It doesn't happen often, but I've occasionally seen questions that raise ethical concerns (but are on topic). A couple examples come to mind: 1) (on stackoverflow) Someone asked about designing a system to defeat CAPTCHAs. This could be used to set up many fake accounts, send spam, and (as has recently come to light) surreptitiously manipulate public opinion. 2) (here on stats) Someone asked for help with a machine learning system for making decisions in what sounds like an industrial/lab environment, where the consequences of a bad decision include "things blowing up". This isn't a situation where someone should take (or need) random advice from the internet, and mistakes could change (or end) lives.

Clearly such cases have to be handled on an individual basis, and it's complicated for many reasons:

  • It's not always clear what the situation is. In example 1, the poster could have been a security researcher (but didn't mention this). In example 2, the poster could have been framing things in a hypothetical or exaggerated way. Meanwhile, asking for clarification doesn't always lead to answers, and others may post answers in the meantime.

  • We also want the site to lead to good and not harm. But we don't want to have (or be) moral police.

  • Public discourse has great benefits and censorship has great costs. In some cases, the lines can be grey. For an extreme example, see the recent controversy about publishing work on the avian flu virus, which was artificially mutated to increase transmissibility.

I think people will already make nefarious or dangerous choices on their own, but I don't want to help them down that path. I also want there to be freedom to discuss technical issues, even when serious consequences are at stake. If there's a concern, I think a good practice would be to ask for clarification, write comments to make clear what's at stake, and shut the question down if there's reasonable suspicion that specific harm may come to pass.

I'd like to hear others' thoughts on these issues. Has anyone seen similar examples? On a lighter note, is this an appropriate use of the 'flag for moderator attention' button? And, is there a way to put a question on hold until it can be determined what's going on?

  • 4
    $\begingroup$ I think this is a very interesting question, which has an abundance of examples in the field I work in: biomedical research. Imagine a decision support system for the overall continuation or withdrawal of medical treatment. 'Supporting' one of these decisions may lead to an irreversible state of death, while the other may lead to a severely impaired remainder of one's life. I think the question we (CV community) should ask ourselves, is whether we want these issues to be on-topic and intertwined with the statististics, or whether this is for the philosophy or ethics stackexchange. $\endgroup$ – IWS Oct 3 '17 at 7:12
  • 11
    $\begingroup$ Perhaps more mundane, but very commonly there is an ethical dimension to many questions here, even if there is no serious practical consequence of jumping one way or the other: 1. How to advise people who are inclined to identify and omit outliers as awkward for a simplistic analysis (and may have been advised that's a good idea). 2. How to report analyses based on cherry-picking from a bundle of significance tests. 3. How to report analyses presented as if first conceived but selected after a multitude of decisions based on looking at the data and trying a range of models. $\endgroup$ – Nick Cox Oct 3 '17 at 8:23
  • 3
    $\begingroup$ law.kuleuven.be/citip/blog/… $\endgroup$ – kjetil b halvorsen Oct 3 '17 at 10:49
  • 4
    $\begingroup$ My guess is that this question has no answers (yet) because it already discusses extremely well the main points to be made. $\endgroup$ – Nick Cox Oct 4 '17 at 11:06
  • 1
    $\begingroup$ @NickCox I agree, the question is well put and both of us have indeed given some examples of similar experiences (first Q in the last paragraph). But, the other questions have not been answered: do we think moderators should be notified of ethically laden questions (through flags perhaps)? Or should these questions be put on hold? (what are ethical questions?) I assume OP has put these forward as mere suggestions, alluding to a general point of what to do with ethical dilemmas which could influence the answers. I believe an answer to that question could still be valuable here. $\endgroup$ – IWS Oct 4 '17 at 11:49
  • 4
    $\begingroup$ @IWS I agree, but I don't see that we need more machinery, e.g. a different kind of flag. The issue doesn't appear to arise commonly (as do questions that are too broad, unclear, etc.). The objection to some questions of this kind might be that they just invite answers based on opinion. I would rather each thread is considered on its merits and that we don't try to erect meta-criteria. In other words, how to answer ethical questions as a category depends on a criterion for "ethical".... $\endgroup$ – Nick Cox Oct 4 '17 at 12:03
  • $\begingroup$ Another example like your second: stats.stackexchange.com/q/142954/17230 $\endgroup$ – Scortchi - Reinstate Monica Oct 4 '17 at 13:49

This turned out to be too long for a comment so it is an attempt now to answer the last two questions.

The examples which occur to me are ones where the OP has conducted a project where human subjects have been subjected to at best inconvenience and at worst the possibility of harm and the OP clearly has absolutely no idea of what to do with the data and so in my view should never have been allowed to start the project. If the results are published with defective analysis subsequent patients may be harmed too. Such studies in most jurisdictions should have had prior ethical approval of course but ethics committees often lack statistical expertise. In most of these cases I think the existing flags are sufficient (unclear, too broad, ...). I would not want to bother the moderators with an extra burden here.

  • 1
    $\begingroup$ Yes, it's true that research in some fields must be approved by an internal ethics committee. But, I don't think this covers the full gamut of ethically questionable situations we might encounter here on the site. E.g. the examples I mentioned in the question probably wouldn't have involved such a committee. Nor would questions @NickCox mentioned about cherry picking, outliers, etc. $\endgroup$ – user20160 Oct 10 '17 at 2:23
  • 1
    $\begingroup$ You raise a good point about moderator burden. One thing we'd need to know is: how many questions require some kind of ethical assessment? I'd guess that they're not too common, so may not impose a huge burden in terms of numbers. But, another aspect of burden is the time, effort, and weight of responsibility involved in assessing ethical issues. I think these could be much greater for assessing ethical issues than for determining whether a question is off topic, too broad, etc. $\endgroup$ – user20160 Oct 10 '17 at 2:24

I see two components, one specific, the other general, both important.

Specifically: To look at how other fields have dealt with this problem, I would point to bio-ethics, which (as a commenter notes) is its own discipline. This is for a concrete, practical, reason -- the morality of some questions simply can't be addressed, unless you're also a technical expert. For example, it would totally be ethical to encourage the donation of both kidneys, to save somebody else's life -- IF, you could survive without kidneys! You can live after you lose one. Why not two? It is precisely because of a medical fact (you need at least one kidney) that makes it unethical.

This is the reason that "give it to the ethicists" is not a general solution. I would be in favor of migrating those questions that do not require technical expertise to evaluate, but don't know how often such a question would arise here, and think statisticians might often be a bad judge of that requirement.

More fundamentally, and when we can't off-shore it, statistics is simply full of ethical problems that only we can assess. What are the ethical problems with not cross-validating a model? How about ignoring outliers? Assuming correlation implies causation? Fudging correlation/causation arguably destroyed Greece's economy. But beyond that ...

Generally: What we do has consequences. As statisticians, and as people. This cannot be avoided. And what we do here is powerful. A few short months ago, Facebook said Targeting is Good! Rules are bad! Now, Facebook says Sorry we turned out the "how to burn Jews" vote! Maybe, some rules! This WILL happen, here. I'm not going to wait for a disaster to conclude that I should consider my actions.

What does this mean for the site? It means that I'm completely in favor of a flag. To me, Flag a Moderator simply means This is Important, and Difficult, and I'd Like Help. Doesn't mean you'll get it. Doesn't mean the question is actually unethical. Doesn't mean anybody knows the answer. But at LEAST, it's something I want to be able to say.

Since I'm also cognizant of the vast amount of work moderators put in (often though, in areas like Culture, which is pretty ethical already), well, Case 1, there are a small number of questions, so it's manageable. Case 2, it turns out there are a ton of questionable questions -- well then it's a really good thing we started looking for them, no? And, to learn from the biologists, I would be completely in favor of having a Statistical Ethics Moderator, to accompany the Technical Moderators, and appropriately divided the specialized work.

Because it IS a lot of work, and it is hard -- so I want to give it the resources it deserves.

  • 3
    $\begingroup$ +1 a follow-up question: do you think a single moderator, or a small group of moderators, can and should decide on the ethical issue raised within such a question? I think this could be problematic if the moderator has moral views of his/her own (which of course the mod is allowed to have!) Maybe the flag shouldn't be directed at the moderators, but at the community at large, asking for a thorough evaluation of the ethical issue in combination with the statistical aspects (I also agree that for Q's lacking a statistical aspect migration would indeed be more suitable). $\endgroup$ – IWS Oct 12 '17 at 10:02
  • 2
    $\begingroup$ I think the biggest point to make is that some, human, must, eventually, decide. Above all, that's really the single point I'd like to emphasize, since the conversation feels like it's going in the direction of, Nobody should ever make an ethical judgment, that would be unethical. Who, among the rest of the community, would then be responsible? I think a Help! flag, in general, would be useful, but the moderators are the ones we rely on now to respond to flags, and especially since the community chooses the moderators, and I think that balance makes a lot of sense. $\endgroup$ – one_observation Oct 20 '17 at 18:38

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .