Wikipedia talk:Request for comment on tone in medical writing

Wording
The RFC seems fine as is. I don't think there's any disagreement on the three issues specified below (i.e., I assume that both of us and the average person would answer "no" to all of them). If you want to keep those in, I suppose that'd be fine, but I don't think they're necessary to include.  Seppi  333  (Insert 2¢) 03:23, 28 April 2018 (UTC)
 * Is uncontroversial medical content, sourced to multiple MEDRS sources, likely to be challenged (WP:V)?
 * With well-sourced uncontroversial medical text, is there a difference between "are advised" and "should"?
 * Is there a difference (relative to medical advice) in saying "should not take X drug" and "X drug is contraindicated"?


 * Actually, the intro could probably use some minor revision. I'll tweak it now.  Seppi  333  (Insert 2¢) 03:26, 28 April 2018 (UTC)
 * Done. My suggestion to delete the full attribution column still stands, but if you want to go ahead and include it as another method for contrast, I'm fine with that.  Seppi  333  (Insert 2¢) 03:31, 28 April 2018 (UTC)
 * After going back over this, I think the issue needs further clarification. The real issue at hand involves prescriptive statements in particular and normative statements in general.
 * Prescriptive statements inform the reader what ought to be done by conveying permission or obligation to perform an action. Some examples of prescriptive statements are listed below (note that the verb "do" in the first two examples could be substituted for any other action verb; regardless of the action verb used, these would still be prescriptive statements because the modal verbs "should" and "must" convey an obligation to perform the associated action).
 * "An individual should do X"
 * "Practicioners must do Y"
 * "Z should be considered when B occurs"
 * "A patient's medical history must be taken into account when list of circumstances."
 * Normative statements convey a value judgment about something and/or specify the way things ought to be (e.g., the right/correct course of action); in other words, they evaluate things, such as actions and outcomes, against a norm. Note that all prescriptive statements are normative statements, but not all normative statements are prescriptive statements. E.g., "genocide is an evil action" – the example statement from WP:WikiVoice – is a normative statement because it evaluates the morality of an action; however, it is not a prescriptive statement because it does not suggest/evaluate a course of action. The statement "genocide is evil; it should never be committed" contains a non-prescriptive clause – "genocide is evil" – and a prescriptive clause – "it should never be committed – both of which are normative statements.
 * Now, keeping what I've said in mind, lets focus on the following two questions from the draft RFC (NB: I've modified these statements to more accurately reflect the issue). I don't think the other 3 questions are relevant to the type of article statements that are in dispute (i.e., I'm pretty sure you and I would both answer them with "no"); you can keep or remove those depending upon whether you're interested in what other Wikipedians think though.
 * Should an uncontroversial and well-sourced normative medical statement be treated (for purposes of ASSERT and WikiVoice) as fact or opinion?
 * Original version: Should uncontroversial and well-sourced medical text be treated (for purposes of ASSERT and WikiVoice) as fact or opinion? - rephrased for specificity; the issues involving this policy pertain only to normative statements.
 * Do statements that specify what action ought to be taken in specific circumstances instruct the reader on how to do something (WP:NOTHOWTO)?
 * Original version: Is asserting well-sourced and uncontroversial medical text instructing the reader on how to do something (NOTHOWTO)? - rephrased for specificity; the issue involving this policy pertains only to prescriptive statements. I don't think the sourcing and "controversial-ness" of a statement are relevant to this policy, so I cut that part of the question.
 * I think the first question is precise in the sense that it directly addresses the more general issue surrounding normative statements that are well-sourced and uncontroversial, but I don't think that the average Wikipedian understands what a "normative statement" is. As for the second question, I think the current phrasing accurately rephrases, in more understandable language, the question: "Do prescriptive statements violate WP:NOTHOWTO?".  Seppi  333  (Insert 2¢) 07:50, 28 April 2018 (UTC)


 * I have some "meh" around this. There is a certain way things are written sometimes, where things come off as prescriptive and i hear the concern in that.  But at the same time, our mission is to transmit accepted knowledge and in some cases there are things that one should do and should not do, that are indeed accepted knowledge. Like, oh, women who are pregnant or who might get pregnant should not take thalidomide.  This has gone over the top a bit. Maybe a lot, and I will be opposing, the way this is written.... Jytdog (talk) 04:25, 28 April 2018 (UTC)
 * I think it's a problem that the RfC is scoped very widely (to all "medical text") but what seems to be the actual concern is statements about treatment in conventional settings which imply some kind of recommendation. I can sort of see the point of the RfC in this narrower context, but implemented as written I think the current proposal would be generally very damaging - especially in the WP:FRINGE topics which I predominantly edit. Alexbrn (talk) 05:14, 28 April 2018 (UTC)
 * could you spell this out better for me as an example? Foggy brain; I'm not following very well today. (Again, found we have no article, but 1 in 6 who get the new shingles vaccine get VERY sick, and that I am ... would have been nice for the doctor to tell me that, since I must miss a nice event tonight :)  Let me know what I can do here to get to the point (do I really have to attribute, "you may need separate bedrooms" if you have REM sleep behavior disorder, and that sort of insistence is really garbling the prose at dementia with Lewy bodies, while on the other hand, Seppi does not attribute a pharmaceutical statement at amphetamine, while insisting that dengue fever do so. Unclear what Seppi's criterion are. If this RFC could cause harm in the fringe area, it might be better just to open an RFC at Talk:Dementia with Lewy bodies, to see if I really need to jump through these hoops to attribute bland statements ... that may be a better option, but then, the lead of dengue fever did not need the change that Seppi introduced, IMO.  So, first thing to decide here is whether we need a broader RFC.  Sandy Georgia  (Talk)  19:07, 28 April 2018 (UTC)


 * Seppi does not attribute a pharmaceutical statement at amphetamine, while insisting that dengue fever do so. No clue what you're talking about.
 * do I really have to attribute, "you may need separate bedrooms" if you have REM sleep behavior disorder, and that sort of insistence is really garbling the prose at dementia with Lewy bodies That's like complaining about how burdensome it is to attribute with a direct quote. If that's what's required by policy, you do it. I did it and I didn't complain about how hard it was to add like 5 words to the relevant sentences in two FAs, much less start an RFC to try to change a Wikipedia content policy.  Seppi  333  (Insert 2¢) 22:39, 28 April 2018 (UTC)

FWIW, I think a third relevant question to pose in this RFC is:  Seppi  333  (Insert 2¢) 11:54, 28 April 2018 (UTC)
 * Should statements that specify what action ought to be taken in specific circumstances be treated (for purposes of ASSERT and WikiVoice) as fact or opinion?
 * It depends, Seppi. It is not something that anybody thinking carefully will be able to give a thumbs up or down to.  You are on a bit of a mission here. Jytdog (talk) 15:12, 28 April 2018 (UTC)

We should be in no hurry to launch this, and make sure it is framed in a way that will result in an answer to the opposition that has come up at dementia with Lewy bodies. To that end, I have tried to generate enough examples to illustrate different classes of concerns.
 * a) The first three are examples that are unlikely to cause harm (get a driving assessment, may need separate bedrooms, and put away the guns).  Jumping though hoops to attribute these bland, inoffensive statements (for Seppi) is garbling the prose.


 * b) The next two are statements about medication, with real potential for harm (no NSAIDs for dengue fever and no amphetamine for people with allergic reaction or taking MAOIs). They each have three sources, no one disagrees, and attributing them would garble the text.  It is curious that you don't attribute your own statement about amphetamine, Seppi333, but want the dengue text attributed, and the dementia articles to attribute far more bland statements.  I also believe the change you introduced at dengue fever is not appropriate for the lead-- it goes in to unnecessary detail.


 * c) The last example, to me at least, is one that absolutely should be attributed at aspirin, even though now it is not. In children with Kawasaki disease, saying they can take aspirin when others can't raises the possibility of those children being harmed by aspirin.  That is the kind of statement that should not be in WikiVoice, IMO.

Perhaps others can come up with more illustrative examples of what kind of text needs attribution, and what doesn't. Seppi, I removed your "normative" edits, because that is precisely where the disgreement is. You have a different view of opinion v scientific fact as some others. The problem here is that Seppi will oppose this article at WP:FAC if I don't attribute these bland statements, and from my point of view, this amounts to one person imposing personal preferences. Sandy Georgia (Talk)  19:11, 28 April 2018 (UTC)
 * Well the question is evolving but as I understood it the proposal originally was that assertions in "medical text" would need to be attributed. So (taking your unfortunate current experience as an example) how would the following two sentences from the opening of vaccination be rewritten: "When a sufficiently large percentage of a population has been vaccinated, herd immunity results. The effectiveness of vaccination has been widely studied and verified." ? Alexbrn (talk) 20:15, 28 April 2018 (UTC)
 * Can I suggest that you explore whether you can transform prescriptive statements into descriptive ones? For example, I would have preferred to rephrase as  or something similar. Assuming that sources will support that approach, might that be a way forward? --RexxS (talk) 21:27, 28 April 2018 (UTC)


 * Thanks for the suggestion,, but I can't source those statements as you wrote them. They should but we have no evidence from sources that they do. So, simple statements, that will cause no harm to anyone, are becoming convoluted, and the prose takes on passive voice, simply to avoid using the word "should". Which is fully and well sourced. As Colin points out, there is no difference between "X is contraindicated for Y" and "X should not be taken for Y".  Passive voice will get flagged at FAC. Sandy Georgia  (Talk)  23:12, 28 April 2018 (UTC)


 * yes aiming for description is the right thing.  Even further in the direction, using the same references: "Driving ability may be impaired early in dementia with Lewy bodies and at some point it becomes unsafe for the person to drive.  Generally family members make the decision about when to stop the person from driving."   Something like that.. that describes instead of prescribes. In most of the instances that are causing trouble this is a style and approach thing that can be solved by aiming a bit differently. The words flow from the aim. Jytdog (talk) 22:02, 28 April 2018 (UTC)


 * But you didn't get in the main issue, which is that they should have their driving assessed early and often. Sandy Georgia  (Talk)  23:21, 28 April 2018 (UTC)


 * Expanding. one of the symptoms should be "impaired driving".  In the treatment or management section, where we describe management (i so prefer "management" to "treatment") it would useful to write: "Driving ability is assessed as part of management, starting soon after diagnosis.  Family members generally determine when driving privilges are removed" or the like. Trying to describe, it comes out differently.... You are kind of aiming prescriptively it seems, Sandy.  Please think about it. Jytdog (talk) 23:42, 28 April 2018 (UTC)
 * I can't think very well today (as someone who actually had dengue fever when I lived in Latin America, that's about how I feel now :) And I may be at my prose limit even on a good day! Driving is listed under Symptoms--> Core features, in the visuospatial function section.  I will give these a whirl once I am over the vaccination ... suggestions are helpful, as my prose is not stellar, but we need to stick to sources also.  Thanks Jyt, Sandy Georgia  (Talk)  23:59, 28 April 2018 (UTC)


 * As for what other kinds of text need attribution, you could ask about the sentence with the underlined value judgment in Wikivoice: For example, an article should not state that "genocide is an evil action", but it may state that "genocide has been described by John X as the epitome of human evil.". Notice how if you cut the underlined text, that opinion becomes a factual statement: "genocide is an action".  Seppi  333  (Insert 2¢) 22:39, 28 April 2018 (UTC)


 * Seppi, you are still talking about value judgments as if they were medicine, which has scientific facts, and facts that are commonly known and accepted (Jytdog's thalidomide example). Your point would be better taken if your examples stuck to medicine.  The underlying issue here is:  "NSAIDs should not be taken for dengue" is not a value judgment.  It's a statement based in scientific fact and backed by three sources, that no one disputes.  Sandy Georgia  (Talk)  23:21, 28 April 2018 (UTC)


 * Seppi, you are still talking about value judgments as if they were medicine - "Taking NSAIDs when you have Dengue is bad". You now have a value judgment pertaining to clinical practice.
 * which has scientific facts - "Scientific facts" have an identical operational definition as "positive statements"; they're both statements that can be verified from empirical evidence. The only distinction between the two is that the former is falsifiable but true by definition whereas the latter is falsifiable with no restriction on the truth value of the statement.
 * On a related note, are you suggesting that a statement which asserts an action that should be taken can be empirically demonstrated? How exactly do you demonstrate "should" empirically? Take a really simple statement: "Bob is running outside right now".  That can be empirically demonstrated by observing that "Bob" is performing that action in that location at the present time. Now, lets add a modal verb! "Bob should run outside right now".   If this type of statement is indeed a "scientific fact", explain to me how you would empirically demonstrate that "Bob should run outside right now". As a medically-relevant alternative, explain to me how the alleged scientific fact - "NSAIDs should not be taken for dengue" - can be empirically demonstrated.
 * "facts that are commonly known and accepted (Jytdog's thalidomide example)" - I've never heard that one before. Lets assume for a moment that an "opinion" actually could become a "fact" simply by being adopted by a large number of people. I don't think the number of medical professionals globally exceeds the number of people who think mass slaughter of an entire racial/ethnic group is "wrong/bad/evil". Frankly, one billion people globally is probably well below the actual number.  WikiVoice still refers to that assertion as an opinion though.  It also mentions how to describe opinions that are held by a large group of people.  It doesn't say "call them facts"; rather, it says: Usually, articles will contain information about the significant opinions that have been expressed about their subjects. However, these opinions should not be stated in Wikipedia's voice. Rather, they should be attributed in the text to particular sources, or where justified, described as widespread views, etc. For example, an article should not state that "genocide is an evil action", but it may state that "genocide has been described by John X as the epitome of human evil." I.e., according to WikiVoice, opinions do not become facts by becoming widespread.
 * FWIW, the only reason I've been talking about opinions as opposed to facts is that these are unequivocally not facts; well, maybe not in cases where a person is experiencing marked confirmation bias, but generally speaking it really is pretty unambiguous since it all follows from a small set of fairly understandable definitions that are easily used to validate statements.  Seppi  333  (Insert 2¢) 01:06, 29 April 2018 (UTC)
 * User:Seppi333, you danced around the thalidomide issue with a paragraph full of distraction. Please don't do that. Let me ask me you something.  The place I use "should" most often is contraindications.  "Contraindicated" is a big old polysyllabic word and I generally try to write in plain English.  In medicalese one would say: "Thalidomide is strictly contraindicated for pregnant women and women who might become pregnant, and is only prescribed with a risk evaluation and mitigation strategy".  Plain english: Women who are pregnant or who might become pregnant should not take thalidomide and there are strict regulations to ensure that this cannot happen".  Can you please suggest some plain english that you would find acceptable with regard to the woman? ("will cause birth defects to the fetus" doesn't get there)  Please answer directly and no, I will not use "contraindicated".  Thanks. Jytdog (talk) 06:36, 29 April 2018 (UTC)

I agree with some of Jytdog's comments above. This is an invented problem/campaign for one or two editors. The red-herring of normative language is mentioned again, despite nobody using that argument anywhere else, ever. RexxS makes the same mistake we earlier saw Tryptofish make at DLB, where medical advice "assessment of driving ability should take place early/regularly" was replaced with unsubstantiated fact that "assessment of driving ability takes place at an early stage/regularly". As I noted on the other page, this is as ridiculous as replacing "People should avoid becoming obese" with "People are generally slim". I think it is a good faith suggestion, but just doesn't work.

One wrong suggestion that keeps being made is that we replace advice with a collection of factoids we think either support the advice or are a consequence of following the advice. Well that perhaps is ok if what occurred was the Wikipedian themselves phrased the facts as advice (which is original research). But often the medical advice is the fact we need to include in the article. When we are writing a disease article, part of our job is to describe (where there is weight in the literature) the medical advice to doctors and patients. For example, for patients with DLB, the best practice is that patients should be counselled about how to make the bedroom a safer place. So part of the "treatment/management/prevention-of-harm" of that disease is that counselling should occur between doctor and patient/carer, and that counselling involves imparting advice on making their bedroom safer. It is not acceptable to baldly state a couple of examples like "Guns in the bedroom are unsafe" (true though this may be) or that "Lowering the bed makes a fall less likely to cause serious injury". Or to claim that people with DLB are at risk of injury in the bedroom, and leave the reader to work out what should be done. Or to list both and hope they join the two together. Missing out the "patients should be counselled about" (however we word that) is a bit like missing out antenatal classes from any discussion of pregnancy. And despite numerous sources repeating this advice about counselling, we don't have sources that say this actually does always happen. So we're left with the problem of either removing important facts from the article (that counselling is recommended) or finding a way to word it that doesn't upset two editors, or misleadingly in-text attributing that advice to just a couple of random journal article authors, even though it appears to be consensus and uncontested.

I don't think continued argument by existing campaigners or an RFC is likely to lead to any solution. Input and discussion from a bright spark with language might help. Or else just drop this whole thing as a distraction. -- 12:16, 30 April 2018 (UTC) — Preceding unsigned comment added by Colin (talk • contribs) 12:16, April 30, 2018 (UTC)

Three columns
I agree that the 3rd "Fuller attribution" column isn't needed and only confuses the issue. RfCs with 2 choices rather than 3 are more likely to result in some kind of consensus. Kaldari (talk) 18:24, 29 April 2018 (UTC)
 * Thanks, but here's the problem, as explained by  at WT:MED.  "Are you arguing that they must all be attributed? If so, that might lead to the problem of attributing to one source practices that are common and that any RS would support. SarahSV (talk) 01:52, 28 April 2018 (UTC)" and "I can't see where policy insists that uncontested positions be attributed. As I said, this leads to the problem of attributing to one source something that any RS would support. SarahSV (talk) 02:13, 28 April 2018 (UTC)"
 * So, by showing partial attribution, and full attribution, we show the problem with any attribution of uncontested statements. Where do we stop?  Perhaps I can better convey this issue in the RFC, but by trying to frame it neutrally, I thought it not right to say why three columns are shown. If we attribute to one source, do we have to attribute to all sources, any source, and if there are so many, why do we have to attribute at all?  Perhaps if I add a line explaining why three columns, but I don't want to appear to lead the !voter. Sandy Georgia  (Talk)  19:40, 29 April 2018 (UTC)
 * I added this, not sure if that leads the !voter. Sandy Georgia  (Talk)  19:43, 29 April 2018 (UTC)

3 questions
These questions ask whether a normative statement should be treated as a fact or an opinion in relation to WP:WikiVoice and whether a prescriptive statement violates WP:NOTHOWTO. That's really all I'm interested in knowing the community's answer to in this RFC.  Seppi  333  (Insert 2¢) 22:39, 28 April 2018 (UTC)
 * Should a well-sourced medical statement which includes a value judgment be treated (for purposes of ASSERT and WikiVoice) as fact or opinion?
 * Should well-sourced medical statements that specify what action ought to be taken in specific circumstances be treated (for purposes of ASSERT and WikiVoice) as fact or opinion?
 * Do statements that specify what action ought to be taken in specific circumstances instruct the reader on how to do something (WP:NOTHOWTO)?
 * I think that these opinions and the proposal itself assume more precision than actually exists in the English language. Wikipedia is a text for humans to read. The level of specificity which this proposal imagines is beyond what a typical reader can use and beyond what academic medical papers use.
 * In medicine the best available information is opinions. Whatever the consensus of medical organizations recommends is still an opinion because so much information gets obsolete with time. The reasonable thing to do is to hold the contradictory position that medical information is facts subject to change. Wikipedia should not qualify and disclaim its every statement because that that would be an extraordinary and inhuman practice. In the context of human culture we expect for any reader of any medical text to have the understanding that medical information is unlike statements of fact such as capitals of countries. Wikipedia's text is not a math equation and instead it should have a level of precision targeted to reader needs and matching normal English language.
 * That said, this is still an interesting proposal because critics endlessly and continually imagine that Wikipedia fails if its medical information is not a perfect equation with accuracy beyond the available published sources and human knowledge. It is great that someone raises the discussion and lays out the case for accuracy beyond human patience to comprehend so that we have a discourse to demonstrate that we deliberated the choice. Although I do not expect this proposal to pass overall it still could surface some major insights, related problems which we could address, or subcases where we do need more precision and attribution of voice. In the examples currently given I do not think that the suggested alternatives with higher attribution are better than the version with the contemporary wiki style of text.
 * I can also agree that Wikipedia's medical text violates WP:NOTHOWTO, WP:WikiVoice, Medical disclaimer, WP:MEDMOS, because we say things like "ibuprofen should not be used". That is two policy violations, which are each ultimate taboos, and a violation of the Wikimedia Foundation legal code, which is a mega ultimate taboo, then violation of the medical community's favorite own guideline, which is the worst of all. Still, the problem is in the text of those policies and guidelines and not in the text of Wikipedia's medical articles. Our first priority is to communicate to humans in a human way. We have no obligation to compromise English language to comply with some rules. Other medical texts have similar disclaimers and they have always used non-logical English also. Non-logical English is standard.
 * Everything in this proposal does makes an excellent model for how to frame information in Wikidata, which is an inhuman and alien rendering of text that is logical.  Blue Rasberry   (talk)  23:58, 28 April 2018 (UTC)
 * The first question begs a further question: Who says that the medical statement includes a value judgement? If the answer is "a Wikipedia editor", then the answer is the statement remains a fact per WP:WikiVoice. If the answer is "a second, equally reliable source", then the answer is that the statement is the opinion of the first source, and both the conflicting sources need to be reported and attributed, also per WP:WikiVoice. You don't need a RfC to answer questions that are already answered by policy. --RexxS (talk) 00:18, 29 April 2018 (UTC)
 * The first question begs a further question: Who says that the medical statement includes a value judgement? If the answer is "a Wikipedia editor", then the answer is the statement remains a fact per WP:WikiVoice. If the answer is "a second, equally reliable source", then the answer is that the statement is the opinion of the first source, and both the conflicting sources need to be reported and attributed, also per WP:WikiVoice. You don't need a RfC to answer questions that are already answered by policy. --RexxS (talk) 00:18, 29 April 2018 (UTC)


 * Blue Rasberry, I am confused by your use of "this proposal", as I tried to frame the thing neutrally. What proposal?  Seppi sees opinion where I see medical statements backed by a multitude of reliable sources and disputed by none.  I don't think we should need an RFC; I think our policy pages have this, but Seppi has a peculiar interpretation.  Sandy Georgia  (Talk)  00:47, 29 April 2018 (UTC)

I think that, for purposes of the eventual RfC, it's important to construct it such that any given question posed to the community should address only one "variable" at a time. In particular, it needs to be clear to responding editors when they are being asked about attribution (whether it is needed, and if so, to one or multiple sources), and when they are being asked about how-to issues (including the use of "should"). Those are really two separate questions. The samples I'm seeing as of this version all seem to focus on what would be best practice in terms of attribution. In the questions section, that corresponds to questions 1 and 4. But there are also questions 2 and 3, about nothowto and should, and I think there should be a separate set of samples for those. It might even be better to divide the RfC into two sections for that purpose. --Tryptofish (talk) 00:48, 30 April 2018 (UTC)
 * Thanks, Tryptofish; I am going to give that a try now. I may self-revert if I mess it up worse.  Sandy Georgia  (Talk)  01:02, 30 April 2018 (UTC)
 * I did that; not sure it is better, revert to here if not. The problem is ... I think ... that all of the examples apply to both sets of questions.  Sandy Georgia  (Talk)  01:17, 30 April 2018 (UTC)
 * It would take some work, but there would have to be a completely different set of examples. Sort of like:
 * Column 1 "Aspirin should not be given to children or adolescents to control cold or influenza symptoms, as this has been linked with Reye's syndrome."
 * Column 2 "Aspirin to control cold or influenza symptoms is contraindicated in children or adolescents, as this has been linked with Reye's syndrome."
 * Column 3 "Aspirin given to children or adolescents to control cold or influenza symptoms has been linked with Reye's syndrome."
 * --Tryptofish (talk) 01:33, 30 April 2018 (UTC)
 * I gave it a try, but can you fill out column C on Samples two? Sandy Georgia  (Talk)  02:40, 30 April 2018 (UTC)

diagnosis and management -- best practices vs description of what actually happens
this rfc is too ham-fisted and will not lead to anything useful.

First, this is really about the main medical activities - namely diagnosis and management, right?

So thinking about that...

in my view there are some places where we can skid over into prescriptiveness and ~perhaps~ shouldn't (!) and some places where "should" is just plain English
 * 1) "contraindicated" means "should not be used" in plain English.  Things are contraindicated because in 'doing them, there is a high risk of doing harm.
 * 2) on the other hand there are other things that are more "best practices" which are really centered on diagnosis and management, where not doing them increases the risk of harm due to misdiagnosis or lack of diagnosis, and poor treatment selection. There are actually two kinds of "best practice" recommendations out there:
 * a) sources like cochrane reviews, where some specific authors (who may have idiosyncratic views) review the evidence and make science-based recommendations.
 * b) The other is clinical guidelines put by various authorities.

I will also throw in
 * 3) descriptive content about what actually is done with respect to diagnosis and management.

I added 3 because it is something to keep in mind, always. And something that we often don't have any sources or data on. But take something like this (am just making this up, this is not from any actual content): "Antibiotics should not be used to treat a cold or flu, because those diseases are caused by viruses, not bacteria. Antibiotics are frequently prescribed for cold or flu". The first sentence is best practice (not to mention scientific fact); the second is what actually happens. Right?

Most WP content is descriptive, like #3. I very much wonder how much of our content in "diagnosis" and "treatment" sections is actually descriptive, and how much is best practices. Hm.

In any case, I think the controversy is around things in bucket #2. I haven't seen any head-on, common sense objection to writing contraindications as "should not" (item #1), but maybe that will emerge. And plain old description of what happens is not being contested (#3).

For the rest, in day to day to editing very often best-practice-recomendations match and we don't think much about attributing and just summarize what those best practices are, but sometimes they directly contradict each other, as in the case of Oseltamivir (tamiflu). It is worth having a look at Oseltamivir which is the result of a pretty fierce negotiation. We definitely attributed there.

There is also an issue of how wide the consensus is, on best practices. I have to imagine that in many, many instances there is little to no controversy about what should be done to diagnose and manage condition X. In some conditions there is a lot (eg tamiflu). I don't know if we regularly think about this. It is generally controversy that drives attribution, right? Part of what is surprising in Seppi's concern here, is the lack of sources recommending other best practices for people with dementia. That makes this all somewhat... abstract and academic.

Turning to the dementia/driving thing. SandyGeorgia, my sense is that this is a "best practices" thing, and not an effort to describe what actually happens. Above I had suggested making it descriptive ("driving ability is tested early and often for people the LBD" or the like) but now i wonder if that is even accurate - the nonprofit source that is cited seems to be describing best practices per 2b -- what "should" happen -- and not trying to describe how people with LBD are actually managed.

The interesting question here is how to handle "best practice" types of content. Right next to that, is considering how we discuss what actually happens vs how we discuss best practices - do we, and should we, distinguish? This is probably something that is worth talking through more. The RfC doesn't get to the meat of the question the way it is framed. Jytdog (talk) 20:35, 29 April 2018 (UTC)
 * Seppi disagreed that we can say "should not" in place of "contraindicated" (over at Talk:Dementia), so that's your Bucket #1. He seems firm on his ctrl-f search on "should" and "must".
 * On the driving thing, it happens that a friend of mine chaired an event on the East Coast to try to get more elder driving assessments, to make "taking the keys away" easier on the family ... I know from that conference that it's not done enough, and is a real source of family conflict ... and I notice that my sources don't say it "is done", but that it "should be done". So while I've tried to adjust the wording on the driving thing, the real recommendation is that it "should" be done, so I guess I must attribute, when everyone knows it should be done?
 * Can you figure out another way to position this RFC? Hopefully, it won't be needed, because still as far as I can tell, the view is not widely held.  Sandy Georgia  (Talk)  21:12, 29 April 2018 (UTC)
 * Here's a better example than driving from dementia with Lewy bodies, ."The Fourth Consensus Report continues to caution against the use of antipsychotics (neuroleptics) for people with DLB." I was expected to attribute that, meaning I picked one source, when that is what EVERY source says.  Antipsychotics kill people with DLB, but we have had to dull the "don't use" "should not use" down, making it vague to the point of useless.  And we end up with the problem pointed out by ; we make it look like that is a position taken only by one group, when it is widespread consensus. Sandy Georgia  (Talk)  21:25, 29 April 2018 (UTC)
 * I don't think contraindications are an issue when it comes to "should not"; that is kooky. This is what "contraindicated" means. Jytdog (talk) 21:28, 29 April 2018 (UTC)
 * Thanks for clarifying that your friend is a public advocate for more driving assessments. Yikes Sandy.  This is where we do have to be careful about SOAP (not saying at all that this is not good advocacy; we just have to be careful not to do it here) Jytdog (talk) 21:31, 29 April 2018 (UTC)
 * Yes, I have a friend who is a volunteer driver for the elderly with a program whose name I cannot recall, and I once attended a conference she organized because I was visiting her town at the time of her conference. I don't think I have a strong stance on the issue; in the article, my aim was to include everything I could that was mentioned in secondary sources, and that was one of them. I have had a problem with the wording per Seppi on every subject. As the sources explain, visuospatial function impairment usually occurs early on in DLB; we can't really ignore driving when it's a big piece of the condition. Seppi at Talk:Dementia with Lewy bodies said: "For example, "antipsychotics shouldn't be used in people with dementia" is a normative statement, but "antipsychotics are contraindicated for dementia" is a positive/factual statement; only the latter one is encyclopedic." And, "What I was trying to get at when I was talking about contraindication above is that we should not be using prescriptive language (i.e., sentences that use words like "can, could, may, might, shall, should, will, would, must, have to, has to, and ought to"." He objects to should not. Colin rebutted with: "The phrases "indicated for" and "contraindicated in" are merely medical jargon for "should be used to treat" and "should be avoided when". A difficulty with in-text attribution is that the reader may wonder if this is the only body/person holding that opinion, or wonder why this body/person is considered an authority on the matter, rather than someone else. " And that discussion went round and round.  The DLB article ended up with "usually avoided" in the lead.  I think/hope it is better explained in the text, but you cannot cover all the whys in the lead. Sandy Georgia  (Talk)  21:48, 29 April 2018 (UTC)
 * And, I deleted the other examples from DLB in the interest of brevity and to work in thalidomide; we can go back to a non-driving example if you feel I cannot be objective about the topic of driving because I have lots of friends all over the world who do lots of interesting things. Sandy Georgia  (Talk)  22:03, 29 April 2018 (UTC)
 * So you are not seeing the differences between 1) (avoiding doing harm) and 2) (optimizing outcomes with "best practices" which in turn can be based solely on evidence per 2a or on "wisdom"+evidence per 2b) and 3) describing what is actually done? In my view these three things ~should~ be considered distinctly in these discussions and not doing so will lead to a very difficult to apply outcome, or none at all. Jytdog (talk) 01:24, 30 April 2018 (UTC) (missing "not" Jytdog (talk) 01:41, 30 April 2018 (UTC))
 * I put in the Kawasaki example because that is a case where harm can be done (the only case of the six examples that I believe needs attribution, but I didn't want to say that), and took out the extra dementia with Lewy bodies examples, because they were all cases of 3, and I think we have 2's in dengue, aspirin and thalidomide. Perhaps the problem with the mucked-up-ness of the RFC is that I gave examples, hoping to let the !voter decide.  Maybe the questions need to be made more explicit, but I didn't want to lead the !voter.  Really, with everyone saying the RFC as framed is mucked up, you all should start editing to find another direction.  I sure don't have a good track record at framing RFCs. Sandy Georgia  (Talk)  01:37, 30 April 2018 (UTC)
 * I don't like this RfC (this is what i mean about FA causing drama))
 * I think too-broad questions lead to bad outcomes.
 * 1) and 2) need to be treated separately, and I think some discussion around how to write about "best practices" would be more productive than an RfC. I think.. Jytdog (talk) 01:47, 30 April 2018 (UTC)
 * Two things, just between you and me (don't tell Seppi :) I don't need the bronze star. Never did.  (Methinks he does.) I think if there's someone on Wikipedia who knows what is and isn't an FA, that just might be me, after reading through more than 3,000 FACs.  I did think it would be nice for the whole dang WP:MED gang to work on something collaboratively as they used to, and work with others I would bring in, and feel the satisfaction of building something together from nothing, but I see that spirit is gone.  So, I already didn't get what I most wanted. And I have no problem forgetting the RFC and forgetting taking an article to FAC.  I am not going to bow to unreasonable demands just for a star.   But, this issue was still contentious enough that several editors left the article because of this problem, and we still have Seppi putting forward a point of view at WT:MED that no one there agrees with.  Whether or not I can take an FA-quality article to FAC is really not the issue and not the problem.  If we can't frame it in a way that won't lead to bad outcomes wrt "fringe" stuff, then I don't really care. On the other hand, if you all can find a way to frame it, to resolve the clear impasse at WT:MED, edit away!   Sandy Georgia  (Talk)  01:55, 30 April 2018 (UTC)
 * It is not just fringe stuff. Thalidomide is not fringey!  At the article about TamiFlu Doc James and i did some serious head-butting (eventually resolved) on how to deal with the conflict between treatment guidelines and the Cochrane reviews.  Underlying that was this same 'should' stuff but we never actually framed it that way.. we ended up jumping right into attribution. I actually wonder a lot, at how drugs are actually used (so much of the biomedical literature is clinical trials which is not medical use...) and how people are actually diagnosed and treated, and very much appreciate clinical review articles where the authors talk about what treatments and diagnostic procedures are actually done.  There is interesting stuff to me, and important too --  in terms of (to kick over into some high falutin language) epistemology and just what it is we are actually communicating as we "summarize" our sources...  That is why i jumped in here. I am bummed about the "mission" and prescriptive tack that Seppi is taking; such a blunt approach misses too much. Jytdog (talk) 02:54, 30 April 2018 (UTC)
 * Well, there's nothing wrong with just letting this sit a while. I'm pretty much done with DLB, and could go ahead and ask Eric to copyedit regardless.  Like I said, what I wanted (a collaboration) I already didn't get.  And who knows, something may give.  When you don't know what to do, the best thing to do is ... nothing!  I feel like the RFC is too complicated for anyone to follow.  And I feel like there is already a demonstrated no-consensus for Seppi's ctrl-f "should" approach, so am not sure what else the RFC will yield.  Sandy Georgia  (Talk)  03:20, 30 April 2018 (UTC)

Scope
I think it's important to plan ahead of time about how the RfC would relate to existing policies and guidelines. Using WP:NOT as an example of what I mean here, it's reasonable to ask whether X or Y properly complies with the policy. However, I think many community members will, rightly, object if it sounds like the RfC is suggesting that there should be exceptions to the NOT policy, without actually being an RfC on a proposal to modify the policy itself. --Tryptofish (talk) 01:04, 30 April 2018 (UTC)
 * Viewed that way, we have tension between:
 * (NPOV) When a statement is a fact (e.g. information that is accepted as true and about which there is no serious dispute), it should be asserted using Wikipedia's own voice without in-text attribution.
 * and
 * (NOT) Describing to the reader how people or things use or do something is encyclopedic; instructing the reader in the imperative mood about how to use or do something is not.
 * If we accept (not all do) that "NSAIDs should not be given to people with dengue fever" can be asserted as uncontested fact in WikiVoice, and it is not an opinion which needs attribution, then ... perhaps we should ask the question ... are we describing to the reader how people do things, or instructing the reader about how to do things? Because in this case, they are the same, no?  Sandy Georgia  (Talk)  01:24, 30 April 2018 (UTC)
 * In WP:NOTHOWTO, that's verbatim what it says. But I cannot find those exact words at WP:NPOV. What I do find is:
 * Avoid stating facts as opinions. Uncontested and uncontroversial factual assertions made by reliable sources should normally be directly stated in Wikipedia's voice. Unless a topic specifically deals with a disagreement over otherwise uncontested information, there is no need for specific attribution for the assertion, although it is helpful to add a reference link to the source in support of verifiability. Further, the passage should not be worded in any way that makes it appear to be contested.
 * That's really quite different from what you quote above, and I don't think that it's at odds with NOTHOWTO. Did I miss something? --Tryptofish (talk) 01:47, 30 April 2018 (UTC)
 * The part you can't find is at WP:ASSERT. All of the links are on the RFC page.  There is (strangely) similar text in two places at NPOV:  WP:ASSERT and WP:WikiVoice.  The differences (I think) come down to one group sees uncontested text in the fact camp, and others in the opinion camp.  What I am saying is that, if we see uncontested, fully sourced, widely held consensus as assertable in WikiVoice, then we are describing to the reader how things are done, even when we use the dreaded word should.  Sandy Georgia  (Talk)  02:03, 30 April 2018 (UTC)
 * Thanks. The ASSERT part is actually at Neutral point of view/FAQ, which is why I didn't see it. --Tryptofish (talk) 17:36, 30 April 2018 (UTC)

structure
Since my opinion was solicited, I would say the structure f this looks fairly well thought out. However, since multiple questions are being asked, it may be better for each question to have its own section for !votes and maybe discussion as well. Beeblebrox (talk) 02:12, 30 April 2018 (UTC)
 * Thanks for looking in.  I kinda think it's a mess, so I'm glad you weren't completely dismayed :)  Sandy Georgia  (Talk)  02:17, 30 April 2018 (UTC)

Comment
I'm catching up and haven't read this in detail. But my initial impression is "Oh no, please not another vote". There are numerous wiki and meta articles on the evils of voting/polling. I strongly recommend removing the "vote" section. Also, recommend removing the closed questions that demand a yes/no response.

I'm not at this moment convinced there is a need for an RFC considering that the complaint that started this was driven by an editor who was confused about grammar rules that apply to academic/journalistic writing-that-advances-a-position for economics and philosophy. There also isn't much disagreement that we should try to avoid being blatantly obvious about offering medical advice in Wikipedia's voice, rather than indicating this is expert medical advice. Nor has anyone explained at all how "indicated for" and "contraindicated in" are at a grammar level any different to "should be used for" and "should be avoided when" except that they indicate expert medical opinion rather than the opinion of Wikipedia.

The difference between opinion and fact is not black and white, and anyone trying to apply policy to require anything that looks like opinion to be in-text attributed is simply being fundamentalist and unhelpful. Anyone following the UK news recently will know that even whether a certain very sick baby should be treated at all is a medical opinion, not a fact. So if it is even opinion to treat, then it is certainly opinion how to treat. Running experimental studies and using the best statistical methods for your RCT does not make the medical advice on treating/caring for people any more factual and doesn't turn the art of medical care into Newton's laws of gravity. I shouldn't have to in-text attribute Bayer Pharmaceutical to state that aspirin is indicated for headache.

Until the issue becomes more than the concern of two editors, I suggest not wasting more time on this. There are far far more important things to get right in an article, and plenty examples of how editors trying to edit text to meet these demands, have ended up making claims unsupported by the sources, or losing the whole point of the text. And those are worse problems. -- Colin°Talk 09:30, 30 April 2018 (UTC)
 * Good enough for me. I don't think I can db-author it, because I am not the only editor (Seppi edited); does anyone know?  Should I put Historical on it?  Sandy Georgia  (Talk)  12:49, 30 April 2018 (UTC)
 * What would be useful, is if editors with great language ability can contribute a variety of alternatives to the obvious "should". In the other discussion, alternatives like "are recommended", "is desirable" were proposed by some. It's passive voice, but active voice becomes clearly Wikipedia's voice. It doesn't change it from being an opinion, but the more we can shift it towards appearing to be an established consensus medical opinion (like "indicated for" does) then the less I think it is reasonable to object.
 * To take another example, what words mean is by very definition (unless you live in France) the consensus opinion of the population at a given time. It isn't a fact like the chemical formula for sucrose or Boyle's law. Occasionally we have technical terms defined by some body (e.g., the definition of "epilepsy" by the ILAE) but mostly word meaning arises organically and only works when we agree and know enough others agree. We don't go around saying "Orange is (according to the Oxford English Dictionary) a colour between red and yellow in the spectrum of visible light". Yet at some point in the 1500s, we started using that word rather than red. You Americans are of the opinion that "pants" are long outer garments worn on the legs, whereas us British are of the opinion they are underclothes. And then there are words, like religious terms, where the meaning really is closely tied with a particular world-view, and means different things to different people. So I maintain this "opinion" thing is a spectrum, not black and white. At one end, where we all agree on the opinion of what a word means, almost like it is a fact, and at the other end, we have to in-text attribute its meaning to Seventh-day Adventists, for example. -- Colin°Talk 14:15, 30 April 2018 (UTC)
 * The most difficult tangles at DLB were two:
 * People with DLB should have driving assessed early and regularly
 * Many people tried to reframe that, but the sources didn't back the text in any case. All the secondary review sources say is that "they should".  In finding a way to rephrase it, supported by sources, it was necessary to pull in the LBDA source, which is not a source I would typically choose. ("Driving ability may be impaired early in DLB because of visual hallucinations, movement issues related to parkinsonism, and fluctuations in cognitive ability, and at some point it becomes unsafe for the person to drive.[51] Driving ability is assessed as part of management and family members generally determine when driving privileges are removed.")
 * Firearms should be removed from the bedroom
 * We could pull in two case reports to explain why DLB docs have this concern (the combination of visual hallucinations and acting out violent dreams per REM sleep behavior disorder, leading to people dying), but the secondary review does not cite the case reports, so I don't feel comfortable pulling them in (relative to SYNTH).
 * I think we might better focus on why we have particularly targeted the word "should" at WP:MEDMOS, because using the word "should" is just common sense in some cases. Like, if you have visual hallucinations and act out violent dreams, you should lock up the guns.  Sandy Georgia  (Talk)  14:40, 30 April 2018 (UTC)
 * I think the overall text at MEDMOS "Do not address the reader directly. Ensure that your writing does not appear to offer medical advice ... Statements using the word should frequently provide inappropriate advice" is reasonable taken as a whole. The active voice with "should" sounds more like addressing the reader, as Wikipedia, to give advice. A passive voice "People with DLB are recommended to ..." doesn't command the reader to the same extent. And it is factual too, because they are recommended, by their physicians, to ... A problem part is the oversimplistic advice at MEDMOS: "instead of plain statements of facts" which has been interpreted as some as "just remove the advice and leave the reader to work it out from 'facts'" which is unsatisfactory in many cases or "assume the advice is followed" which is plain wrong and OR. There are plenty cases where the best course is for Wikipedia to dispassionately state the facts and not begin to suggest what the advice is in a certain case: Wikipedia is not a how-to manual for intensive-care for example. In other cases, the medical advice is itself a notable fact that is worth mentioning. The dogmatic wikilawyering and misapplication of inappropriate grammar rules has combined here with a rather naive assumption that writing this sort of thing is easy, or that this issue is the most important concern for medical article writers. A little knowledge is a dangerous thing? -- Colin°Talk 08:28, 1 May 2018 (UTC)