even if you don't consider it worthwhile to address it as a problem. Pundit | utter  18:01, 27 July 2012 (UTC)
 * I see no difference between writing our own receptions or labeling images depended on our point of view. We would rate a movie or we would rate an image. Please note that i differentiate between the inclusion/exclusion of content and the rating of included content. Not covering a topic does not expose our judgment about a topic and if we cover it, then we follow the NPOV policy.
 * If you need real life example, then let me compare our relationship with our content to a court case as an analogy. Please read with care to not confuse the roles. If we cover a topic then we opened a court case at court, which is run by the government, the WMF. The sources of the articles are the lawyers representing various evidence and opinions. It is the judges, the author(s), role to hear the arguments and to structure the process, the article. It is also the judges role to select which evidence is accepted. He has the difficult role to hold an unbiased process, like our authors have to follow the NPOV principle. On the other end is the jury, the readers. It hears the arguments, looks at the evidence presented by the parties involved and comes to a conclusion about the case, like the reader is free to come to a conclusion after reading the article. Thats how a law case tries to achieve equal treatment and fairness in every case. If i stay put with this analogy, then what you propose is like a judge that asks the jury to look away, because some evidence might be shocking. A fair trial, neutral covering of a topic would be something else. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 19:20, 27 July 2012 (UTC)
 * Let's not make analogies too distant and too fuzzy. I gave you a clear example: we, as the community, ultimately decide if an image is classified as porn and deleted, or if it is classified as art or educational and kept. These decisions are "arbitrary" and "POV", and yet as a community we handle them reasonably well. There is absolutely no ground to assume we would not with the labels. Also, as said before, it is not up to you or me to force projects not to use labels, if they want to. You can make your argument in each project you are active in, but I can't possibly see how you would like to forbid projects you are not even active in from using labels on your whim/point of view/personal principle. Pundit | utter  03:30, 28 July 2012 (UTC)
 * We don't delete porn if it can be used in educational context (a combination that your words usually exclude). We have this more and more boring discussion, because of the accusation that the community wouldn't be able to handle this cases reasonably well. This fact contradicts your wording inside your last comment. Additionally you still don't recognize the difference between covering a topic (or to retain an image in a collection) and rating the subject itself as X, while X is some kind of non neutral arbitrary criteria. As long you don't agree to this difference, I'm not really motivated to continue the discussion, because we would never find a solution or an acceptable compromise.
 * PS: Yes i know that you dislike analogies, especially well established ones. But I'm curious why models from the "real world", that proofed workable and viable in daily practice, that stand in for sharing knowledge and/or treating every subject as equal as possible, suit the "unfiltered model" of Wikipedia so perfectly, while filters and advisory messages contradict them. -- ／人 ◕ ‿‿ ◕ 人＼  署名の宣言 08:22, 28 July 2012 (UTC)
 * Pornography is "the depiction of erotic behavior (as in pictures or writing) intended to cause sexual excitement" (Merriam-Webster). Under no circumstances we publish porn. Educational or artistic nude pictures are NOT porn. It is quite sad that you do not see this distinction and yet enter the personal filter debate. Again, whatever your personal point of view and beliefs are, you have no right to dictate to projects whether they should be allowed to use labels for their own purposes. This actually sounds pretty much like a fundamentalist approach. The projects have the right to decide for themselves. Pundit | utter  15:32, 28 July 2012 (UTC)
 * I'm well aware of the difference inside the definition, but inside this debates history everything close to nudity was already called "porn" by filtering/labeling supporters, no matter what intention it had. Thats why i used this wide scope wording, assuming that you would use the same definition. As i see now, you don't want to use it. Thats fine with me and i would like to see our ever raging "Commons dishes out porn" group to agree with the same definition.
 * I don't speak for the projects, neither am I in the position to enforce anything. That isn't my goal. I just tell you my viewpoint on filtering and labeling, how likely the implementations would be accepted by our diverse communities, readers and donors; and how they relate to our primary principals to create a encyclopedia as a community. From my perspective such a filtering or labeling (as proposed inside this discussion) is contra-productive to the mission of the projects, because it is working against the NPOV principle. Thats why i don't want to see the WMF putting any effort into such an effort, as long the proposals need to violate the policy. There are several cheap, easy to implement, inherently neutral proposals around, that got ignored all the time, while this discussion still circles around the category -> label -> advisory -> "next new word for the same thing?" proposals. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 16:00, 28 July 2012 (UTC)
 * Great, so we agree more or less with the definition of porn, with the fact that the community is good at deleting it, with the fact that the projects have the right to make own decisions. I also agree with you that WMF should not enforce any global solution not chosen by the projects. We disagree about particulars of solution, which of course is understandable. Pundit | utter  18:24, 28 July 2012 (UTC)

Yes we have no problem with deleting porn after the definition of Merriam-Webster, but it is said that we have a problem dealing with porn after the definition used inside the usual deletion requests on Commons, by the same people that write the deletion requests and argument in favor of deletion on whichever basis. Projects can also make their own decisions, but there are some rules that every project has to follow or has followed so far. One of them is the NPOV policy. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 18:46, 28 July 2012 (UTC)
 * Of course. And yet, I believe that each project should be able to make decisions about labeling on their own. NPOV is ultimately reached through consensus decision-making. Please note also that if labels are project-dependent, the Commons community's decision will not affect other projects. I think the problem of several users criticizing any nude picture would dissolve when more people are involved. Pundit | utter  19:04, 28 July 2012 (UTC)
 * I can't agree with your sentence: "NPOV is ultimately reached through consensus decision-making." Even if our or some other community is able to reach consensus, it is still the point of view of that community. This view doesn't equal neutrality, because being neutral does mean not to take the stance of any side. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 20:25, 28 July 2012 (UTC)
 * Fine, but then NPOV is only an abstract, non-achievable ideal. Pundit | utter  21:12, 28 July 2012 (UTC)
 * Of course NPOV is an abstract ideal. But that is part of the goal of this project, because we strife to get as close to it as possible. Thats why we argue about content and try to improve it. A major part of our work is to balance the arguments of different viewpoints, to represent them in the respective weight regarding the topic. You could also call it the decision to make the best compromise in representing arguments, but not to propose a compromise. It is because we strife for an ideal which can't be achieved we often rely on a compromise as an workaround, which should be very close to the ideal to call it a good compromise.
 * I see a filter as the opposite to a compromise or the ideal. He only accepts two answers. They are "yes" or "no". The room for compromise would be between both answers, which does not exist in case of a filter or labeling. A tag is present or not, even in 50:50 cases. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 22:05, 28 July 2012 (UTC)
 * Again, as I tried to point out many times before, this 0-1 decision is already made in case of porn. Clearly you realize that between porn and pictures of kitties there are also images which, while valuable, prove to be still controversial to the vast majority of users on some projects. Our decision, made exactly in the same mode as the 0-1 decision whether we want to keep an image or not, would introduce a layer of information to what we already do. But I think you know this argument already, since you decide not to address it, and I don't see how repeating everything all over can change this discussion :) Pundit | utter  23:01, 28 July 2012 (UTC)
 * As i stated many times already: The inclusion of topics or images inside our collection does not indicate endorsement of their contents. Thats the part of my argumentation you ignored so far and it is also the reason why we don't find a common level for constructive discussion. So i ask you: Do you agree with my view (shared by law, schools and libraries) that the decision to include or exclude a topic/lemma/image does not indicate endorsement or reluctance of the subject, while adding our point of view to the representation of the subject does indicate endorsement or reluctance of toward the subject? -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 08:36, 29 July 2012 (UTC)
 * @Pundit. The 0-1 decision is not and cannot be made for porn. Also NPOV is the wrong approach for a filtering system, or rather the only way to combine NPOV with a filter system is to be neutral between all POVs and enable any group of people who share a view as to what is offensive to filter out what they don't want to see (enable the creation of a filter rather than populate one as we as we can only do this if the responsibility for deciding whether an image offends them is put on those who opt in to the filter). The problem with doing this at a project basis is that our projects are mainly divided by language, not by cultural preferences re porn etc. So a project based solution will fail those who are likely to want it most - the people whose filter preferences differ furthest from the mainstream in the editing community for their preferred languages. At the same time it has all the disadvantages of a public filter - endless arguments over borderline cases, the impossibility of classifying all our images as porn or not porn when the world is divided as to how one defines porn, and the probability that we would be assisting others to censor and huge scandals over various religious or artistic works that would be "classified as porn by wikipedia". Better in my view to have private, personal filters. In fact I'd go as far as to say that the community is unlikely to give consent to a filter unless it is designed to work for as many as possible of those that want a filter and otherwise impact on as few as possible of the rest.  Ϣere Spiel  Chequers  09:53, 29 July 2012 (UTC)
 * Well, it is ironic that you say that 0-1 decision cannot be made for porn, and yet this is exactly what we do :) We remove porn (images intended for sexual arousal) and leave educational and artistic images, even if they include explicit nudity. It is the community's judgement. You speak of filters, I describe labels (the main difference is that labeling in the description is project-dependent, and users can use labels at their liberty). As you are surely aware, language and culture are not independent variables. There is a huge overlap of cultural norms in same-language groups (in other words, language is a reasonably good estimate of cultural sensitivity: people speaking the same language do not necessarily have the same sensitivity, but are much better aware of sensitivities of others speaking the same language). While I respect your view on personal filters, I want to remind you that these do not and will not solve anything for 99% of our readers. Also, please remember that "the community" as a whole cannot forbid projects to introduce labels. It is the projects, which decide about their rules, not the vast collective of all of them. It would be outrageous and somewhat dictatorial if "the community" understood as everybody in Wikimedia forced solutions on separate projects, and in the name of theoretical liberty limited the actual one. Pundit | utter  13:55, 29 July 2012 (UTC)
 * Let me conclude: You want your labels, no matter what someone would say to you or which arguments he would present.
 * I write this conclusion, because i lost interest in this discussion. I asked you multiple times to answer some simple questions, but you continued, ignoring them, giving no answer, but repeating the same unproven (i asked for proof, still waiting) arguments over and over again. At this basis i consider a further exchange of arguments as useless and the discussion as a fail. It will only run in circles. We can continue, if you react to my questions and break with the circle. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 14:27, 29 July 2012 (UTC)
 * Lol, what a neutral conclusion indeed :) I could remind you, that you refused to address my concerns, refused to talk to OTRS people (any would easily confirm that the issue of shocking images is an often recurring topic), and so on, and so on. I am not going to summarize your position, I'll just repeat mine: projects cannot be forbidden to use labels if they chose to. Just as global filtering is wrong, so is the dictate to projects what they can and what they cannot use for themselves, if they choose to. Pundit | utter  19:08, 29 July 2012 (UTC)
 * The distinction between "porn", "art", and "education" is illusory. Pornography is at its heart a scientific endeavor, as any little child first interested in "playing doctor" could attest.  Science is at its heart pornographic, as the crude cabal of medical students and "Resurrection Men" who charted out the human body could attest.  Art does not even make a pretense not to be "prurient", if by that you mean, attractive in a sexual way, though not often does it aim to itch.  And pornography, though it is usually very badly and inartfully executed, is by far most effective for its designated purpose when it is produced with a true artistic sensibility, because the essence that fires the mind in daydreams of love and lust is not scientific, but spiritual; truly pornography appeals to the dream of love, and therefore is only effective to the degree that it is able to conceal within its self an honest rendition of the spirit of innocence. Wnt (talk) 02:59, 30 July 2012 (UTC)
 * You are right. It strongly depends on the definition. If we go after the definition of Merriam-Webster, then we will have a hard time to find any porn. If we choose some other definition then we would have to agree that porn, art and education are not mutually exclusive, creating a blurry border line. I believe that the discussion is unnecessary exaggerated.
 * Just to throw in some facts:
 * The Chinese and Russian Wikipedia have no problem to feature and represent "controversial content" on the main page. ZH (main page entry 2012-02-16), RU.
 * The Arabian Wikipedia featuring images like this two. Curiously both images had a very hard time at EN with the voters calling for "never show it on the main page" or "can't support as long Howcheng would put it on the main page".
 * -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 08:42, 30 July 2012 (UTC)
 * This is exactly why we should have project-based labels, and not global filters. Pundit | utter  13:35, 30 July 2012 (UTC)
 * That the solution, if there is a good solution, should be project based is out of question, but if it does need labels is another story (the long story above). -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 14:33, 30 July 2012 (UTC)

Content neutral implementation
I think the foundation should only get involved in providing the facility but it very much should not be involved i choosing content to filter. I believe it should be left to the community to set up filters and the most popular ones be made easily available.

The work I see is:

Provide a means of generating a filter with a short description. The generated filter would just be a list of files and the generation could involve other filters, categories, files to include, files to always allow. Which filter is used by a user would be private. Filter generator files would be subject to the same protections as files and the most popular would probably be only editable by admins. It would be an implementation problem whether the generated filter was only updated if say older than a day or if a dependency changed.

The most popular filters would be made easily selectable by users, it might stop controversy if this was automatic! Less popular one could be searched for using the description or could be used in links.

The filtering would normally be via javascript downloaded to the browser but if that is not supported then it should be done on the server.

I would allow the filer to stop any file, whether media, article or external. An easily identifiable mark should always be put in showing the removal and be easy to override. I would allow schools to set a default preference for a range f ips but if they wanted to absolutely stop override they would have to use an outside facility or something like Wikipedia for schools. Dmcq (talk) 00:51, 16 July 2012 (UTC)
 * I believe such solutions are overkill and are not possible to gain consensus for. I like your concept in principle, but I'm in favor of doing something very straightforward, industry standard.  I am proposing that the Foundation could staff it if necessary only because I think getting to a 99% effective industry-standard solution is incredibly easy, and most complications are attempts to solve edge cases that no one else in the industry is even attempting to solve.
 * A javascript show/hide solution makes it really easy for category errors to be dealt with by end users in a very low-trouble way.--Jimbo Wales (talk) 12:13, 16 July 2012 (UTC)
 * A bit of overkill is necessry for a number of reasons


 * A very simple solution does not allow for instance Muslims to hide images of Mahomet easily and it would be good to be rid of that problem.


 * It would be better for the foundation not to be associated with questions of exactly how much breast is okay or whether cartoon violence should be included and differences between different countries in their ratings.


 * What you are talking about is specific filtering that will cause controversy. IF the foundation gives the problem to the community then it's it is peoples own choice if they don't wish to unwittingly see pictures of Xenu. Do you really think it would be harder to get acceptance of something where the foundation says what is acceptable or not in the workspaces they define? Do you really want to decide on amounts of breast shown and yet say you are not interested in the Mahomet business? You're much better off just providing the technical and moral support rather than getting into the morass of specific arguments. 13:05, 16 July 2012 (UTC) — Preceding unsigned comment added by Dmcq (talk • contribs)

1. I am unaware of any user-controlled website filter from any major Internet provider that allows people to turn on and off a preference relating to pictures of Mohammad. I conclude that our version 1 need not solve a problem that no one else has bothered to solve. Making a system try to do too much is a constant temptation, and given our resource constraints and need to get widespread agreement, proposing a system to try to do too much is a sure recipe for getting nothing at all done. Don't make the perfect the enemy of the good!

2. I think it is better if the community provides clear guidance on what NSFW means, but I note that it is industry standard for such systems to be managed by employees at various firms. I see no real barrier to us doing the same thing. I am actually completely indifferent to who does the technical job of going through the categories and flagging things as NSFW for the primary reason that because any image hiding for NSFW purposes will be one-click to turn on or off in all cases at all times. This is not an effort to censor Wikipedia or prevent people from seeing whatever they want to see. It's an effort to give people reasonable control over their own experience of Wikipedia.--Jimbo Wales (talk) 16:18, 16 July 2012 (UTC)
 * NSFW means very different things in Mississippi, Madrid and Medina. If we were a firm operating in and for the residents of one of those cities then I believe we could agree what NSFW was fairly easily. But we are avowedly global and that means we need a more complex approach. Rejecting the idea of allowing Moslems to opt out of seeing photos of Mohamed because other major Internet providers don't do that sets a simple precedent - you intend to base the filter on western mainstream concerns.  Ϣere Spiel  Chequers  17:57, 16 July 2012 (UTC)


 * Go around and flag things as NSFW, I don't think you can have listened at all. You are just dooming the whole business if you start from that point of view. Something useful can be made without too much effort that provides a useful facility and brings in more readers and it need not require a continuing commitment by the foundation. But if you want it done the way you're going you'll get peoples' backs up badly. If you can't be bothered working with the community and dealing with the actual problems why not just do it as a separate package and you can avoid any hassle from the community? Dmcq (talk) 21:54, 16 July 2012 (UTC)

Code to re-use for this
Would submitting a patch to mw:Extension:Bad Image List adding a user preference to add one or more URLs with arbitrary media files to block instead of using only the centralized list require the approval of the community or just the developers? Line 17 here performs image censorship in the centralized, top-down way that the community and board rejected, so a patch to add a distributed filter list should be in line with community decisions, right? Maybe whatever fetches the blacklist should use a special HTTP user agent string so the best filter sources don't give away lists of porn. Also, this is completely pointless. Early exposure to pornography is associated with a decline in sex crimes, not to mention witch hunts. 71.212.249.178 (talk) 00:54, 16 July 2012 (UTC)
 * Whether it is beneficial or not is not the point. It isn't our job to cure the problems of the world. What's being attempted is allowing access to Wikipedia when people might otherwise avoid or censor its use but still remain faithful to the no censorship policy. And you must admit a 'do not press' sign is a strong magnet for children! ;-) Dmcq (talk) 01:08, 16 July 2012 (UTC)
 * Good point. 71.212.249.178 (talk) 01:21, 16 July 2012 (UTC)


 * http://www.php.net/manual/en/features.remote-files.php (the http_get in comments there using @fsockopen is decent and can set arbitrary headers including user agent so as to prevent listing all the porn images instead of just an ordinary web page with instructions when a blocklist url is used in a browser, and it supports some redirects, although using $buffer as a global is bad) Alternatively:
 * http://www.php.net/manual/en/function.file-get-contents.php can also set HTTP headers with stream_context_create
 * http://www.php.net/manual/en/filesystem.configuration.php#ini.allow-url-fopen as in ini_set("allow_url_fopen", true); just to be sure. (I have no idea whether Foundation production wikis prefer to leave that set to false for security or other considerations.)

$wgHooks['GetPreferences'][] = 'wfSendToIdenticaPreferences'; ... function wfSendToIdenticaPreferences ( $user, &$preferences ) { $preferences['SendToStatusNetActivate'] = array(           'type' => 'toggle',            'section' => 'misc',            'label-message' => 'sendtostatusnet_activate',        ); ...       $preferences['WikiRootUrl'] = array(            'type' => 'text',            'section' => 'misc',            'label-message' => 'wiki_root_url',        ); ... } ...    $url = $wgUser->getOption('WikiRootUrl'); ....
 * http://www.mediawiki.org/wiki/Manual:$wgDefaultUserOptions adding options to user Preferences
 * http://www.mediawiki.org/wiki/Extension:Send2StatusNet useful for showing how to put text field options in Preferences:

That extension has a reasonable lightweight i18n ("messages array") model. It also uses libcurl, which is probably on production everywere, but not part of the recent default PHP builds.


 * http://en.wikipedia.org/wiki/Help:Gadget-ImageAnnotator a production enwiki gadget which deals with images -- we could do something like this with black rectangles or pixelation in the future, maybe, if people think that is better. 71.212.249.178 (talk) 19:22, 16 July 2012 (UTC)

IP Editors
How will this work with IP editors/readers? For those without an account, will they inherit the settings of the previous user of the IP? That would allow an IP reader to opt in future readers without their agreement. Better would be to use a session cookie that will reset the settings when the browser window closes. Those wishing for long term filtering would still have the option to register an account to enable it. Monty 845  01:27, 16 July 2012 (UTC)
 * Because it is one click to turn it off or on, I don't think it really matters much. I'm totally happy to follow whatever Google Images or Flickr do about it.  I would say that a session cookie is a fine idea for anonymous users.  But if the cookie were longer lasting, that would not be a very big problem, because it would be one click to see anything you want to see, and one click to turn off filtering altogether.--Jimbo Wales (talk) 12:15, 16 July 2012 (UTC)

Saving state for logged-out users across sessions presents a problem. Dynamically assigned IPs will probably want to use cookies for this, or will they? If it's easy to toggle, should it be associated with the browser or the IP address, or some combination? (Some related discussion in code comments here.) Associating the filter state with an IP address would mean one person could turn filtering on for a whole house or school, for example, but anyone could turn it off. 71.212.249.178 (talk) 01:26, 17 July 2012 (UTC)
 * If you use an internet café any cookies set up by the previous user should be removed and not affect the next user. If it is some place that can't be bothered to do that then that's their business. Since it would be pretty obvious some images are blanked and how to set the facility I can't see a big problem even in that case.


 * However it my be worthwhile to have a way for an internet cafe or library or school set a persistent default. They could use an image blocker but Wikipedia is widely enough used for some special facility to be added even to such software. Dmcq (talk) 10:12, 17 July 2012 (UTC)

What should a blocked image look like and other pressing implementation details
71.212.249.178 (talk) 19:24, 16 July 2012 (UTC)
 * 1) Blank? White? Black? Pixelated? Collapsed? File:Do_Not_Enter_sign.svg?
 * 2) Have only the most offensive rectangle of the image censored and keep the rest of the image? [deferred per principle 5.]
 * 3) Keep the thumbnail caption displayed or not?  Append, prepend, or replace with an explanation of the filtering? (With links? To what?)
 * 4) What happens when you click on a blocked image? Link to instructions for unblocking? Toggle the preference off? Link to Interlude of Youth?
 * 5) Given that IP users don't have Preferences, should administrators be able to turn on filtering for IP ranges such as schools or Islamic nations? If so, how should they select the blocklist(s)?
 * 6) How should IP users turn on filtering? [essential per principle 3.]
 * 7) How should IP users turn off filtering? [essential per principle 3.]
 * 8) More than one blocklist can be used in any combination at a time, right? (Just a list of urls)
 * 9) Do we want to be able to offer the most popular and/or official blocklists as checkboxes instead of urls?
 * 10) Should there be a table expanding easy-to-remember tokens (e.g. "official") to the corresponding lengthier URLs (e.g. http :// wikimedia.org/blocklists/unified)?


 * In accordance with principles 2, 3, and 5, here are the answers which I believe would be the least contentious and make implementation easiest:
 * The ISO prohibition sign: [[Image:ProhibitionSign2.svg|15px]]
 * Defer until after testing; then decide.
 * Prepend an explanation of the filtering when a thumbnail caption is available, with links to a Special page with further detail and instructions for deactivating filtering.
 * Clicking on a blocked image should link to the further detail and instructions for deactivating filtering.
 * This is pretty important I would think, but it should be deferred in accordance with principle 5 for now.
 * A "hide controversial media" link on the upper right of every page next to "create an account" and/or "log in"; using the session cookie to store this information.
 * by clicking on any blocked image or link from a thumbnail caption and following the instructions, or by clicking on the link on the upper right which will be "show controversial media" when filtering is on -- all of which would link to the same Special page in (3) above.
 * yes
 * One unified official blocklist should be offered as a checkbox. After implementation and testing, the top three most popular other blocklists should also be offered as checkboxes.
 * Yes; syntax errors for tokens not found in the table or URLs not returning blocklist format data should be indicated as such to the right of the text field (on Preferences or the Special logged out instructions/filter configuration page mentioned in (3) above) where such tokens and URLs are entered. 71.212.249.178 (talk) 01:26, 17 July 2012 (UTC)
 * I'd prefer the caption/alt text in simple lettering and "show image" and "about filtering" buttons on a blank field in the same size and place as the blanked thumbnail. The ISO prohibition sign would be unnecessarily distracting.
 * See 1.
 * See 1.
 * Too complicated and possibly controversial. Maybe later.
 * Like Google: a "Safe search on/off" button at the top of each page.
 * Ditto.
 * Too complicated for this exercise. A good idea for later, though.
 * Ditto.
 * Ditto. --Anthonyhcole (talk) 02:33, 17 July 2012 (UTC)
 * Ditto. --Anthonyhcole (talk) 02:33, 17 July 2012 (UTC)


 * I agree, your answers are better and make the task easier. Although "show/hide controversial media" may be substantially more accurate than "safe search." 71.212.249.178 (talk) 03:45, 17 July 2012 (UTC)

Motivation tag
According to the pseudocode I have sketched out on some napkins, the answers above amount to about two pages of PHP context diffs, or 3-4 pages with comments when implemented with the suggestions above. However, I now have that all-too-familiar feeling of utter pointlessness which has been the reason nobody has bothered to do this yet, so I gladly announce I will be working on other things until the feeling goes away (e.g. with sleep). And I would be ever so happy if someone getting paid for PHP programming takes over here, or even an enthusiastic volunteer developer. Or both, maybe with some kind of a junior-aprentenceship for making the Foundation edit lists of porn. Darn it, it happened again. /me visualizes Jimmy and Larry sitting down for a beer in total unspoken forgiveness Okay someone make it real if you really want it. 71.212.249.178 (talk) 21:51, 16 July 2012 (UTC)

From the subject-space page
(Copied from .)
 * I haven't been following this debate at all, so tell me if I'm just saying things everybody knows. Has it been proposed that "Not safe for work" or "Not suitable for children" tags and the like could be put on the image's description page by the person uploading the image at the time of uploading?  These would presumably be subject to editing and discussion by other users. Michael Hardy (talk) 22:48, 15 July 2012 (UTC)
 * If that happens, it needs to be phrased such that it can't be construed to constitute incorrect legal or medical advice. 71.212.249.178 (talk) 00:46, 16 July 2012 (UTC)
 * Both "not safe for work" and "not suitable for children" are entirely subjective, and thus not appropriate for a categorization or tagging scheme. Is an anatomical drawing of a penis suitable for children? Is an image of a woman in a string bikini safe for work? It would be an endless battle that would do nothing to actually improve the projects involved. Monty  845  19:55, 16 July 2012 (UTC)
 * These were among the reasons why the previous category based proposal was so contentious. Yes if we go for a system based on our categories we will inevitably come under pressure to follow the route that Flickr took and require our uploaders to classify their images as contentious or otherwise. Which is a practical and ethical difficulty for the GLAM community and a philosophical difficulty for others. On a global project we cannot seriously expect to ask our uploaders to classify their images not in terms of whether they would be safe for work in their culture but whether they would be safe for work in the US. Unfortunately Jimmy seems to have restarted the debate with a version of the scheme that the community was most opposed to.....  Ϣere Spiel  Chequers  20:19, 16 July 2012 (UTC)
 * On the contrary, he asked for the least contentious proposal. 71.212.249.178 (talk) 20:31, 16 July 2012 (UTC)
 * The least contentious proposal would be to add a collapsible header (like a system message) on top of every page that Wikipedia is not censored and readers have to expect controversial content due to the fact that Wikipedia tries to depict all common knowledge and that knowledge can be disturbing or unpleasant. But that could be wrong since we have a discussion like this a the moment instead of continuing to reach the project goal. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 21:21, 16 July 2012 (UTC)
 * How are you measuring contentiousness? MediaWiki:Bad image list exists today, its architecture has been rejected by the community and the board, and it strictly contradicts your proposal. How would you put it in line with community standards? 71.212.249.178 (talk) 21:29, 16 July 2012 (UTC)
 * The bad images lists intention is to stop vandalism in the case that an image is repeatedly inserted in various places where it doesn't belong. That is the initial and true intention of this list. But i have to admit that this list, today, right now, contains various images that were never used for vandalism, which is misuse of the feature as a tool to rate content as inappropriate. A sad case. Thats why I'm convinced that any kind of additional centralized filtering and tagging will do more harm then good. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 22:10, 16 July 2012 (UTC)
 * I'm convinced decentralization is correct, too; not just for filtering, but for greater income equality, general libertarian freedom, etc. But then the question becomes: How do you prevent the churches from destroying biology and radiochemistry education? I'm worried that isn't off-topic because, for example, what if someone made a list which censored all the evolution-related media? Would we even be able to see that in the logs if it happened? 71.212.249.178 (talk) 22:24, 16 July 2012 (UTC)
 * As long we don't actively contribute to the creation of such lists, joining isn't mandatory and the tagging is private, it should be fine. But Jimbo was talking about a quick "one hat suits everyone" solution, with public tagging - a special categorization for this purpose -, which is interfering with the projects missions. I hope that you did not misinterpret my words. From my point of view any needles tagging/categorizing (like "nude x in y") is bad if it doesn't serve as an directional aid for the user/reader that is interested in a particular topic. I consider such kind of tagging as "evil" and share the view of the ALA and similar organizations like Elisad, LIANZA, vdb and so on. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 23:15, 16 July 2012 (UTC)
 * We already categorize images. If you aren't actively contributing to decentralization, then you are inactively contributing to the status quo of centralized censorship. 71.212.249.178 (talk) 23:30, 16 July 2012 (UTC)
 * We categorize images in the means of "directional aid" and not in the means of "prejudicial labels". Please read this short introduction to know the difference between good categories and bad categories. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 03:16, 17 July 2012 (UTC)
 * That is well worth reading. 71.212.249.178 (talk) 03:53, 17 July 2012 (UTC)

Old thoughts from a MediaWiki developer
mw:User:Simetrical/Censorship --MZMcBride (talk) 19:52, 16 July 2012 (UTC)
 * That is useful to read, but it amounts to much more work than I'm contemplating, and there are plenty of arguments all over the place about why using the category system directly will likely be more trouble than its worth. If blocklists are URLs and not categories, then the blocklist publishers, whether official or third-party, can certainly use API or database category queries (along with recent changes upload/category edit feeds) to seed their lists and keep them current. On the other hand, maybe categories should be able to be placed in such lists, but then the problem becomes: How does an extension running on enwiki enumerate a set of Commons categories? You wouldn't want every index.php process from a school or Islamic country to have to do separate API calls from the enwiki servers to commons just to enumerate tens of thousands of filenames. You want to cache those anyway, so why not keep them in a flat file on the other end of an ordinary URL? (Or maybe not so ordinary in that it would require a special user agent string to actually see the text flat file list instead of a web page with instructions for use which is what you would see when you plug the same URL into a browser.)
 * This would require the Foundation to maintain such a cache (or preferably multiple such blocklists for sex, violence, spiders, pictures of Mohammed, etc.) which is okay per principle 4 and can probably be done mostly automatically. If the Foundation (or third parties) chose to publish their current blocklists by placing the images in them in categories, then everyone could easily review them, which would probably make Fox News, too.
 * Maybe there could be a "weary of censorship" mode where a tiny ProhibitionSign2.svg sign appears superimposed over the corners of images which would have been blocked so that users can review the extent and characteristics of censorship choices. Of course then wikimedia-l would probably fill up with personal taste debates. (On second thought, that would just codify current practice.) In any case, this can be deferred in accordance with principle 5. 71.212.249.178 (talk) 20:49, 16 July 2012 (UTC)
 * It might be a good idea to have a complicated way of generating such a list, but I'd have though any actual run time implementation would in effect just list files. I would hope the actual run time for people using javascript could be kept practically identical to as present except for some extra javascript and the downloaded url list. How many files are we talking about do you think? I would be interested to know exactly how big a list of urls would be if compressed to still allow quick lookup but distinguish between files - it would be reasonable for instance I think to just put in a short checksum of the end of a file name after the first character that distinguishes it from other files and the start before that character would also be common so each url would only need a short string. Also how many people don't have Javascript and so such a facility would require more server work?


 * And by the way what's your favourite way of doing something like this or what cheaper facilty were you thinking of? Dmcq (talk) 22:14, 16 July 2012 (UTC)


 * Start with the files in the existing bad image list, I guess, and start adding whatever else the critics complain about. I'm not contemplating adding any kind of Javascript "[show]" which pulls the image name through NSFW-detecting firewall logs even if the user doesn't look at it. You already need to send all the information in the existing Bad Image List as it is, so adding extra fields to a flat file text cache is no problem. There are checksums in the TCP layer, but if you mean a version number or timestamp, then I agree, and if you are suggesting putting an abbreviation token in the file as it is fetched to go on the token table, then I agree with that, too. That way, you need only retrieve the first few lines of the file, and can close the socket if it's identical to what is cached locally.


 * My favorite way of doing it would be so that I wouldn't be the one who made looking through porn part of Sue Gardner's job description. If you can round up volunteers on IRC #mediawiki and #wikipedia I bet you can find dozens of people who would do it for less than what I'd charge. 71.212.249.178 (talk) 23:46, 16 July 2012 (UTC)

keep it simple
Something that is simple and can easily switched on or off would be a button: "Images on/off".

Whoever wants to read an article, let's assume, about the human penis without any picture of a human penis, switches all images off before going to Human Penis- done. After that he wants to read something about the iPad: Images back on, done. Even more convenient (although I have no idea how to program that): Implement a right-click-option for any blue link: "open without images". That would allow context related DIY-filtering satisfying all needs without any categorisation or intercultural struggling. --Superbass (talk) 02:35, 17 July 2012 (UTC)
 * I agree that would be superior, but there is no cross-platform way to add right click behaviors. We could make just a button to toggle all images on and off with the MediaWiki session cookie. Would that fill the bill? 71.212.249.178 (talk) 03:38, 17 July 2012 (UTC)


 * Controversial content/Brainstorming --MZMcBride (talk) 03:50, 17 July 2012 (UTC)
 * Not only could that turn out to be entirely sufficient, but it's worth doing first as a subset of the tasks described above. Yes! 71.212.249.178 (talk) 03:57, 17 July 2012 (UTC)


 * This could also be very useful for mobile phones to cut down costs and only look at a picture where it seems it might actually be helpful. Dmcq (talk) 10:20, 17 July 2012 (UTC)
 * Please give us an all images off option as soon as possible. Having one would be immediately helpful. 76 Strat String da Broke da (talk) 15:43, 17 July 2012 (UTC)
 * Don't most browsers already have this option? Powers T 15:14, 18 July 2012 (UTC)
 * Yes but it is normally two or three levels down, for instance on Chrome you click on the spanner then settings then Advanced then Content settings then Do not show any images and you can associate that with particular websites. Not exactly straightforward. What we'd be talking about here is probably two clicks - filter options, hide all images as the initial filter option. Plus hopefully just one click to display the images you actually do want to see. Dmcq (talk) 15:47, 18 July 2012 (UTC)

Does someone want to try to come up with the smallest patch possible to the Bad image list extension which toggles all images? 75.166.200.250 (talk) 21:01, 19 July 2012 (UTC)

Directing Wikimedia Foundation staff
Hi. I've moved the following text to the talk page for discussion:

I don't believe there's any structure in place currently that allows a Board member to direct staff in this way. Perhaps there could be, but at the moment, such powers simply don't exist, as far as I'm aware. The Executive Director reports to the Board as a body; individual members (even the Founder seat) can't commit staff resources like this, as individual members have no direct control over the staff.

I suppose as a general principle, not worrying about staff resources could be included, though I think it greatly discounts the (high) cost of human resources here.

(There's also the much larger issue of staff getting involved in editorial policies such as this. I believe such a direct intervention by staff in this context would be unprecedented. It would be viewed as paid editing in the best light, and paid censorship in the worst.) --MZMcBride (talk) 05:26, 17 July 2012 (UTC)
 * Someone mentioned that the remainder of the resolutions in place (the two of them taken together) still authorizes direction for this effort \o/ but not the previous prototype with (5-7?) categories. 71.212.249.178 (talk) 06:23, 17 July 2012 (UTC)

WHY categories?
I don't understand the insistence on using categories, which are made on Commons for a different purpose, to sort content, as a means of "filtering". A category is a list of articles. So why not have one individual user go through the category, either with minimal effort (copy and paste), or reviewing each entry carefully, then create the list of names as a text file in his userspace, and allow others to transclude that into their own filter blacklists? Per User:Wnt/Personal image blocking. Wnt (talk) 08:06, 17 July 2012 (UTC)
 * Agreed. I'm quite happy for categories to be used in setting up such lists if anyone can set up a list. However having categories marked with a limited set of centralized approval rating tags strikes me as a barrel of trouble. Dmcq (talk) 10:26, 17 July 2012 (UTC)


 * If you're suggesting readers have to create their own filter blacklists, that fails #3 on Jimbo's list of specs. But perhaps I've misunderstood you. --Anthonyhcole (talk) 12:39, 17 July 2012 (UTC)


 * To "turn this on" a user could create a new file User:(your name)/image-blacklist.js with the contents .  If there is some way I don't know about to automatically fill in the user's own username, this can be done with two clicks; otherwise it takes a bit more effort, but even so, it would be easier to add a feature to automate this action than many other ideas. Wnt (talk) 13:25, 17 July 2012 (UTC)


 * Special:MyPage/image-blacklist.js is the syntax you're looking for. Using JavaScript subpages in this way is patently retarded, though. --MZMcBride (talk) 15:18, 17 July 2012 (UTC)


 * If there's any other way to mark a page as uneditable by anyone but yourself, I'm all ears - but the .js suffix would allow it to be done. Otherwise we might have editors complaining that vandals blanked their blacklist and filled their userpages with vulvae to harass them for their censorial beliefs.  At least this way they wouldn't see such images. Wnt (talk) 19:18, 17 July 2012 (UTC)

Classes of people
A key objection for me revolves around questions of groups of people. Images of individuals from particular ethnic minorities and castes might be a helpful example. Yes, there is something pedantically "neutral" about a system that allows one to avoid photographs of Tutsi but not Hutu or the other way around. But the effect of that allows either on human beings is distinctly non-neutral. So might be the case for particular castes, religions, races, or for photographs of transgendered individuals (who many people find inherently offensive).

In some of those cases, there isn't even pedantic neutrality from a category system, as a category exists for the minority but not the majority--we have no category for Gentiles, or cisgender people.

The idea that images of particular classes of individuals are "not safe for children" is commonplace. Flexible filtering systems promote the point of view that such distaste is acceptable through availability and easy sharing. Simple five-button systems promote this point of view through subjective application: Same-race vs. interracial kisses, etc., are found disparately "offensive to children", so are platonic kisses based on the sex of the people involved.

In either framing, filtering ends up being at odds with any basic presumption of humanity of our readers. In the case of a genocide, it might even seriously contribute to harm against living people.

Perhaps this danger can be guarded against with a simple restriction or prohibition, simply removing categories of human beings from any use in image filters (although you'd need to make a few exceptions there, too). However, the history of the so-called neutrality of image filtering is not cause for optimism. While I have no doubt about the good faith of those pushing for image filtering, I remain wary of the consequences. --j⚛e deckertalk 16:00, 17 July 2012 (UTC)

I've got a very long list of images already...and a method that works...but it requires a Firefox add-on
Hi! I've taken a previous stab at this problem (detailed at User:Bob the Wikipedian/Virgo). This uses the free AdBlock Plus extension for Firefox (unfortunately, it's browser-dependent and requires an add-on!). It basically comments out all images that show up in the filter list, and exception rules can be added (for example, some images that are generally inappropriate might be a wise choice for anatomical articles). My list of censored images is naturally tailored to my own religion-- and I'd recommend we have multiple lists for different cultures. Of course, I've tried to host this list for others interested in using this filter but have had my account suspended from the host website for "violating their terms of use" with links to obscene images. Anyway, it's something that I know has worked for me quite well, and I'd love to see something everyone can use similar to this. Bob the WikipediaN (talk • contribs) 17:45, 17 July 2012 (UTC)


 * I don't think I'll actually test the program myself, but provided other people can confirm this works, this is an extremely important and credit-worthy milestone. From now on, people should not say that image filtering is unavailable from Wikipedia - only that it is user-generated, not easy to install and doesn't have all the content options everyone wants. Wnt (talk) 19:24, 17 July 2012 (UTC)

The Booru model, to learn/copy from
Figuring that, since there's a whole bunch of *boorus out there that deal with a lot of the same problems we have (User-defined content and moderation, extreme diversity of content and user desire to view said content, over a million images, etc), I figured I'd lay out the tools that sites like Safebooru use to deal with filtering images:

Images on a *booru can be given multiple tags, roughly equivalent to our categories, but tending more towards inclusion. Users have individual filters where they can choose not to view certain categories of images, and some sites have a global/default filter that is used for all users that have not already modified their filters.

Most sites use exactly three global content ratings for sexual content: "Safe", "Questionable", and "Explicit". That the tags are self-descriptive is more important than whatever rules could be found on a page somewhere (it took me a while to find Danbooru's rating guidelines, for example. These ratings are typically uncontroversial because it's understood there's an audience for each, i.e. being rated explicit can make it easier to find for some users, even if it's hidden by default.

Filtered tags can be hidden (removed from view entirely) or spoilered (thumbnail replaced with a placeholder image, clicking the image reveals it) On the page for a spoilered image, it may say something like: "This image has been hidden upon loading this page because this image is tagged with the tag _______ which is spoilered in image listings by default. To un-spoiler content with this tag, edit your content settings . Alternatively, click the image above to reveal the image just this once." (Blue text links to the appropriate user preferences page or login page). Often different spoiler thumbnails are used for different content tags - for instance a thumbnail for hiding explicit images would actually say "Explicit" on it.

One site I know of, Derpibooru (a My Little Pony *booru that hides explicit and grotesque images by default), puts at the top of all image lists "X tags hidden" which links to the user filter, clearly communicating that, yes, some images are being hidden by default and you can turn them on/off if you *really* want to.

I'm not suggesting we turn Commons into Wikibooru, but I'm putting this here as an example of a model that works for both SFW and NSFW communities. Nifboy (talk) 17:59, 17 July 2012 (UTC)


 * I like the Booru concept, but i don't think that it suits Wikipedia or Commons. Most Boorus have a very specialized audience and mission. This simplifies the task significantly, since the tagging and viewing is done by users with very similar backgrounds and interests. But this isn't the case for the multicultural Wikimedia projects, which have the basic mission to wrap up all human knowledge. We have various interest groups with very different opinions on what is acceptable and what isn't. Thats why a simple "one hat suits everyone" approach doesn't work well, while the Boorus have not such big problems, even so they also have their edit wars about borderline cases. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 23:08, 18 July 2012 (UTC)
 * Thanks for that feedback. I'm assuming that contentious borderline cases are pretty much an inevitability of the simplistic filter approach. Of course if someone can produce an example of a site with a global audience and a similar diversity ethos to us that has managed to implement such a simple system in a multicultural way, then maybe we should resurrect the Simple NSFW filter idea. In the meantime I for one will continue to treat it as a blind alley one has to extricate oneself from before you can try to develop an image filter that would be acceptable to the community, and indeed one that would work for us as a project.  Ϣere Spiel  Chequers  05:03, 19 July 2012 (UTC)

Expanding the Commons porn collection with effective indexing
Joe Decker makes a good point above with his "classes of people". Minorities like Commons:Category:Ngbandi people might be targeted by a filter because one or two of a few images display female breasts. But, to be fair, people would not and have not settled for this, and a solution is possible. Instead of having an image like File:Sango girls at Banzyville-1905.jpg which is categorized according to non-sexual features, we could ensure that every Commons image that would be prone to a filter is categorized like File:TweiMadchen vonSangaflussCentralafrika.jpg, as "nude standing girls", "nude sitting girls", "ethnographic female toplessness".

Now, some may say that by doing so, we fall from the erudite standard of a 1950s National Geographic reader, who simply accepted that different people had different fashions, even though the magazine might eventually end up in the sticky hands of a dreamy teenager. Instead, like the teenager, we objectify women; our learned response to legitimate cultural photography becomes "Oooh, look at the boobies!"

But I think that will be more persuasive to you is that maybe creating immense categories full of "nude girls" on Commons isn't the best way to reduce the "porn" perception. You get one means, an unnecessary means, of creating a content filter, at the cost of enlisting, indeed I suspect eventually dragooning, volunteer editors in an unwanted crusade to label every Commons image in terms of what body parts a person might wank to. Wnt (talk) 12:10, 18 July 2012 (UTC)


 * I wish Jimbo would give up the idea of top down western categorizing and support community sourced ones. That way this sort of question becomes irrelevant as far as a page like this is concerned. You choose a filter that seems to say the right things as far as you're concerned and if you're still unhappy and can't get just the right one you can always start up a mini project to create one with the sort of standards you approve of.


 * As to porn, Wikipedia is supposed to be educational. There is a problem with some people exploiting t in stupid ways but overall the porn business is a categorization people apply rather than something intrinsic in the images. The filter should be a personal filter to cater for the people's wishes not a filter of Wikipedia because it has problems that can't be dealt with otherwise. IF there are problems in Wikipedia they should be dealt with irrespective of whether people have personal filters. Dmcq (talk) 14:46, 18 July 2012 (UTC)

Who decides?
Most image filters and/or rules about what is and isn't acceptable on sites comparable to WP are very strictly limited by ToS wording or similar. This type of filtering would be something I would happily support.

But I'm concerned that discussions about a filter for Wikipedia tend to assume something very different and basically experimental - a system that is entirely community and user controlled. IMO, this was the central with the previous proposal. It didn't limit itself to content which, for example, reasonable people might want to shield their children from. Instead, it proposed a tool of censorship or self-censorship which was, in principle, open-ended.

What might be put in place to ensure that images such as these, , , , , are not capable of being filtered?

I don't think that it is good enough that people will be able to undo the filter, because categorising an image for filtering in the first place sends out a strong moral message about its contents, particularly if that categorisation ostensibly has the backing of a community. Formerip (talk) 01:29, 20 July 2012 (UTC)
 * Well your last sentence says it for why the community has a problem with a central filter. And there is a very strong no censoring ethos. The community simply will not as a group back any decision that particular images are morally one thing or another. You're talking about a very major loss of good editors if that was imposed. Censorship of any one thing is seen as a thin edge of the wedge to having things like state control of the news. That's why they would far prefer that anything like that be done outside of Wikipedia altogether. If there is going to be any support for filters in Wikipedia it has to be for purely personal decisions, and it would help if it ameliorated or solved real problems on Wikipedia like what some see as profane images of their religious figures. Fixing problems like that would help extend the reach of Wikipedia to new readers. Yes personal filters would allow creationists to avoid seeing anything about evolution if they didn't want to but it must be better than having them get all their information from Conservapedia and Fox News Dmcq (talk) 11:20, 20 July 2012 (UTC)
 * So far the community has been acting with common sense. Crowdsourcing to people is usually smarter than any scripted filters, I'd actually believe it should work just fine. It has worked, e.g. for deleting porn and yet keeping art, so even the examples you give show that the community is reasonable in its judgments. Pundit | utter  13:45, 20 July 2012 (UTC)
 * Did you ever take part in deletion or inclusion discussions of controversial images? If you did, then you must have noticed that most solutions are one-sided (one happy side and one disappointed side). Sometimes you will find a compromise, but over the time it is an exception and not the rule. It is easy to find a compromise if not many people are involved, but it gets significantly harder with any additional opinion. Such discussions do not stop with or without an filter. The filter just breaks it up into two parts, while the first part (editorial judgment) remains as it is, the second part is adding an additional layer of personal preference. It does not prevent conflicts. Actually it creates a second battlefield and burden without improving the first part.
 * If you say that common sense is already in place and a good solution - then, why do we need a filter? Both things do exclude each other somehow. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 14:36, 20 July 2012 (UTC)
 * Not really. What we have now is a 0-1 choice. A community can either take a potentially controversial image, or leave it. We don't have a choice to use it with a cautionary warning, for example. See also my comments below. Pundit | utter  14:59, 20 July 2012 (UTC)

Pundit, when you say "so far the community has been acting with common sense", I find it hard to imagine that you are speaking from experience. Sure, in broad terms, our needle tends towards common sense, but we also make bad decisions all the time.

As Niabot suggests, an image filter will inevitable bring with it a whole heap of new drama and stress for the community. What I would say is that it had better be worth that price. I think the fundamental design of the filter should focus on what will genuinely benefit the project. But I'm concerned that many users take an abstract (and IMO, old-fashioned) Theory of Wikipedia approach, which focuses instead on maximising collaboration and designing-in a flat structure for its own sake. Why would it benefit the project to allow interminable discussion about whether Olympia or whatever should be subject to filtering? For me, the answer is that it wouldn't, so we should make sure that we don't end up in that place. Formerip (talk) 16:39, 20 July 2012 (UTC)
 * We're all theorizing, but I'd be VERY surprised if ancient famous art was subject to such debates. Discussions would be involved only if there was a disagreement that a given picture may or may not be disturbing to a significant minority of our readers. All in all, I believe that we have been very successful as a community in eliminating porn, also by using fuzzy, consensus based criteria. There has been a lot of drama, tons of disputes, etc., but it works. For advisory labeling there could be even simpler procedures for protesting a label. Pundit | utter  18:52, 20 July 2012 (UTC)
 * If we make the mistake of trying a filter system based on the current Commons categories then we can pretty much guarantee that some art will be caught up in the filter. Especially if this is used to filter out female toplessness. As for having a simple procedure to protest an advisory label, I've no doubt that we can setup as simple a system as WP:AFD. But simple systems to handle vast numbers of borderline cases can be a huge timesink. We already have incredibly complex arguments as to the notability of borderline articles and edge cases over copyright and freedom of panorama; Do you really want endless arguments as to how translucent/clingy a top needs to be before an image goes into a female toplessness filter? Or indeed how much cleavage is acceptable before an image goes in the filter? There are ways to do this that allow for shades of grey, or more pertinently different tolerances for cleavage and opacity, but if you go for a simple system based on the Commons categories you are heading for a traincrash.   Ϣere  Spiel  Chequers  10:43, 21 July 2012 (UTC)
 * But some art is disturbing to some populations - Félicien Rops with his drawing of a saint Theresa playing with a dildo is just an example. Yet, the thing is, that categorizing a picture as known to be or potentially disturbing does not really change much - it is the viewers, who can use such categorization (or projects). Categorizing only enables them to have more choice. Of course, there are other ways than following the current categories, but I'm very doubtful if the community would actually try to enforce a mechanistic rule cleavage=caution rule.  Pundit | utter  17:31, 21 July 2012 (UTC)
 * I would disagree that there is any such thing as a community in Wikipedia and I would resist any such tendency towards being a social site. This may be agreed by consensus but that is a consensus of individuals. And the individuals have widely different opinions and philosophies and backgrounds. Some people are like others but that's about as far as it goes. Dmcq (talk) 19:45, 21 July 2012 (UTC)


 * Pundit, what I really, really take issue with is your assumption that offering users more choice is a virtue in and of itself. That's contrary to the way Wikipedia has traditionally worked (at its heart is the idea of consensus - a single question having a single, collective response and not giving rise to a range of alternatives). Giving users choice (actually or potentially) is surely a good thing if it improves the enyclopaedia and a bad thing if it doesn't. I'm worried that we may be about to simply shortcut consideration of what we actually want to achieve with a filter. We may end up with a filter designed to generate work and grief for the community, rather than to serve its goal.
 * I think I agree with WSC, that the thing needs thinking from the ground up. If we simply grab hold of the nearest thing to hand (i.e. Wikipedia's existing system of categorisation) then I think it is only natural that we will end up with a pig's ear.
 * But, more than that, we seem to be headed for the deployment of a tool before we get clarity about what it supposed to be for. Getting that clarity - either a community discussion or a WMF fiat - may mean a delay to implementation, but I think it would also make the implementation much less of a nightmare. Formerip (talk) 01:02, 22 July 2012 (UTC)
 * I agree that categories may not be the best solution, since they are too global (and there are other issues). I think that ultimately each project should have the right to develop their own tagging/labeling, and decide if they want to use if for anything (filtering is just one option, there could be textual warnings, and many projects could refuse to use any labels at all). Pundit | utter  01:24, 22 July 2012 (UTC)

Comments from Pundit
On the same day that Jimbo published his proposal I came up with a (lengthy and personal) essay on the same topic on meta. It makes more sense to discuss all solutions here. After reading through the discussions, I think a couple of points are worth emphasizing/reiterating: These are my two-cents. As some of you know, I am personally quite insensitive to controversial images. But again, it is not about our sensitivities, but of our readers, as long as we don't really start limiting access to knowledge. Pundit | utter 14:59, 20 July 2012 (UTC)
 * it should not be about filtering, but about image advisory. While most of us oppose censorship, and many of us support unrestricted free speech, I believe it is still possible to seek consensus for common sense warning signs.
 * any solutions affecting the articles' outcome should not be global. Pushing global filtering down projects' throats is not a way, and it is good it is not on the table. Yet, if separate projects could decide about how to use provided warning signs/categories (and which ones) for the reader communities they serve, they would most likely know best what works for them (A factoid: in the US, computer games avoid sexual imagery, while in Germany they avoid showing red blood.). Some could not use them at all, some could add cautionary tags, some would allow personalizing the preferences by the logged-in users, etc.
 * people do stumble upon images shocking to them unexpectedly. It is a complete illusion to assume that people only find what they're looking for. There's plenty of severed limbs, inflammatory caricatures (we seem to care more about deleting racial slur than religious one, it would seem), sexually explicit materials, and it is fine, but we should make reasonable precautions that people don't see them if they don't want to. A personal example: as a non-native speaker, I've learnt the meaning of the word "fisting" from Wikipedia. I wasn't shocked or moved in any way, but was not really expecting anything sexual. I think it is common sense to recognize that some people may be shocked by some images. Also, even common terms searches may bring unexpected results.
 * Readers of our projects are predominantly not logged-in, therefore any solutions we seek should take this into account. There is a lot of space for discussion between global solutions and complete personalization at individual user level. Separate communities understand the cultures they operate in well enough to decide whether they want to use advisory labels in any way. Ultimately, we serve our readers, not just ourselves. If German Wikipedia decides about an explicit image on their front page as a FA (as they did), it is up to them, they know what they're doing. Ultimately, publication standards for different cultures have to differ, and the communities are the judges of those standards. But projects should have a choice whether they want to be able to use some advisory, and they don't now. The only choice they have is to take the images or leave them, and this is very limiting.
 * (obvious) there are images on the commons which are controversial to some people and still needed for encyclopedia. Educational (e.g. detailed explicit explanations of sexual positions), artistic (Félicien Rops has been known to cause stir, for example), informational (e.g. severed limbs from accidents, etc.), religious (e.g. Muhammad caricatures). I am totally insensitive to any of these, but again - it is not about us, as the editors, but about the readers. We need to serve also those who are sensitive to some images, without limiting universal access to knowledge to the others. The only reasonable way to both be able to keep and use the images, and still care for the ones who find them shocking, is to allow advisory categories. How individuals and communities use these categories, and if they are going to use them at all, is up to them.
 * the purpose of image advisory goes beyond Wikimedia projects - people use images from Commons outside Wikimedia. They use them for conference presentations, websites, books. They often trust that what is considered encyclopedic by Commons standards, is safe for unrestricted use, but it clearly not always is. I myself, even though I've been involved with Wikimedia projects for 6 years, have served as an admin/crat/stew (although with no experience with dealing with controversial images on the Commons), made this mistake as well. Adding a warning sign ("use caution", basically) is thus a sensible thing to do. Advisory labels, in fact, would protect also the insensitive part of the population, who needs being advised on what may be shocking to the others.
 * all filtering solutions should be crowd-sourced to the community, and not to any hired staff or algorithms. Just as deciding about notability is an essential community's right, so is deciding what may need advisory label. The community has been able to perfectly sort out porn, so we should be able just as well to select images requiring caution. It is just one layer of advisory, as simple as that.
 * using categories for advisory labels makes a lot of sense. First, it is easy. Second, it allows smart sub-categorizing. But thirdly, it also allows to browse controversial images more easily. Yes, this is an asset. Some of our readers seek them, and it is also of tremendous academic value to find out what people find requiring advisory. A side effect of advisory labeling is better sorting of controversial images, for those who seek them, too.
 * what requires advisory is not really that subjective. All cultural standards are arbitrary, different groups have different sensitivities, all true. But down to business, in practice, we have proven that we are able to e.g. discern art from porn, really well. Why shouldn't be able to recognize what images may be shocking to some populations?
 * It is about freedom to choose, ultimately. Nobody is going to increase censorship by allowing projects to independently decide how to use advisory categories. If there is any censorship, it is the preventive one we face now: projects decide not to use images, because they're controversial, even if these images would be good illustrations to the articles. By not giving the projects a choice whether to use cautionary categories or not, we are contributing to the result which is in contrast with our principle, to make knowledge universally accessible.


 * Advisory and providing a NSFW Filter is the same. Prejudicial labeling is designed to restrict access¹, based on a value judgment that the content, language, or themes of the material, or the background or views of the creator(s) of the material, render it inappropriate or offensive for all or certain groups of users. The previous sentence is a slight variation in wording of what the American Library Association calls a "censor's tool". Adding a Warning sign is POV, our own POV, discrimination and judgment about a topic. It is in violation with the five pillars (II).
 * ¹ This is meant in a broad sense of to warn, discourage, or prohibit users or certain groups of users from accessing the material.
 * I reply in each thread separately. You are absolutely right, filtering is not the same thing as categorizing/labeling. I believe we need categorizations globally, while project/individual filters cannot be introduced globally. Image advisory is "POV" in the same sense that decisions about notability are.  Pundit | utter  16:45, 20 July 2012 (UTC)
 * No it isn't. We judge about the notability. That means if we want to cover a topic. But if we do, then we don't write inside the article: "The worst shit i have ever seen. You really should not look at it". Thats is a small but mighty difference between covering a topic and sharing you personal view about a topic. Thats why we use secondary sources. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 17:47, 20 July 2012 (UTC)
 * I see your point - we have established rules for notability in terms of referring to sources. You're right. However, we do make "fuzzy" ad-hoc decisions e.g. about whether an image is porn or art. Thus, ultimately, if there is a consensus that something might use advisory (and keep in mind, just neutral label, with no comments), it should work fine. Pundit | utter  18:36, 20 July 2012 (UTC)
 * There is no such thing as a neutral warning. A warning requires a reason/threat to be present, which indirectly tells the reader that something must be bad. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 20:30, 20 July 2012 (UTC)
 * See, here I finally completely disagree. E.g. parental advisory labels have nothing to do with considering materials to be "bad". So isn't marking materials as inappropriate for certain age. It is not good/bad judgement, it is just a simple sensitivity warning. Pundit | utter  21:31, 20 July 2012 (UTC)
 * I oppose such warnings if they dictate an expected behavior or reaction. There simply is no good advisory label if it comes to knowledge. But i guess we can stop at this point when you don't accept that filtering and adding advisories of/to selected topics is an unacceptable violation of NPOV. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 21:45, 20 July 2012 (UTC)


 * Any solutions affecting the articles or content is bad It doesn't matter if all projects or a single project is affected. It would be support for censorship going against the mission of the projects, - the definition of an encyclopedia itself.
 * I miss your point here. How is proper categorization censorship? Pundit | utter  16:45, 20 July 2012 (UTC)
 * I guess you did not read the linked text above. There are two kinds of categories. Directional aids (help to find what you look for, NPOV) and prejudicial labels/warnings/... (POV) -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 17:36, 20 July 2012 (UTC)
 * Community consensus is not prejudice. One person's POV is. If there is a consensus that something could use advisory, it is exactly that - A CONSENSUS, allowing to issue a warning. It has nothing to do with censorship. Pundit | utter  18:36, 20 July 2012 (UTC)
 * If all authors (the community) of Avatar agree that the movie was the worst piece ever, then we would rightfully add a advisory message to the article that it may be boring to watch the movie? -- ／人 ◕ ‿‿ ◕ 人＼  署名の宣言 20:30, 20 July 2012 (UTC)
 * Perhaps? But I don't think our readers expect us to pass judgments on movies, while they do voice concerns about our explicit content available without warning. Pundit | utter  21:31, 20 July 2012 (UTC)
 * I say no. They don't expect us to add advisories to our content. If they want that, then they can do it on their own, like any reader can have it's own opinion about a movie. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 21:45, 20 July 2012 (UTC)


 * You get what you want or have to expect If you are not able to be offended by reading or seeing something you dislike without knowing what it is, then don't click on anything that looks suspicious and don't even think about buying a book. The cover itself may be offensive, as well as the summary at it's backside. When people search for futanari without knowing what it is, but blaming Wikipedia for telling them, then it is their own fault. If they use Google, then they will also find the answer. Maybe you want to compare both results.
 * There's been plenty of examples - learning does not have to assume automatic acceptance of explicit images. Pundit | utter  16:45, 20 July 2012 (UTC)
 * No there isn't an automatic acceptance. I don't expect it. But we should not encourage the reader to automated nonacceptance either. A filter with presets encourages nonacceptance. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 17:36, 20 July 2012 (UTC)
 * Yes, you're absolutely right. This is why I don't support global filters, but rather simple advisory. At least with some images we can reach consensus, that they may be disturbing to many people. And images for which there is not such consensus, clearly stay without additional labels. Surely you'd agree that at least in some cases we can quite responsibly know that they may be disturbing. Pundit | utter  18:36, 20 July 2012 (UTC)
 * You speak about advisory. But advisory messages are useless (in your sense) if you see the warning and "disturbing content" at the same time. I proofed in my previous comment that advisory is as bad as filtering since it serves the same purpose. (prejudicial labeling) We won't reach a consent in this discussion if you try to weaken the statement that both are bad things to do, with or without consensus.
 * I would have nothing against a general advisory that "Wikipedia isn't censored" and displays all human knowledge. But adding such information only to some articles or images is discrimination and discouragement. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 20:30, 20 July 2012 (UTC)
 * The purpose of filtering is disabling displaying images. The purpose of advisory is allowing a user to be warned about content, which is still extremely easy to access. Pundit | utter  21:31, 20 July 2012 (UTC)
 * I don't see much difference between a filter that can be disabled at any time and an advisory. Both label the content as inappropriate for some kind of audience, before the audience itself looked at the content. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 21:45, 20 July 2012 (UTC)


 * The search can be easily improved, if someone would invest some time into improving it, without additional tagging: A little bit of intelligence
 * Absolutely! But sort of sideways of our discussion. Pundit | utter  16:45, 20 July 2012 (UTC)
 * You started it. ;-) -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 17:36, 20 July 2012 (UTC)


 * The community doesn't need a filtering solution The community hasn't been able to perfectly sort out porn, but it tries it's best to find suitable illustrations to illustrate a topic. The editorial process handles such cases already. If it does well, then we don't need a filter. If it doesn't do a good job, then it wouldn't help to provide a new tool which can't get any better as the community itself.
 * The community and the editorial process have been both very effective so far. What we need is not a filter, but "advisory" category, adding a layer between stuff which is totally non-controversial, and the one which is overly controversial. Also, it seems you didn't read my comment - what I'm saying is EXACTLY that the community should make the decisions about where would the advisory label be applied. Again, NOT a filter. Pundit | utter  16:45, 20 July 2012 (UTC)
 * It is controversial to the core. We want to share unbiased knowledge and say as group A "This stuff might upset you" while at the next corner someone belonging to group B is offended by content not tagged by A. This shows clearly that such labeling is our (group A) POV. Its the same as writing "This movie is shit" as the conclusion inside a movie article while deleting the various opinions belonging to critics. We cite them for a reason. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 17:36, 20 July 2012 (UTC)
 * But again, don't you see that your argument may be extended to porn? Yet, we have common sense to not allow it. Also, the message is not "this stuff might upset you" but "we know that some sensitive people might find this image upsetting. If you're one of them, or a minor, be prepared". Pundit | utter  18:36, 20 July 2012 (UTC)
 * Look at my answer in the previous section. I would only repeat myself. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 20:30, 20 July 2012 (UTC)


 * What requires advisory is entirely subjective All cultural standards are arbitrary, different groups have different sensitivities. We respect them by staying neutral and not to follow one major group with it's own kind of sensitivities while ignoring others. Actually we have proven, Jimbo himself contributed to it, that we aren't always able to e.g. discern art from porn, really well.
 * We as individuals, definitely. But we, as a community, with consensus procedures, have proven extremely well that we are good at making such judgments. Pundit | utter  16:45, 20 July 2012 (UTC)
 * You belief your own words, seriously? How often did i hear the story that our community isn't representative, because it consists mostly of educated, young men, with an own overall bias. But as soon it comes to a/the filter it is somehow the pinnacle of evolution that does anything right and perfect. The typical group A vs groups B or majority A vs minority B, C and D problem, while group Z is the average reader. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 17:36, 20 July 2012 (UTC)
 * Not filter, advisory label :) I believe we are as sensible as it can get in making a communal judgement that some pictures may be disturbing, but are still needed for encyclopedia. Pundit | utter  18:36, 20 July 2012 (UTC)
 * Same again. I guess we can start to narrow it down. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 20:30, 20 July 2012 (UTC)


 * The reader has the right to choose If an reader isn't happy with a book, newspaper, writing, picture etc. then he doesn't need to look at it for 24h a day. He can just put it back, don't read it, don't look at it, don't buy it. But if a filtering system is exploited, then we have readers that want to look at some interesting content, but they can't for 24h a day. Any category (not limited to existing categories) based proposal does enable this possibility. It limits the freedom of choice for the users.
 * You're right. Yet, opting out from Wikipedia simply because we refuse to use advisory categories seems a bit extreme choice you suggest we should force our readers to face. Also, how does additional categorizing limit anything? Pundit | utter  16:45, 20 July 2012 (UTC)
 * It's about not having another choice. Do you have numbers on how many people in the world are bound to one, only one, provider? What if that provider does accept our (group A) NSFW sentiment and enforces it (a simple task for a proxy, you use one right now), allowing members of group B only to see what group A approved to be acceptable for everyone? -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 17:36, 20 July 2012 (UTC)
 * This is a very fair theoretical point. But in practice, we could use advisory categorization without details as well (not differentiating within it). Also, I don't think that currently there is a threat of censorship from major Internet providers. If there was, as a community we could take action, too (e.g. by cutting out the censoring providers). Let's not avert from addressing real problems by inventing theoretical ones. Pundit | utter  18:36, 20 July 2012 (UTC)
 * It isn't really theoretical. Googles safe search was already exploited by multiple providers, denying the access to images and pages found in the uncensored searches, but not inside the "safe searches". A pretty easy way to let Google work for them. The same could easily be done with Wikipedia. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 20:31, 20 July 2012 (UTC)
 * And you seriously believe Wikimedia community would not act then? I'm convinced any attempt of this sort would end up in blocking certain providers, if necessary. Pundit | utter  21:31, 20 July 2012 (UTC)
 * Looking back at the discussions about the "Virgin Killer" case, i doubt that the community would find a quick consensus. The lengthy discussion would be if it does more harm then good to take down Wikipedia in protest. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 21:57, 20 July 2012 (UTC)


 * The purpose of image advisory goes beyond Wikimedia projects It affects the daily life of many people and would strengthen the position of censors that point at Wikipedia telling the story that their doing is justified and generally accepted.
 * Censors? Why and how? Not any more than "sex" category, if at all. Pundit | utter  16:45, 20 July 2012 (UTC)
 * You should ask some people from UK that just recently switched their provider, if possible and affordable, to avoid censorship. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 17:36, 20 July 2012 (UTC)
 * If it affected Wikimedia, as a community we could take a stance. There have been Wikipedia blackouts on it-wiki, en-wiki, ru-wiki and they were successful.  Pundit | utter  18:36, 20 July 2012 (UTC)
 * We could. But do we really want to wait until we have to do so? -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 20:30, 20 July 2012 (UTC)
 * We want to serve the readers as we can best. At present we have a theoretical censorship threat, and a practical problem of explicit images thrown at readers not wanting/expecting them. Moreover, we even know how we can address this theoretical threat, if it ever becomes even remotely real. Pundit | utter  21:31, 20 July 2012 (UTC)
 * I think that this "practical problem" is also a theory. At least we don't have a single neutral study to back it up and of course not even a single study directed at the readers. But we have studies that show that the "theoretical threat" might be existent and already occurred in similar cases. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 21:51, 20 July 2012 (UTC)


 * I guess that we have an strong disagreement and very different views about this points. Maybe we can sort out the details in a further discussion, but from my point of view you have a very different understanding about collecting, providing and spreading of knowledge. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 16:30, 20 July 2012 (UTC)
 * This very may well be so. But do keep in mind, that we all serve the readers, who may benefit from categorizations. Those who don't need them, won't even notice. Also, preventive censorship (projects not using images because they are too explicit) is much more real than theoretical concepts of global censorship (by whom? how?...) you invoke. Pundit | utter  16:45, 20 July 2012 (UTC)
 * We serve the readers. But i don't think that we serve our readers by telling them what is good and bad. We give them the opportunity to inform themselves and to decide for their own what they see as good or bad. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 17:36, 20 July 2012 (UTC)~
 * Absolutely, I can't agree more. Yet, cautionary signs are not filters. Nobody would be limited in what they can see, they'd only have an option to choose if they want to see it. You seem to be anxious about corporate or totalitarian regimes censoring the content, and this threat should not be treated lightly. However, as a community we have many different measures to take if some of the intermediaries tried to handpick from our content and make some of the categories non-accessible. Pundit | utter  18:36, 20 July 2012 (UTC)
 * Advisory labels do not stop you from accessing content, but they can discourage people from looking at the content, which is some kind of discrimination of a topic, even if they would have no problem with it at all. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 20:30, 20 July 2012 (UTC)
 * If this was really the only concern, it could be addressed by textual caveats describing in general what is in the image, on which the advisory is issued. Pundit | utter  21:31, 20 July 2012 (UTC)
 * That doesn't make it any better. I guess you missed the important part of the last comment. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 21:51, 20 July 2012 (UTC)
 * I simply consider using categories as an artificial form of discrimination. Stating that e.g. some images would be considered inappropriate for minors under many legal systems is not "discrimination", "prejudice", or "arbitrary POV", but simply a statement of fact. The crucial thing is that some users do express their disappointment with the project after stumbling upon explicit content, and also the feedback from the general social environment is that we are somewhat inept with tackling with the problem (if the strongest criticizers of wiki pick up on this, you can be sure they perceive it us our weak point). Yet, let's agree to disagree.  Pundit | utter  22:48, 20 July 2012 (UTC)
 * To really tackle this problem we would need to change the entire society. Adding advisories is no solution. Do you think that it would stop the criticizers if we add "disclaimers"? I doubt that. As long the "child/minor argument" is used, it is entirely useless. It may sound like irony, but thats actually the best solution i could find and that already works: Ignore the problem and give to the parents. Let them take responsibility. They have it anyway, since the internet is no place to park children. If we want to create "Wikipedia for children", then we could try to do it. This would at least circumvent the "shocked child" hypothesis. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 00:33, 21 July 2012 (UTC)
 * You seem to believe people don't want advisory and maybe the majority indeed don't. It still does not change the fact that a minority may want it, and be harmed by stumbling upon explicit images, which you clearly don't care about at all, because of the abstract anti-censorship ideals. While I probably share your lack of sensitivity in terms of images, I don't in terms of real impact our project have, and I would like to keep it as close to positive as possible. Pundit | utter  02:08, 21 July 2012 (UTC)
 * Any within Wikipedia solution wouldn't actually prevent children from accessing images. However it would certainly be possible for some external tool to use the information from Wikipedia to do actual prevention. I know even that possibility will cause disquiet to many anti censorship people but I believe it is something that they can eventually live with. It isn't as though external tools can't or don't do filtering already anyway, they'd just get some extra help for free. At most what that would do extra if they had a facility to use Wikipedia filters would to allow for instance Muslims to prevent their children seeing images of Mahomet on Wikipedia. Personally I think if Muslims are so interested in that why have they not asked internet filter suppliers for such an option. For governments they have no problem paying loads of their own people to make filter lists and would not trust us anyway so it is not as though anything like this would make a difference. What we are talking about really is a personal filter facility. Fox News is not going to be satisfied with it but I believe allowing people to avoid pictures of spiders for instance is a very useful facility. Dmcq (talk) 10:15, 21 July 2012 (UTC)
 * p.s. if you search for 'Mahomet' in Google it shows a list of images for Mahomet, if you put in 'spider' it shows images of spiders, and if you put in 'violence' it shows videos of violence. However if you put in 'sex' it shows no images or videos, quite interesting I think :) Dmcq (talk) 10:24, 21 July 2012 (UTC)
 * @Pundit: I believe that adding advisory labels only to some of our content is discrimination of that content. It would be our POV influencing the attitude of the readers towards those topics. We can't remove the fact that our readers have their personal preferences, but it isn't the goal of the project to influence this preferences, which advisory labels inherently do. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 11:46, 21 July 2012 (UTC)
 * Would you be okay with what I advocate which is to allow User:RedNeck say to set up a filter User:Redneck/Sex_and_Blasphemy and share its use with others of a like mind or even eventually get it into WP:Filter/Alabama_Sex_and_Blasphemy with a description at the top and the remit 'Filters out images normally considered indecent or blasphemous in Alabama'. The images would still be accessible with one click even if the filter is on. No tags would be added to the images themselves. Dmcq (talk) 15:19, 21 July 2012 (UTC)
 * Wouldn't that be the reinvention of adblock? As long such lists aren't associated and supported (created, maintained) by Wiki(p/m)edia i don't have a big problem with them, but i have problem with "community approved" (at least labeled in similar manner) filter lists. They are not part of the mission of this community. It would be part of the "Alabama..." community, even so single users might contribute to both communities. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 16:48, 21 July 2012 (UTC)
 * WP:ESSAY says 'Essays are the opinion or advice of an editor or group of editors (such as a WikiProject) for which widespread consensus has not been established'. It is okay to put things into WP: if they don't cause some concern. That is not the same as being approved by the community. I don't see the connection with adblocks..The foundations's mission says 'The mission of the Wikimedia Foundation is to empower and engage people around the world to collect and develop educational content under a free license or in the public domain, and to disseminate it effectively and globally.' If we don't cater for peoples taboos we can't disseminate the information to them effectively. Dmcq (talk) 17:20, 21 July 2012 (UTC)
 * See also vision 'Imagine a world in which every single human being can freely share in the sum of all knowledge. That's our commitment.' It says nothing about only if they are willing to abandon their principles or work past their fears. Dmcq (talk) 17:24, 21 July 2012 (UTC)
 * Let me call this "Bullshit!" at once. Not the quotes, but their interpretation. That you are free to do something doesn't mean that you have to do it. It means that you can do it if you want to do it. No one forces you to look at things you dislike or to share them. What you described is word bending in excellence. It's like saying: "I can't go to school in Florida as long there is bad weather in Norway". -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 18:23, 21 July 2012 (UTC)
 * Could you be more specific about what you don't like about having a personal filter please. Why do you wish to stop a person preventing pictures of what they consider blasphemy being presented to them when they consult Wikipedia, or is it some other kind of problem you see? And you did not explain the reference to adblock. Dmcq (talk) 19:37, 21 July 2012 (UTC)
 * I have nothing against a private personal filter. Everyone could add the content that he dislikes to such a list, like everyone could just cover up pictures in a book that he does not want to see or to show to someone else. But i don't think that we (as the community) should work on such filter lists. Neither should we support their creation. It makes us responsible on what we filter and what not, while at the same time we have to apply moral judgment for others.
 * Adblock has two options. Either you can suppress advertisements manually or you can select from a list of presets. Naturally none of this lists is perfect. But you should keep in mind that this presets don't rate some adverts as good or bad. In fact this lists have the goal to remove any kind of advertisement (like an show all or nothing filter). -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 20:13, 21 July 2012 (UTC)
 * By community here I think you mean a central consensus and I certainly have not said anything like that. I was talking about User:RedNeck setting up and sharing an Alabama_sex and blasphemy filter. That's hardly an overall consensus and people could choose to use or not use it after looking at the description. So why did you go on about bullshit? Dmcq (talk) 20:25, 21 July 2012 (UTC)
 * If i go after Jimbos introduction it should be as simple as possible. How should that sharing work? How do readers choose/find a list? What about users that only want to have a private list, without letting others know what they blocked? How will lists be updated, on user basis or by group of users? You mentioned global lists like "WP:Filter/Alabama_Sex_and_Blasphemy". How will they be established/choosen/supported?
 * I know that are a lot of questions, but I'm interested what kind of system you have in mind. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 20:37, 21 July 2012 (UTC)


 * I think the purpose of advisory labeling is not as much prevention of showing images, as prevention of seeing them unprepared. Pundit | utter  17:34, 21 July 2012 (UTC)
 * Does a warning sign like "Warning, trespassing may be dangerous" encourage you to go further or does it urge you not to go any further? -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 18:27, 21 July 2012 (UTC)
 * If we're to stick to road signs metaphors, a sign "Use caution, slippery road" makes me, well, use caution. It does not make me think the road is morally bad, it does not urge me not to travel, it does not make me think that the state tries to limit my liberties. Pundit | utter  18:59, 21 July 2012 (UTC)
 * Right. It rates a part of a street as dangerous because a slippery road is life threatening for everyone that drives to fast. Advisory messages rate an topic as a whole as dangerous and threatening even so there is no real danger to all while reading through it with full speed. A road sign is no moral issue, while an advisory imposes moral judgment on a topic. -- ／人 ◕ ‿‿ ◕ 人＼  署名の宣言 19:31, 21 July 2012 (UTC)
 * Wouldn't we just say the image is not shown becuse filtering is on, with clickable links on filtering to explain that and another to dhow the image anyway. Dmcq (talk) 19:37, 21 July 2012 (UTC)
 * Let's not make it overly theoretical. It is not about morality. It is a simple information sign - "Look, some other people consider this image to be potentially disturbing. We don't really care if you are one of them, we are just letting you know to let you decide". No moral judgments. No right or wrong. In fact, I bet most of the images requiring advisory are morally good (how can a good photo of a burned limb be morally bad? but in the same time, it is common sense to accept the fact that some people would e.g. not like to see it while eating lunch, or the fact that some kids could have nightmares after watching severed bodies). Pundit | utter  19:45, 21 July 2012 (UTC)
 * I'm trying to figure out why you need say anything like that. If they see the image they'll know if they find it disturbing or not, if they don't it'll be because they have a filter on and that can be pointed out to them. We don't have to worry about feelings or why. Dmcq (talk) 19:49, 21 July 2012 (UTC)
 * I don't think we understand each other. I'm only discussing labeling at this stage (tagging images with "advisory" notes), not actual filtering applications, which would be up to projects/users. Also, I'm not really suggesting any comments besides "advisory", I was only using the metaphor provided by Niabot to explain that some form of common sense tagging does not have to involve moral judgments. Pundit | utter  20:32, 21 July 2012 (UTC)
 * @Dmcq: It is not about the explanation why the image is hidden, it is about who decided to hide it for which reason. If we hide an image then there must be reason to do so. But what is this reason? Isn't it not just a moral judgment that a certain group of users/readers doesn't want to see it, because it is X? But let us grab an real example and make a small test for ourself. I ask you to write down X for the first image in hentai (japanese pornography). Then we can go through your explanation and see on what basis you would like to hide it and to which filter list(s) it should belong. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 20:28, 21 July 2012 (UTC)
 * "We" don't hide anything. Specific projects or specific users may choose to. We're only discussing labeling/tagging here. It is not a moral judgment, but basically a result of a simplified poll, of utter interest for many viewers.  Pundit | utter  20:32, 21 July 2012 (UTC)
 * Why do we need to label any images? Labelling implies either a central consensus to me or else a proliferation of fairly random tags, I don't see either of those options as very desirable. Dmcq (talk) 20:41, 21 July 2012 (UTC)
 * Labeling does not have to be central, it can be project dependent (and rely on tags similar to cross-wiki links). Unless we want to enable hiding all images, some form of tagging pictures which may require advisory (in a wide community consensus) is necessary. Pundit | utter  20:47, 21 July 2012 (UTC)
 * What are those "specific projects or specific users"? What is this "simplified poll" and how do you define "many viewers"? We never had an official poll directed at the readers before (i asked for one) and i explained previously (not this discussion, but somewhere on Meta) that "many" is very stretchable. I once counted how many people are viewing articles like hentai, bdsm or futanari in a month and compared that with the possibility that 0.1 % of the offended readers would be able to leave a comment on the discussion page or would remove the images from the articles. The irony is that either less then 0.1 % of the offended people are willing or able to make and edit or to leave a comment, or that the actual number of offended people is below 0.2 %, which is way lower if we could expect that more then 0.1 % are able to leave some kind of message. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 20:53, 21 July 2012 (UTC)
 * Apologies if I sounded unclear. Specific projects or people means simply that projects or people would have the possibility to use labeling at their own discretion. A "simplified poll" is, basically, what we do when reaching consensus. I hope you realize that most of our viewers don't even understand they can edit articles, not to mention leaving comments on the talk pages. Pundit | utter  21:00, 21 July 2012 (UTC)
 * Why not just use a list of files when filtering? A separate generator program could do complex operations to generate a list like logical operations between categories and specific exclusions or inclusions, but I see no requirement to label any images. Dmcq (talk) 21:03, 21 July 2012 (UTC)


 * (ec) Which project should be responsible to add such label, and would it not mean that we, the community, add moral value judgment (POV) to articles? By the way, i assumed that only 1 of 1000 readers would be able to edit the pages. The number of vandals that leave the images in place is much higher in comparison. -- ／人 ◕ ‿‿ ◕ 人＼ 署名の宣言 21:06, 21 July 2012 (UTC)
 * Dmcq - absolutely, a list of files could be a working solution. Yet, I think that keeping tags near images makes more sense, since sometimes discussions will be needed. "Which project should be responsible for adding a label" - each project could make a decision for themselves. I could imagine e.g. a standard discussion syntax such as /talk/en/NSFW for establishing consensus. ONLY projects interested in using such labels would even have to bother. Individual users, however, could decide by themselves if they want to rely on some project's consensus in this respect, even if their "own" project does not use labeling. Pundit | utter  21:23, 21 July 2012 (UTC)
 * I think a discussion should be associated with the filter not spread all over the place. Dmcq (talk) 21:26, 21 July 2012 (UTC)
 * I respectfully disagree. I believe we're discussing solutions to the presented problem. Filtering is just one solution. I believe that tagging/labeling is another one, and that it addresses the biggest problem of adjusting to the projects/people's needs (opinions vary to the level of forking). Pundit | utter  21:34, 21 July 2012 (UTC)
 * You have completely lost me there. What you said there just seems a jumble of words with filter and tags etc stuck in. If one were talking about NSFW one would probably have a load of images to talk about that had some specific aspect that needed discussion. It would not be associated with a specific image. What has labels or tags therefore got to do with discussion? Dmcq (talk) 21:54, 21 July 2012 (UTC)
 * Again, apologies for the confusion. The only thing I'm saying is that we could use interwiki-like tagging (e.g. adding en:nsfw) to specify which projects have decided that the image may need advisory. Then, and only then, projects would independently decide what they want to do with it. Some would propose filtering, some would add text advisory, some would do something totally else. This is different from individual filters, even though individual filters could be based on such tagging. Pundit | utter  22:03, 21 July 2012 (UTC)
 * Btw, maybe I misunderstood you - when you say that "a discussion should be associated with the filter", do you mean that there should be one discussion about each image? I more or less mean the same thing, even though I believe that each project should be able to decide separately. Pundit | utter  23:09, 21 July 2012 (UTC)
 * By the discussion should be associated with the filter I mean exactly what I say, the filter NSFW for instance would have a talk page associated with it. No it would not be associated with the image. Dmcq (talk) 00:05, 22 July 2012 (UTC)

How many images are we talking about?
Sorry to ask again, but does anyone have any estimate of how many images we're talking about as being possibly filtered out assuming a pretty strictly no bare breasts no violence type reader where even nude art is included? 10,000? 100,000? 1,000,000? more than a million? Dmcq (talk) 00:54, 22 July 2012 (UTC)
 * I've categorised thousands of images on Commons and yet I have no idea what the true answer would be to this. I'd suggest that you restate the problem, if the community were to accept a filter based on our category structure, and if the community could agree a single clear definition of NSFW, then how many edge cases would there be, and how many millions of images would need to be reviewed against that criteria? There are over 13 million files on Commons, plus a large number on other projects, and Commons is still growing quite rapidly, so we have an idea of the task if we decide to go down the category route. Of course neither an agreed definition of NSFW nor community consensus for a filter based on our category structure are plausible, viable or likely scenarios. If we really want an image filter we need to start by accepting that reality.  Ϣere Spiel  Chequers  10:05, 22 July 2012 (UTC)
 * I was just hoping for a very broad brush estimate but it looks like that even isn't available. It could make a difference to how the stuff should be implemented, what is possible and what would just clog things up, how quickly could changes be effected, what would caching do if changes were delayed. It is possible to use tricks to fix many problems and for instance how Netflix rates preferences could help shrink the data but it is better to have something straightforward which isn't too slow and expensive to start with. Dmcq (talk) 12:04, 22 July 2012 (UTC)
 * I'm fairly sure that of the 15% of Commons images that have been imported from the Geograph less than 1% would offend anyone who doesn't want to see statues of nude or seminude people. Though that includes images of statues of completely nude males as well as topless or skimpily clad ladies. If you add vertigo then there will be at least as many more. But I honestly don't know what proportion of the other 85% are potentially offensive images, and I'm pretty sure that we have some uncategorised and undercategorised image releases from cultural partners. Of course clicking on a few hundred random images would give you a ballpark re Commons - by the time you've found twenty you could be fairly confident that you knew whether on the definition of NSFW you were working to there were thousands, tens on thousands or a million. But the project has a shedload of other files such as the fair use ones here on EN wiki, and I doubt if anyone is familiar with all those projects.  Ϣere Spiel  Chequers  13:39, 22 July 2012 (UTC)
 * Well I clicked on random image two hundred times on commons and got
 * One image of a person with a gun at a practice range
 * One image of the back of a semi naked lady
 * One image of a nude statue
 * One image of some religious art with one of the figures semi naked
 * I didn't find any images of spiders or Mahomet or Nazis. So that would give at most 2% that someone would find objectionable I think which comes to about 200 thousand files I'd guess. I don't think many of the fair use ones would have problems. Most filters would be quite a bit smaller - probably only the back of the semi naked lady might be counted as NSFW for instance. It is a large number but still reasonable to download as a compressed list of files but if downloaded it is definitely worth cacheing for a while. With a half decent format I don't think checking would be a problem even on a mobile phone. It is a pity the proportion wasn't much bigger in which case a list of files would become a bit unworkable, or much smaller in which case it becomes the obvious choice.


 * The second alternative is for the server to put various tags into the html where the images should be and the Javascript just checks those tags. There would have to be limited number of such tags or perhaps it could just put in a bit mask for say the 64 most popular filters.


 * If javascript is not supported the server itself would have to do the filtering and if there are a very limited number of filters or the main ones are very popular it could cache the result. I believe we'd want to make this efficient - it seems a waste of time to check 98% of the images and then pass them. It would be silly to actually read a file for each image so the filter information should either be in a database associated with the file name or else a filter should be checked for the file name - basically depending on whether the number of filters is fixed or if they are user defined. Dmcq (talk) 15:39, 22 July 2012 (UTC)
 * 4 out of 200 is 2% which would equate to 260,000 images, but your sample size is too low for that sort of precision. I'd suggest that a more cautious extrapolation from your sample would put it at probably between 100,000 and 500,000 images. If we imposed one culture's NSFW definition then it would be much less. But if you were specifying a system you'd also need to factor in anticipated growth, and Commons is still growing quite fast. Incidentally could I tempt you to run an eye over meta:Controversial content/Brainstorming/personal private filters? I'd be interested in your views as to its practicality  Ϣere Spiel  Chequers  05:28, 23 July 2012 (UTC)