Wikipedia:Editing scientific articles

Editing scientific articles
Wikipedia's content policies are often unambiguous when working with sources which use ordinary language and everyday concepts. But scientific terms are much more precise, and require much more care. When writing about science and mathematics, accuracy is paramount. Following the letter of Wikipedia policies often is not enough to guarantee that an article won't contain serious factual mistakes or misleading statements. When editing scientific articles, bear in mind the following:


 * 1) Check any non-trivial statements you intend to insert into an article. Determine whether your statement could be invalid under some circumstances. To find out, you may need to study the entire source in which the statement is made, or look in other sources. The validity of a statement made on some particular page of a technical book may well rely upon necessary conditions mentioned many pages earlier, or even in another source. If you find that the statement is valid only within a specific context, you should take the effort to find out what this context is, and to include that context in the article.
 * 2) After checking carefully, you may find that a statement you want to insert still disagrees with other statements made in the same article. It may be that the conflicting statements are true under some conditions not explicitly mentioned in the article. Any apparent conflicts should be worked out by the editors through discussions.
 * 3) In resolving technical disputes, the exact wording of quotes from sources is often unhelpful. It is essential that the editors sort out the scientific issues from first principles as much as possible, referring to all the ideas in the sources, not to out-of-context fragments. This means that editors should read the technical literature with the goal of acquiring a full understanding of all the relevant points while editing the article. It is important that every disputed point be explained clearly, so that any remaining disagreements which appear in the encyclopedia reflect actual diverging points of view in the literature, not the misunderstandings between editors.
 * 4) When editing a scientific article, be careful to be as complete as possible. Filling in intermediate steps which are omitted from more condensed literature sources is not OR. Rephrasing content is not synthesis, as long as the ideas faithfully represent the ideas in the sources; remember that “Carefully summarizing or rephrasing source material without changing its meaning is not synthesis—it is good editing.”
 * 5) Assume from the outset that multiple meanings of technical terms are likely to occur, whether or not you are aware of them, so search for meanings proposed by other editors, rather than searching only to back up your own understanding. The goal is for the editors to gain a comprehensive familiarity with as much of the literature as possible.
 * 6) Different approaches or explanatory models are often all correct, and different readers will find different explanations useful.  Don't delete existing explanations just because they use a different model; add your explanation to the article, so long as it is carefully sourced.
 * 7) Discussions from first principles or evidence, alongside citing sources, are not violations of the ban on original research if they are conducted on the talk pages.  Such discussions have two benefits:
 * 8) educating less knowledgeable editors to the obvious benefit of the article
 * 9) identifying science cranks. Although cranks, by definition, cannot be swayed, such discussions are useful for identifying them as cranks, so that further remedial action can be taken.
 * 10) The perception of what is original research varies according to the level of expertise of the editor; experts often do not consider it original research to provide the logical connection between sourced premises and sourced conclusions; but this might be disputed by someone less familiar with the topic.
 * 11) Experts should be wary of deleting material that seems "obvious" to them, and take care to avoid the over-use of unexplained or underlinked jargon.  Remember, not everyone has a PhD in the topic.

The truth does matter
Contrary to what WP:V suggests, whether editors believe a statement is true is of crucial importance when editing a scientific article. The reason why WP:V depreciates "truth" is because, in the humanities, it is customary to interpret "truth" as mere opinion, as well as because the postmodern view that objective truth doesn't really exist is commonplace there. These views are not the norm in the hard sciences. The Sokal affair demonstrates very clearly what is wrong with this attitude. But since most of Wikipedia's articles are not about such subjects, we can understand why WP:V depreciates the truth.

Depreciating truth does have many supporters among editors who are active in scientific topics. This is because dealing with editors who have a wrong understanding of a theory, who are very fanatical about defending their POV, is then a lot easier. However, we need to keep in mind that the focus of WP:V is not on dealing with disruptive editors, we have many other policies for this. Indeed, in discussions on the policy talk pages when amendments are proposed, one is often asked to demonstrate why the current wording does not work. But examples where disruptive editors are involved are rejected as good examples, the argument being that an editor hell bent on having their way typically cannot be persuaded by what a policy says.

Many editors have expressed the opposite opinion that truth is an important factor to consider. Truth in terms of the current understanding of the existing scientific literature is one of the most important things to consider when editing articles. Truth is certainly not to be depreciated.

One may ask what the big deal is with "truth" if we demand verifiability anyway. The issue here is the way verifiability is defined on Wikipedia. The policy WP:V demands that anything challenged or likely to be challenged be attributed in the form of an inline citation that directly supports the material. This is problematic if an editor who is not familiar with a certain theory, would challenge a statement that can only be verified by studying the theory in its entirety and not by a single citation to a source that directly supports the statement. Such a case will be rare and exceptional. Editors who have indeed studied a scientific theory in its entirety are very few and are certainly not the majority of Wikipedia editors. Editors who have indeed studied a scientific theory in its entirety will readily produce adequate multiple citations that, when immediately and validly combined, provide reliably inferred support for a statement; if a Wikipedia entry that relies on such a combination is reasonably questioned, that entry should be removed until it has been agreed to by a consensus of editors, and even then it should be explicitly indicated in the Wikipedia entry that the combining inference is made by the Wikipedia. Such support by editorially agreed combination of multiple citations is always open to further reasonable question, and a former consensus cannot stand against fresh reasonable questioning.

Editors who do this in a disruptive way are of no concern considering the relevant policies dealing with disruption. The big issue here is that constructive editors can be unfamiliar with the basics of a certain theory, and they may well demand in good faith that certain statements, which are unverifiable by merely giving single inline citations, be clarified. One can then argue that this isn't a problem either; just give the citations to all of the necessary existing literature from which the statement follows. However, this would violate the synthesis policy that forbids combining statements given in different sources to arrive at a conclusion that is not given explicitly in a reliable source. The problem here being the clause "explicatively", which prevented one from giving a single inline citation directly supporting the statement in the first place. Clearly, one needs to violate the WP:Synthesis policy when necessary for verification purposes.

We can gain a deeper insight into this problem by considering a model of Wikipedia editors that the core policies implicitly assume.

The bot model of Wikipedia editors
Consider how one should program a bot to edit Wikipedia in the way we are used to doing. Let's focus first in non-scientific articles where there arguably are no issues with the core policies. Clearly, such a bot would have to be an artificially intelligent entity, and will thus contain a huge amount of information to simply be able to understand the meaning of simple statements in sources. We thus can't assume that the bot is a simple transcription monkey.

Let's call the information present in the bot "internal knowledge". This internal knowledge is essential to correctly interpret what a source says. It is then possible for the bot to stick to the core policies, provided the internal knowledge is sufficient to understand information form sources. One then doesn't have to have a concept of truth that applies to information from sources, truth only refers to the internal knowledge. In case of many bots collaborating with each other, all that is needed is that the bots have approximately the same internal knowledge.

The policies as currently formulated, model real Wikipedians as such collaborating bots with similar internal knowledge. They don't mention the internal knowledge explicitly, as this is taken for granted and considered to be irrelevant. That the internal knowledge is a relevant factor, even in the ideal case where the model seems to work ok., becomes clear if we consider statements that are already verified by the internal knowledge and where verification from sources is inappropriate. Take e.g. the statement "the sky is blue". Entire essays have been written on this problem on Wikipedia, but they all miss the core issue about the relevance of internal knowledge.

The bot model can go wrong when different editors have radically different internal knowledge. Correctly interpreting statements in sources from which an article is written can require some editors to acquire additional internal knowledge. Verification of statements without violating WP:Synth can then be impossible.

Conclusion
Wikipedia editing should both respect truth and provide reliable sourcing.

WP:SYNTH is not a rule against proper logical explication from stated reliable sources: it is a rule against inference or innuendo that combines source materials in a way that is loose, reasonably questionable, or unsound, or that misrepresents reliable sources.

If a Wikipedia statement is logically implied by reliable sources but not actually explicitly stated in them, and if there is reasonable question of the immediacy and validity of the logical implication, then the Wikipedia entry should explicitly indicate that the statement is inferred by Wikipedia. Unless verification of a statement is to be found in reliable sources either explicitly, or implicitly with valid logical inference from them—either immediately and unquestionedly inferred, or in the Wikipedia entry explicitly indicated as inference by Wikipedia—, then the statement would be original research and is not admissible as a Wikipedia statement.

Editing Wikipedia scientific articles requires editors to be careful, diligent, and knowledgeable. The requirement for truth cannot justify laziness by Wikipedia editors, on the spurious ground that a reliable source is too hard to find, or that a hardly reliable source is good enough. It may require editors to engage in extensive discussions on the talk page to ensure that what they write in the article is consistent with the current scientific understanding of the topic that is to be found in reliable sources. An editor who knows the current scientific understanding will be familiar with enough current reliable sources to provide verification from them. Someone who can only parrot the current understanding does not know it. Scientific knowledge is not merely verifiable, it has actually been verified, directly or indirectly, by the knower.