On the Feasibility of Homework Questions

Current policy is to close all questions that "feel like homework" to at least five (high-reputated) people. Those more militant even downvote answers given by others even if those were hints rather than complete solutions. I want to argue for stopping this crusade and allowing all kind of questions that are posed well (to be defined) and concern TCS.

First, rational reasons:

1. We cannot for sure distinguish homework from other questions¹
2. Getting help/inspiration is not always forbidden, sometimes even encouraged
3. We are not responsible for other people cheating
4. Some people are willing to help with low-level question

Second, opinions:

1. By providing help for homework questions we can reach new users
2. We should let people help that want to
3. A well-posed question deserves at least some hints
4. Such questions will continue to pop up, always, thanks to Google. We may as well answer them.

The only two reasons I ever read against homework/low-level questions are cheating and scope. Cheating I consider not an issue (see above) and scope should, imho, be defined by active users. If such answer low-level questions, so be it. Those who have some pure utopia in their minds should evaluate wether or not they really speak for most people and/or the greater good. Using RSS it is easy enough to preselect questions based on their titles.

Ad (1): Let me link to some accepted questions one might consider as homework level:

• -1: This site is elitist like MO trying to attract experts and researchers working in TCS, not undergrads. We have discussed it before several times. – Kaveh Nov 8 '10 at 23:08
• @Kaveh why do you want this site to be elitist? I would hope this site is welcoming to everyone including undergrads. I am still very much against homework questions because they will turn this site into a homework forum. But questions relating to TCS research should be welcome. – Lev Reyzin Nov 8 '10 at 23:16
• On opinions 1,2,3: they were not and are not the intentions of cstheory. Contrary to that, we don't want the site to be overrun by non-experts. On 4, MO is dealing with for a considerable time and they seem to be good at it, and we will be better in closing them as we get more high reputation users. – Kaveh Nov 8 '10 at 23:17
• None of the questions you link to seems homeworky. There is no way to quantify this of course, but "being basic" and "smells like homework" are not the same thing. I agree with Peter Shor (and others) that, when in doubt, we should ask for motivation. If nothing else, that will separate those genuinely interested from those who just want the answer already, dammit. – Aaron Sterling Nov 8 '10 at 23:20
• @Lev: Let me clarify what I said, I want this site to be elitist in the same sense that MO is, all interesting questions to researchers in TCS are welcome in my opinion. – Kaveh Nov 8 '10 at 23:20
• @Aaron, I think we really should put something about homework questions and providing motivation in the FAQ. @Kaveh, my feeling is that math is a pure science and TCS is an applied science, and as an applied science, we cannot let ourselves lose touch with the applications. – Peter Shor Nov 8 '10 at 23:22
• On rational 2,3,4: (I don't agree with them but even putting this aside) they are not among the objectives of this site and what it intends to be. On 1, we don't need to be always correct, and there are ways to increase the probability of making the right decision: don't rush to answer elementary questions posted by new/drive-by low reputation users, ask them for motivation and background and what they have done to answer their questions themselves. – Kaveh Nov 8 '10 at 23:27
• [continued] I think MO has dealt with these issues quite successfully, if there are enough users who want to have an elementary/homework level site, they can try math.SE or start something similar to it (as people on Math.SE have done). – Kaveh Nov 8 '10 at 23:28
• ps: Since you have linked to two questions that I have answered I want to add this: asking for an answer to a elementary homework question is completely different from asking for insight/intuition. I am perfectly OK with the questions asking for explanations of this kind when the OP shows that they have learned the technical part/proof, but does not see what is going on behind the senses. A very good example of this is the question answered by Russel on padding, this is something that one usually cannot find in textbooks and is not given as assignment. – Kaveh Nov 8 '10 at 23:42
• [continued] They are interesting even to researchers or at least graduate students trying to understand what is going on, what are the limitations of a technique, what are its advantages, why it works there but does not work here, ... I think they can be interesting to researchers and therefore they are welcome. These are completely different from homework/elementary questions. – Kaveh Nov 8 '10 at 23:49
• @Kaveh: Regarding "elitism": I see this site as similar in a way to Dick Lipton's blog. The level of technical discussion is challenging, and sets a high bar, but Lipton goes out of his way to be welcoming to people who post "dumb" comments. If we do the same thing, we can maintain a high standard of discussion, while at the same time be welcoming of broad participation. – Aaron Sterling Nov 8 '10 at 23:49
• @Aaron: I understand what you say, but I prefer the level on MO, there are dumb question on MO, but they are not inviting them. By the way don't forget that Dick needs to approve the comments made by a new commenter before it shows up on his blog. A closer example would be Lance and Bill's blog at the time they were not moderating the comments and the high level of annoying comments made some people stop reading it. – Kaveh Nov 9 '10 at 0:02
• The researchers have the right to have a research-level site, that was the reason this site was proposed, a TCS version of MO, and we have spent considerable time to get this going and on convincing other researchers to use it, it is obvious that we don't want it to become useless for us. I seriously don't understand why you want to morph this site into something it was not intended to be. There is Math.SE and the people there are completely welcoming to elementary/homework level TCS questions. (in the case that you are not happy with Math.SE, you can propose a new site on SE area51.) – Kaveh Nov 9 '10 at 9:24
• @Raphael, the site is community managed as you see and new users can gain reputation by contributing to the site and then start helping in managing it, I think you don't understand what community managed means. As far as I understand, the intended community that this site is trying to create is researchers in theoretical computer science, others are welcome if they are interested, but they are not the main group that this site is aiming at. This is not a playground, and I think you have a strong misconception about what this site is about if you think so. It is for helping researchers – Kaveh Nov 10 '10 at 1:03
• @Kaveh: Exactly. I would also argue that helping people with easy non-homework problems is good for us theory researchers in two ways, as long as it doesn't swamp the site. First, it enhances the reputation of theory, and thus makes it marginally easier for our grad students to get jobs. Second (and I saw this at Bell Labs/AT&T Labs) every once in a while someone will ask a question they need help with which actually turns into an interesting theory question (e.g., my paper on the Rectilinear Steiner Arborescence Problem). Homework problems don't do either of these things. – Peter Shor Nov 11 '10 at 21:04

The danger of allowing homework problems, besides the ethical issue, is that we may be swamped by low-level, easy homework problems, and this forum may degrade into a homework-solution forum.

The danger of banning low-level non-homework problems is that people who are not theoretical computer scientists may stop asking questions here, and we may miss connections with practice. For a notable example, by far the most important recent advance in information theory came from the interaction of theory and practice, and I would like to encourage as much of this interaction here as feasible. Most non-TCS people won't be able to tell the difference between an easy and an interesting TCS problem.

Of course, we may have to answer a hundred low-level questions from practitioners for every interesting one we get, but I don't think that's bad, because (1) Answering low-level questions doesn't take much time, (2) It's being done on a purely volunteer basis, and (3) We'll end up helping lots of people this way.

I don't think there's any chance we will be swamped by low-level problems if we don't take homework, and homework problems have no possibility of leading to important theoretical advances. The problem is distinguishing homework from non-homework.

One way of distinguishing homework problems might be asking (in the FAQ's) people to give some small amount of motivation or background for their problems. That way, if something looked suspiciously like a homework problem, we could ask for more motivation. I'd be interested in seeing a little bit of background or motivation for some of these problems, anyway.

• +1: I see people on twitter complaining about the Physics.SE exactly for this reason and asking for a "research-level site". – Kaveh Nov 8 '10 at 23:12
• "We'll end up helping lots of people this way." <-- providing a good argument for the value of theory, and theoreticians – Aaron Sterling Nov 8 '10 at 23:16
• @Kaveh: I just looked at the Physics.SE site. It's in private beta until tomorrow, but you can see some of the questions on the public part, and their level is very worrying. – Peter Shor Nov 8 '10 at 23:48
• @Peter: I think that may have been me he saw complaining about it. I'm doing my best to fulfill my commitment, but once that is done, I will very likely stop using the Physics.SE. I was really hoping for a TP version of here or MathOverflow, but instead it is all pop sci questions, which renders it useless to me. – Joe Fitzsimons Nov 9 '10 at 6:17
• Distinguishing homework from non-homework is top-priority, then. Motivation could be a good indicator but can be faked. Also, are follow-up questions ok? "We had this and that as homework, now I wonder..." And what about higher-level homework? – Raphael Nov 9 '10 at 7:58
• Can closed questions be edited? Will edited closed questions be pushed to the top again? Meaning, can we close questions until they meet the minimum standards wrt wording and motivation and then reopen them, regardless of level? – Raphael Nov 9 '10 at 8:06
• @Joe: I thought that it might be inappropriate to mention your name in my comment, sorry. – Kaveh Nov 9 '10 at 9:37
• @Kaveh: Nothing to apologize for! I didn't know for sure if it was me you had seen. – Joe Fitzsimons Nov 9 '10 at 9:47
• One point: From a purely practical point of view, if a student figures out how to fake homework questions as non-homework well enough to fool us, they will presumably also fool all the students who are looking for good places to get solutions to homework, and their questions won't turn into a flood. – Peter Shor Nov 9 '10 at 11:53
• It seems reasonable to ask for clarification when the question smells like homework. In my experience someone asking a genuine question won't usually mind expanding their question based on the comments asking for clarification. Someone who is going to stoop to the level of asking for an answer to a homework question is likely to need a quick response to make an assignment deadline. So delaying answers to such questions by asking for clarification results in both better genuine questions and filters out drive-by homework questions. – András Salamon Nov 10 '10 at 15:47

I think it's a little unfortunate that the question has conflated two separate issues that Peter masterfully disentangled (insert bad quantum joke here). As Peter says, homework questions BAD, non-homework low-level questions FINE.

Consider the question of NP-hardness that the question mentions. The OP specifically indicated that they were blocked on an NP-hardness question related to their actual work, and Tsuyoshi's comment was all they needed. This might be wishful thinking, but if we can encourage people to ask NP-hardness questions and help with answers, more applied researchers might start doing this on a regular basis, which is great for theoryCS.

But homework questions cause a flood of more homework questions: this has been well documented on both SO (which has a homework policy) and MO (which doesn't allow it). I am not interested in visiting a site that only helps with homework, because it doesn't help me with my work, doesn't teach me anything, and doesn't even help with the larger goal of theory outreach. Secondly, homework questions are usually posed by drive-by questioners who have no interest in sticking around and helping grow this community, so I see no reason to cater to their needs.

To say that since people will continue to ask homework questions, we should answer them is analogous to saying that since cranks will continue to post P vs NP proofs on comp.theory, we should entertain their posts. We all know what happened to comp.theory and I have no desire to let that happen here.

Just to reiterate: this is not just a moral issue, and in fact all the reasons I cited above are purely selfish. I want a site that I find useful and can derive value from, not one where my only task is answering homework questions.

• "We all know what happened to comp.theory" -- Actually, no, we don't all know what happened. I take it that it something VERY VERY BAD. – Aaron Sterling Nov 9 '10 at 2:57
• sorry. my age is showing. just go visit it right now. it used to be a good site before the cranks took over – Suresh Venkat Nov 9 '10 at 3:14
• +1 for "I am not interested in visiting a site that only helps with homework." I think that is the crux of the matter. – Robin Kothari Nov 9 '10 at 4:03
• You acknowledge these are selfish reasons but you will downvote people that feel differently? – Raphael Nov 9 '10 at 7:59
• The problem is, I think: five people >500 rep that are not interested can close-vote, ignoring the opinion of potentially many that have less reputation. Those people might never get enough rep to vote for reopen if they are mainly interested in/able to answering low-level questions. This yields a self-fulfilling prophecy, kind of. – Raphael Nov 9 '10 at 8:08
• yes, if the only questions one wishes to answer are homework questions (not low-level - that's different), then one will have a hard time gaining enough rep to change the policy. That's just tha nature of how community voting works, I'd say. For now, it seems like our policy on homework will stand, given the lack of support for your proposal. – Suresh Venkat Nov 9 '10 at 9:07
• Let's not forget we had multiple discussions about this already, we came to a decision, and we just sent off an article to a publication stating no homework questions are permitted. At the beginning, I didn't care about this issue, to be honest. I do think @Peter and @Suresh make compelling arguments in their answers to this question. But even if they didn't, I'd oppose the inclusion of HW at this point, because it would mean we would be lying in our marketing of the site. Not exactly the kind of reputation I'd like to build. – Aaron Sterling Nov 9 '10 at 14:18

Let think about it as a TCS problem. Most "homework question" are easy to spot, these are the one that are $\epsilon$-far from being at the research level. Then there is a grey zone of "hum, is this homework, or is this a question of someone not in the field?", for this one some questions stay, some are killed by the crowd, and I think it's fine.

Originally, the site is about research level question, and in my opinion it is OK if some basic question stay as long as their density is low. To be more clear, I won't keep reading/asking/answering if I have to spent too much time to separate the wheat from the chaff.

Concerning non TCS researchers that could be disappointed by this policy, I think this is not a real problem. a homework question by a student and a basic question by someone of an other field are clearly stated in a different manner (more documented for instance, and the latter is likely not to use the usual jargon).

Finally, you are saying that "scope should, imho, be defined by active users.", this is currently the case, unless you mean something I don't understand. You can vote for reopening question, and if enough persons are in agreement the question will be reopen, so I don't really see your point.

• Good point. closed questions can be reopened if enough people think that way. – Suresh Venkat Nov 9 '10 at 4:31
• and have enough rep. – Raphael Nov 9 '10 at 8:10
• clearly - otherwise a spammer could create 5 fake accounts and reopen his own questions. – Lev Reyzin Nov 10 '10 at 13:06