# Towards a policy for hint-only answers

This question is getting old, but the matter is still unresolved. Please contribute by voting on the answers here, or by voicing your opinion if you have something to add to the existing answers!

Our help center provides basic guidance towards writing good answers, including:

Brevity is acceptable, but fuller explanations are better.

And an important bit:

And only such, I daresay.

Now, given our homework policy (allow homework in the sense that we don't care where a question comes from), we get a lot of questions of the form

Now, was far as our help center is concerned -- and our own gut feeling -- this does not make for a good question. Full answers are unlikely to help the asker beyond gaining exercise credit.

So, being a helpful lot, our community has developed a compromise between "ignore and/or shoot down" and "answer for glory": the hint-only answer.
There are quite a few.. If you scroll through that list, you'll see that many if not most answers are one-liners, with some exceptions.

Now, these are clearly bad answers. Granted, they might have done their job in actually helping the original asker without doing their work for them, but now we have a question-answer-pair that is almost useless for anyone landing there from, say, Google -- except that person attends the same course in the next iteration.

In essence, the problem is the contradiction of these two statements:

1. The Stack Exchange platform only works well with full answers.
2. We want to help people without giving full answers.

So, the goal of this discussion is to find out

• if the community agrees that hint-only answers are a problem, and
• what to do about it.

Feel free to vote to your desire; should we decide that we need to vote, there will be another thread with clearly defined rules.

• appears to me some ppl eg some high rep users have quite a bit of rep due to "hint answers". the upvoting suggests that they are useful to someone although that benefitting audience may be narrow (maybe nearly only the original asker). so a better analysis would try to figure out quality via upvoting if these are acceptable on the site. seems overall, while not all that worthwhile or having much value eg wrt se stds etc, they are not a "problem". – vzn Jun 15 '15 at 15:13
• – Raphael Jun 15 '15 at 15:22
• I'm sympathetic. Thanks for the well-reasoned argument! Before I form a strong opinion, I could use some help understanding better how to think about questions that explicitly ask only for a hint. Do you have any suggestions or thoughts about what to do with them? Is your thinking that such questions are bad (i.e., a bad fit for SE), and they should be closed as off-topic/not suitable for this site? And then my follow-up query will be: What about questions that implicitly seem to be seeking only a hint -- e.g., "Can you tell me how to get started?" Should they be treated the same way? – D.W. Jun 15 '15 at 18:14
• The reason why I ask: It seems like what to do with "hint-only answers" is related to what we do with questions that call for hints. If we treat "hint-only answers" as bad, it seems like we're logically forced to nuke all questions that are looking only for a hint (as there's no way to answer such a question except in a way we consider bad). Are we comfortable doing that? – D.W. Jun 15 '15 at 18:16
• @D.W. Good question. I daresay that a good question that asks only for a hint will contain all the usual good stuff: independent thought, attempts, and such. They then ask for a hint on how to proceed. This should usually allow for richer answers then the type of question I assume above. Or the problem is actually hard and we don't expect a full answer, anyway; in such a case, a (list of) possible attack points may be a good answer (since best effort). "[Problem statement] Please give me a hint!", however, is just as bad as "I don't have a clue!", and should be nuked. – Raphael Jun 15 '15 at 21:41
• @D.W. Do you have examples of specific questions that trouble you? – Raphael Jun 15 '15 at 21:41
• @D.W. If I think about it, I guess the majority of all answers to "real" questions are rather elaborate hints. If the question is for logarithmic-time selection (in some context), "use AVL trees augmented by subtree sizes" is an idea that allows anyone above some level of knowledge/skill to derive a full solution, but in itself it's only an idea. Many TCS questions receive proof sketches as opposed to full proofs. Very rarely do we see very detailed, full answers to "real" questions. So yes, there is probably no strict line. – Raphael Jun 15 '15 at 21:49
• R, thx for linking the older relevant thread, just relooked at it; why are you reraising this when basically the old thread is exactly the same issue? and the top answer is by DC, ex cstheory mod (although apparently not much active anymore), in favor of hints as answers. it appears you dont want to follow the prior established convention/ consensus, ie overturn it with new policy. think that should be addressed somehow. re DW/ "questions that ask for hints", isnt that excessively rare? – vzn Jun 23 '15 at 4:40
• The older discussion is three years old; things and opinions have changed. Also, we did not establish policy back then, which is my goal now. – Raphael Jun 23 '15 at 6:57

Here is my proposal:

We do not want hint-only answers.

If a question does not deserve a full answer (now), don't answer (now). Use the comment feature for providing hints. Down- and/or close-vote the question if and as appropriate, i.e. in accordance with current policy.

If you happen upon a hint-only answer, kindly refer the author to this policy, downvote and flag for moderator attention (so that it can be converted into a comment).

### Rationale

1. Questions that scream for hint-only answers are usually SE-bad themselves; if you think it does not admit/deserve more than a hint, there probably is something wrong with the question. We should not encourage bad questions.

2. Hint-only answers are SE-bad. They are motivated by the behaviour of the original asker (and reasonable as a response to a problem dump!) but this motiviation goes away quickly. In the long run, any question-answer-pair has to be independently useful to search-engine users.

3. If the hint is a good one, even a hint-only answer encourages the original asker to ask bad questions (because they got the help they needed).
Yes, this applies to hints as comments, too.

5. If you want to provide a hint now and answer later (maybe because you think the problem is interesting but you don't want to encourage cheating), you can comment now and post a full answer later. The hint + spoiler format can be useful.

In summary, the pool of people whom hint-only posts help is small, but posting such is harmful to the site in several ways.

• A possible (and maybe necessary) consequence is to convert most old hint-only answers to comments, provided the posters don't expand upon them in a reasonable, t.b.d. timeframe. – Raphael Jun 15 '15 at 15:25
• When you say "Down- and/or close-vote the question as appropriate.", can you be more specific about when close-voting is appropriate? Is your proposal that if the question only seeks hint-only answers, then the question should be closed? You say you shouldn't encourage those questions, but does this mean they should be closed as off-topic? I'm not opposed, I just think it'd be helpful to have an explicit statement of what this proposed policy is. I view the policy for what to do with hints-only answers as connected to the policy for questions that scream for hint-only answers. – D.W. Jun 15 '15 at 18:09
• @D.W. I left that open deliberately; define appropriate as placeholder for the policies active at the moment of action. So, at the moment, close problem statements (unclear); downvote problem statements plus some query; and so on. My thought was, if you think the question does not admit/deserve a full answer, there's probably something wrong with it. (Does my edit clarify sufficiently?) – Raphael Jun 15 '15 at 21:43
• how do we discourage the bad reading of 'bad' questions? the demand for explicit representation of questions makes perfect sense; the mathematical symbolism removes all possible semantic misunderstandings and when refined is optimally succinct. learning to communicate in this way is an art in itself and requires a broad knowledge base that, perhaps, a pool of people coming here with questions do not have let alone are capable of expressing original thought in. just an idea to consider. it is a shame that only people who care can be berated effectively ;) – Illimitable Jun 17 '15 at 9:11
• @Illimitable I don't think communicating in mathematical and/or CS terms is an art; it's a language skill like any other. As far as I can tell, most posters of particularly problematic questions (problem dumps) are students of CS, so they have to learn the language anyway. By demanding they actually do so, we may be doing them a favor. (Besides insisting on high post quality in the sense that experts -- those you want to answer -- get a clear picture fast.) – Raphael Jun 17 '15 at 10:22
• is poetry a language skill? ohhhoo interpretive. you do know i am being complimentary about what you are trying to persuade me of. whos relating what to what now? – Illimitable Jun 17 '15 at 10:36
• i am sorry, not "whos relating what to what" but "if it does not relate, down it is shot!" :D – Illimitable Jun 17 '15 at 10:51
• i guess this difference is rarely considered, rarely matters and even more rarely matters for long. i aim to show that this time of mattering is [excessively extending or needs addressing or [something?]]. we do our best eh? – Illimitable Jun 17 '15 at 11:12
• I just want to point out, that bad questions sometimes are result of bad answer politics. Someone has to solve task X, and have some idea. If he describe it and his problem Y with solution... he may receive answer like "Forget about your idea, you can do it easier...". So better to write "I have problem Y", because then answer will be about OP's problem. End in next comment. – Tacet Aug 8 '16 at 23:37
• You can say, that he could describe his idea... then he may receive answer "Your idea is wrong, because...". This answers are high scored, so other answers normally doesn't appear. It's better to OP, to write "How to solve Y?" without any explanation. Therefore I stopped flagging most question like this on MSE. I can understand questioners. – Tacet Aug 8 '16 at 23:38
• @Tacet "Better" in which sense? This is a Q&A site, not a homework solving service. We don't like questions that just dump a problem here, and I don't see why that should contradict a simultaneous dislike of hint-only answers. – Raphael Aug 9 '16 at 8:17
• I agree with Tacet. While used on homework assignments it is good point that this site is not a homework solving service, but sometimes it is evidently not a homework, at first blatantly ignoring the question some answers appear, then people get annoyed, flamewar begins, problem X Y comments, the answer that exactly follows the request is lost, then accepted, then nuked with downvotes from others. In such cases these answers are hint only (not on purpose) and become eligible for nuking - this is a contradiction. – Evil Dec 21 '16 at 6:34
• @Evil Can you link some examples of helpful and SE-good answers that got "nuked"? – Raphael Dec 21 '16 at 8:07

Many, even most, of the questions on this site are homework questions. Despite the homework policy, in practice the site is used mainly to help students with their homework, and indeed traffic varies predictably with the academic year. If you discourage such questions, traffic would decrease dramatically.

While the homework policy suggests that we do students' homework for them, I have never agreed with this sentiment, and will continue providing hints in many cases. You can count me out of this site if policy prohibits this behavior.

• "traffic would decrease dramatically" -- I for one don't care. Moderating fewer dumps saves me time (and almost all of them require attention). It's quality traffic we are after, anyway. "I have never agreed with this sentiment, and will continue providing hints in many cases" -- so you are saying, I see the policy but I disagree so I'll not adhere? That's harmful behaviour. I'm saddened by the fact that you as a user many will look to for reference have decided to set such a bad example on purpose. – Raphael Jun 18 '15 at 14:36
• "You can count me out of this site if policy prohibits this behavior." -- that's your prerogative. Note that my proposal at least does not forbid hints, it only forbids posting them as answers. We can use comments and chat for guiding people to insight (even though their TAs should do so). I'm all for helping people, but maintaining site quality comes first. (That's a somewhat sad thought, but without quality, the whole will not work.) – Raphael Jun 18 '15 at 14:37
• Anyway, you post a prediction and a personal preference. Do you have any idea about how to move forward in a constructive fashion? Or, is your answer to my first bullet of inquiry, "I don't see hint-only answers as a problem"? – Raphael Jun 18 '15 at 14:40
• Exactly, I don't see hint-only answers as a problem. The problem is that there is so much homework traffic, a problem which is shared with math.se. Quality traffic often belongs to cstheory or mathoverflow, while math and cs are, by and large, low quality sites, by design. – Yuval Filmus Jun 18 '15 at 14:42
• I find that view unnecessarily pessimistic and in case of cs/cstheory plainly wrong. We do get quality traffic outside the scope of Theoretical Computer Science, which is most of CS, after all. This site has certainly not been designed to be low-quality, and I for one won't stand for it becoming such. In particular not with the line of reasoning "the only good stuff is on cstheory.SE, anyway". – Raphael Jun 18 '15 at 14:57
• I agree that there are some good questions on cs.se, but in most cases they are out of scope for TCS not because they're not "theoretical computer science", but rather because they're too basic. More advanced questions tend to be migrated to TCS. – Yuval Filmus Jun 18 '15 at 18:24
• By the way, the original reason behind adding hint-only answers was to make sure that questions get (officially) answered. A comment doesn't count as an answer as far as the system is concerned. – Yuval Filmus Jun 19 '15 at 0:58
• Count me in with @Raphael. I'm not sure where the idea that CS.SE was designed for low-quality questions came from -- I don't think that was ever the intent. (And, at the risk of stating the obvious, I view "research-level" as orthogonal to quality. I've seen plenty of low-quality research-level questions, and plenty of high-quality non-research-level questions.) We get more than enough high-quality questions that are not suitable for CSTheory or MathOverflow, so amount of traffic is not a concern for me. Just my personal take/preferences. – D.W. Jun 21 '15 at 2:13
• By the way, is the statement "the homework policy suggests that we do students' homework for them" a typo? Did you mean "the homework policy suggests that we avoid doing students' homework for them"? – D.W. Jun 21 '15 at 2:14
• @D.W. Our site invites questions on all levels, so it is the site of choice for homework questions. It could have specifically rejected homework questions, but it doesn't, by design. As for my statement which you don't like, the homework policy says that we answer questions ignoring their homework status. If we ignore the fact that a question is homework and we answer it, then we are doing students' homework for them. This include many or even most of the questions on the site, like it or not. – Yuval Filmus Jun 21 '15 at 5:34
• I think you are victim to a few misconceptions. 1) low-level != homework. 2) don't reject homework(-inspired) questions != accept problem dumps. 3) Help with homework questions != doing their work for them. – Raphael Jun 21 '15 at 7:16
• YF has top rep on this site. strange that nobody "else" is supporting him on this. think mods should yield to "other" high rep users on some basic issues. feel sometimes mods here are too inflexible & dont allow difference of opinion and leadership by other top rep users, possibly forcing them to take dramatic stands as like above. would like to see more striving for agreement/ consensus, more conciliatory atmosphere. "find a way to agree". the constant refrain that mods are the top interpretors of se policy rings hollow sometimes. do concede that se design may exacerbate all this. – vzn Jun 23 '15 at 4:16
• however, have long also wished YF would show up in chat at least occasionally to hash out some of this "behind scenes" mgt/ vision so to speak (mod R is in chat frequently) & maybe other misc stuff too. think this is really lame that a top user of the site would threaten to leave the site aka "fall on sword" over something that seems not so big a deal.... also it sounds like YF has an issue with homework policy, which has been a frequent topic of chat/ mods also... you guys are all highly emotionally intelligent adults right? can this be worked out? chat as a viable "blow off steam" valve? – vzn Jun 23 '15 at 4:22
• Well, I wasn't really serious about leaving the site. I don't think any such attempts would have been successful anyway... – Yuval Filmus Jun 23 '15 at 4:56
• In some cases I also consider a hint a better answer than giving all details. For example "want to prove $L$ is regular" the hint "try to store XXX in the states" might be a better answer than giving the automaton. But it is hard to decide, and depends on how the question is asked and mostly on my mood I guess. It is in my opinion certainly better than closing as duplicate and referring to the standard answer that basically says "build a finite automaton". – Hendrik Jan Jul 20 '16 at 10:20

I think the issue can be resolved through rigorous semantics. A "question" in the form "I don't know where to even start, please help!" Isn't really a question at all. It's a request for help. Bad questions in this form can be edited into the form "What are the initial steps for starting to solve the problem...". This form has two advantages: 1) it is actually a question. 2) A "hint-only answer" is actually a complete answer to this question, since the hint will provide the initial steps for getting started. So semantically this solution holds. I think it is useful as well, because other people from, say, Google, may also be looking for how to get started.

• This does not solve the problem at all, it only hides it beneath a thin layer of posts fulfilling technical definitions. – Raphael Jun 16 '15 at 22:40
• will still get relevant answers. – Tobi Alafin Dec 23 '16 at 22:23

Edit: Somehow I thought I was on Math Meta. So my examples don't quite fit. But I think the overall point applies, and I'm out of time anyway, so I'll leave it as-is.

My personal feeling is that by holding the line on question quality, this problem will disappear.

The comment from Raphael clued me in to this opinion (though I'm not implying he shares it):

I daresay that a good question that asks only for a hint will contain all the usual good stuff: independent thought, attempts, and such. They then ask for a hint on how to proceed. This should usually allow for richer answers then the type of question I assume above. ...

Just a problem statement by itself that is purely a copy of a homework question with no attempts made to answer any part of it, should be closed.

One key point: Most homework problems are not questions about math. They may be questions about Farmer Brown's potato field, or questions about the value of X, or about the height of the rod or the volume of Johnny's swimming pool, but they are not questions about math. Sometimes they aren't questions at all, but only instructions: "Find the value of X if Y is greater than 23."

Thus if they are questions they are off-topic, as swimming pools, potato fields and such are not the topic of this site, and if they are instructions only they should be closed as "unclear what you are asking." ;)

(Lest this be perceived as splitting hairs, the above is very slightly tongue-in-cheek. I believe this is a case where a purely literal interpretation of the rules, as I describe, has the advantage of also being correct. I'm not saying "follow the rules because those are the rules," but rather "I believe the policy we should apply is ______, and hey, look, you can derive that from an overly literal interpretation of the rules!") :)

A question such as "How can you find the surface area of a geometric solid?" is perfectly valid. The answer could be applied to a question about Farmer Brown's potatoes; that's the nature of math.

• I don't agree that questions about mathematical modelling of concrete situations are not about math. And it is definitely not true that questions about computer science modelling of concrete situations are not about computer science: that's the definition of applied computer science. – Gilles 'SO- stop being evil' Aug 6 '16 at 9:37