tea.mathoverflow.net - Discussion Feed (Structural bias on MathOverflow)

gilkalai comments on "Structural bias on MathOverflow" (20939)

2012-12-20T23:15:05-08:00

I dont think that structural changes in MO are needed or will make a large difference, and, in particular, I dont regard the closing policy as important in the context of our discussion about bias. When it comes to new areas we saw that sometimes the enterance of a single person can make a large change. (The newer versions of stackexchane have a slight advantage in showing most reputable users in a short time-window which allow new users to be more visible.)

Mark Meckes comments on "Structural bias on MathOverflow" (20928)

2012-12-20T12:00:05-08:00

@Ryan: I'm not sure that Laba meant to suggest that MO emulate the specific features of Google+ mentioned in the page she linked to. As I understood her comment she was using that page simply to demonstrate that, in her words, "there are actually things that site owners and moderators can do to make [a] site more attractive" to a particular audience, not necessarily to suggest specific things that should be done here.

Michael Bächtold comments on "Structural bias on MathOverflow" (20926)

2012-12-20T10:29:38-08:00

I agree that if we use the 32 categories here http://arxiv.org/archive/math it might be aninfinte process, but maybe we could group some of these categories together to have roughly 5-10 areas in the beginning, and maybe there are already enough experts in each area. But I don't want to insist on the idea. Just wanted to make a suggestion.

François G. Dorais comments on "Structural bias on MathOverflow" (20921)

2012-12-20T09:39:07-08:00

Emil identified the key problem with this idea. As in the early days of MO, moderators would have to actively moderate questions from less popular fields until a critical mass of users in that field make their way to 3000 points. Given the number of distinguishable fields, that process would last forever and would be severely taxing on the moderators. As it stands, it's not a realistic proposal.

Michael Bächtold comments on "Structural bias on MathOverflow" (20919)

2012-12-20T09:37:37-08:00

Yes I meant what Mariano said. I admit that I haven't thought about all the implications of such a rule and the details of implementing it. Concerning Emil's last comment: I suppose in the early days of MO there was a similar situation, so it would just be a matter of time for enough high rep users to accumulate in an area.

Emil J comments on "Structural bias on MathOverflow" (20917)

2012-12-20T09:13:44-08:00

I also understood Michael to mean “active on MO in that field”. This can be tracked by the software easily enough using tags of questions and answers posted by the user, as can be seen already on current MO user pages, but I have no idea whether SE 2.0 supports restricting rights based on this information. In any case, I think this is not such a good idea. For one thing, it might mean there are not enough high rep users in less popular fields to do this kind of administration.

Mariano comments on "Structural bias on MathOverflow" (20916)

2012-12-20T08:45:27-08:00

Micheal might mean «active on MO in that field» as opposed to «with more than two MathSciNet pages» :-)

Todd Trimble comments on "Structural bias on MathOverflow" (20914)

2012-12-20T07:03:23-08:00

I really don't like that idea, Michael, because it sounds like that judgment would be highly publication-dependent. It is quite possible for users to have acquired some level of expertise or good judgment in an area where they haven't published at all. (I'm also not sure how that would work for those who are active but prefer to remain anonymous, such as our friend quid.)

Michael Bächtold comments on "Structural bias on MathOverflow" (20912)

2012-12-20T02:57:05-08:00

Concerning the situation of question getting closed by people who are not neccesarily specialists on the field: could Stackechange 2.0 identify which high reputation users are active in which field and only give these the right to vote to close in that area?

Ryan Budney comments on "Structural bias on MathOverflow" (20911)

2012-12-19T20:12:14-08:00

The example she gave is of Google+ compared to Flickr. This gives you the option to shut people out, and invite people in. So if enough people shut one individual out, they're effectively ostracized from the community. This is a little more extreme than what MO does with its reputation system and a little more personal, it would require people to make judgements on whether or not to shun an individual or to give them more access to the community. I suspect most people would consider this prohibitive to passive participation on MO.

Timothy Chow comments on "Structural bias on MathOverflow" (20908)

2012-12-19T19:13:25-08:00

I don't visit meta very often but this is an interesting thread.

I see that on her blog, Izabella Laba says that "there are actually things that site owners and moderators can do to make the site more attractive." What are these things? I didn't see specific suggestions but I read only one blog post and its associated comments. Perhaps she can be persuaded to give specific suggestions even if she does not want to participate on MO.

The list of subject areas by popularity on MO seems to correlate fairly well (though not perfectly) with the list given by Joseph Grcar in his December 2010 Notices article on "Topical bias in generalist mathematical journals." To the extent that there is a correlation, it suggests to me that much of the subject-matter bias on MO is not a function of MO structure per se, but more of a reflection of pre-existing biases in the mathematical community.

One question that I have not seen explicitly addressed is, given that there is subject-matter bias on MO, why does that bias need to be altered? I don't see the practitioners in the fields that are less represented on MO saying that they want MO to serve their sub-community better. For example, it's conceivable to me that those other fields have a better mechanism than MO to address the needs that MO is trying to address, and that MO's "bias" gives just the right amount of help to certain fields so that there is now perfect balance across all subfields of mathematics, once all mechanisms are taken into account. While I don't actually believe that there is perfect balance, I say this to illustrate the point that I'd like to see more evidence that the existing bias ought to be changed.

Will Jagy comments on "Structural bias on MathOverflow" (20874)

2012-12-17T12:10:57-08:00

fedja, I had a short email exchange with Francois after my long post. The relevant part would be that Francois and Anton are considering various kinds of outreach. As silly as that always sounds, it is in a helpful direction. You have described very well the circle of under-representation within MO, few askers means little for experts to answer means fewer experts mean fewer asking, something like that. Change entirely from within has real limitations in this case.

Also, you make a very cute victim. We could have an MO picture calendar.

Michael Greinecker comments on "Structural bias on MathOverflow" (20859)

2012-12-17T00:14:07-08:00

I think the subproblem of people closing questions as too elementary they do not actually understand can be met somehow and reduce the bias towards "areas with lingo that sounds sophisticated". In the Stackexchange 2.0 system, you get an actual description of the closing categories. So one can have a category "too elementary" and a description that says one should only use it if one can easily answer the question oneself completely.

fedja comments on "Structural bias on MathOverflow" (20857)

2012-12-16T23:20:14-08:00

Well, since we are discussing the "field bias" and, as an analyst, I'm assumed to be a victim here, I'll comment a bit on it.

First of all, what partly amuses and partly irritates me is an occasional outright lack of elementary analytic education in question posters and voters to close alike. The posters can still be excused because, after all, the whole point of asking a question is to discuss something you don't know, but a short comment with a totally wrong answer followed by the suggestion to close the question as too elementary can look REALLY strange.

Second, the fact that the only real answer to any of my mathematical questions that I could accept with satisfaction was given by myself can hardly improve my general opinion of the usefulness of MO as a research tool for me.

Can any of these two problems be solved by tweaking with the software? I doubt that very much. Will people change their behavior and will I get all my questions answered in the near future? Sorry, but I doubt that very much too.

Why am I still here then? Well, what I derive my satisfaction from is my ability to help other people out occasionally, from trying to solve some unusual problems I normally do not deal with in my professional life, and from some metamathematical discussions. I do not think any of those will go away soon, so I'm hooked pretty well for now. Do I feel uncomfortable on MO? Certainly not. If anything, I feel too comfortable here. What I usually do is just to completely ignore all that algebraic geometry and category theory stuff and look at the rest, which is more than enough to keep me busy during my free time. Would I like to see more analysts on MO? Yes, but there is absolutely nothing you can do with the software, reputation points, etc. to attract them like there was absolutely nothing you could do with such things to attract me in the first place. It is the content of the site and the general atmosphere plus my personal attitudes and preferences that matter for my decision whether to enter or not and whether to stay or not, and no single person, be he a moderator, a system administrator, or a user can change them.

Of course, there are things the admins can do to improve my experience on MO. The first one is to make LaTeX fully compatible with the markup features of the site and to introduce something for drawing pictures (if they don't know how or what, let them look at AoPS, where these problems had been solved to everyone's satisfaction long ago). The second one is to disable that infamous captcha for users with sufficiently high reputation. There are some other technical matters as well, but they are less irritating so they can wait.

The only people who can correct the "underrepresentation" of analysts on MO are analysts themselves and they can do it not by telling horror stories like how terrible it is that nobody in their field is here, or how topologists and algebraic geometers dismiss analysis questions as not suitable for the site, or how the typical number of points for a good answer to an analysis question is far below that for a question about categorification of something, but by entering one at a time and starting doing some real work here. I'm happy to be able to say these words from the "victim side" because when I say them from the "offending" or the "neutral" sides, they are never really heard.

François G. Dorais comments on "Structural bias on MathOverflow" (20852)

2012-12-16T16:55:41-08:00

Izabella Laba posted a follow-up post on her blog.

Harry Gindi comments on "Structural bias on MathOverflow" (20819)

2012-12-14T13:21:24-08:00

I stand by my position in the earlier arguments, albeit not their (sometimes) unfortunate choice language.

That is, I see all of these ideas as solutions in search of a problem. We've seen rep-hiding suggestions to practically every single perceived problem on MO.

Also, the idea that a little friendly competition (through rep scores) would make women feel uncomfortable sounds patronizing in extremis, but what do I know.

Scott Carnahan comments on "Structural bias on MathOverflow" (20816)

2012-12-13T19:43:36-08:00

Dear testcomment: as long as you are testing, please check that the "edit" link next to your post works, and expand your clever comment into something that is better developed and constructive. At the moment, your intentions are not completely clear. For example, are you suggesting that we should choose a population that does not participate in MO, and have them conduct this discussion instead?

testcomment comments on "Structural bias on MathOverflow" (20808)

2012-12-13T08:49:39-08:00

If MO really does have this bias, then it's the wrong place for this discussion.

Alexander Chervov comments on "Structural bias on MathOverflow" (20805)

2012-12-12T22:08:10-08:00

If the join of say http://scicomp.stackexchange.com/ and MO happens than there will be more user on MO who are in applied math. However I would think that since http://scicomp.stackexchange.com/ was created, people find this is not a good solution. Why ?

Nilima already pointed out one reason - "aggressive/unprofessional behaviour".

But I do not think this is the only (main) reason.

I just browesed 10 questions on top of http://scicomp.stackexchange.com/ I would say that 8 will be welcomed on MO, but

http://scicomp.stackexchange.com/questions/1194/how-much-better-are-fortran-compilers-really

http://scicomp.stackexchange.com/questions/4831/where-to-find-data-to-start-a-graph-data-structure-to-practice-on

These two would probably be immediately closed on MO.

And this problem seems to be unsoluable. It is the same that I would think that join of MSE and MO is not good, since
there are just toooooooooo many questions on MSE I would not be able to find MO-quality level questions in such a joined site at least it would not be easy.

PS
@Benjamin
"I know countless examples (like the one quid linked to above) where questions were closed by people outside the area because it somehow hit their radar as being below the level of the site."
Is it really many such questions ? My feeling that it is like exceptions...

Nilima comments on "Structural bias on MathOverflow" (20803)

2012-12-12T20:54:22-08:00

Will Jagy pointed me to this interesting discussion. I think it's an important one to have.

I honestly believe the broad membership of MO doesn't have any particular axes to grind, or none that one is conscious of. This makes it challenging to suggest structural changes.

There are two specific reasons I don't participate except as a passive reader on MO. Neither warrants (or suggests) a structural change.

I personally don't bother posting questions to MO, because the forum doesn't have enough of a critical mass of people working in numerical analysis/scientific computing to get expert answers quickly. This means any question I post would need to have a more detailed background/explanation of technical terms than is efficient for me. In some sense I've contributed to the 'tragedy of the commons' by not participating. The comp sci stack exchange is a better forum for me.

I'm somewhat old-fashioned and am not comfortable with what to me seems aggressive/unprofessional behaviour. The Internet has its own norms and many MO users are comfortable with these. I'm not. I dislike the dismissiveness and troll-like behaviour which occasionally surfaces on MO and this forum.

A specific episode involving a post by Doron Zeilberger (see http://tea.mathoverflow.net/discussion/1091/how-many-integer-partitions-of-a-googol-10100-into-at-most-60-parts/ for a gory history) proved a bit much for my taste.

The moderators do a heroic job, and most of the users are courteous. But some are not. Even though I dislike their commentary, they have a perfect right to it. I cannot advocate a structural change to limit discourse I don't like. Instead, I have simply wandered off to a forum which was more welcoming, both scientifically and culturally, for me.

This post is already very long, for which I apologize! Again, my best wishes to the site, I think it is a valuable resource.

bsteinberg comments on "Structural bias on MathOverflow" (20793)

2012-12-12T10:16:40-08:00

@Francois,

I did say it would be a big change and I am not 100% sure I even support it but I thought it was a good idea to bring out the fact that comments are in some sense cheap and therefore too easy to make. You may remember a long time back I had made a comment that a question about an exercise in Thurston's book would have generated less votes to close/comments if the questioner's background and motivation was presented and the questioner, a new user, suggested my comment was made because her username was a woman's name. This was not my intent and I had previously made similar comments on male users questions. Now I have learned the hard way that comments can be construed in very negative ways by the OP regardless of the intent and they are best used with care. I suspect comments have the most power to lead to misunderstandings and a sense of bias of any structural component on MO.

François G. Dorais comments on "Structural bias on MathOverflow" (20791)

2012-12-12T09:37:02-08:00

The other structural thing, but this would be a major change, would be to have upvoting/downvoting on comments and have this affect reputation. If people could loose a fair amount of rep for dismissive (but not blatantly offensive) comments, then you might get less of them. I think such comments can discourage people.

There are reasons why the comment system is the way it is but I definitely think there is room for improvement. I'm glad you brought this up and I would love to hear more opinions about this.

Currently, comments can only be upvoted and this has no impact on reputation at all. One of the reasons for this is to de-emphasize comments. The other is that comments are sorted chronologically and not by votes, except for the selection of comments that appear in the unexpanded view.

Another issue with comments is the much smaller font size used, which may be problematic for some users. I haven't heard complaints, but I do remember a big mayhem because someone misread one letter in one of my comments.

Donu Arapura comments on "Structural bias on MathOverflow" (20790)

2012-12-12T09:33:14-08:00

(Willie: I was using a hyperbolic metric.)

bsteinberg comments on "Structural bias on MathOverflow" (20789)

2012-12-12T09:14:57-08:00

My feeling is that somehow the biases in MO are not structural in the sense that it is something that can easily be changed by tweaking the software, removing the reputation system or changing the moderation style. I think many of the biases are biases that exist throughout the mathematical community. For instance, is the percentage of women using MO comparable to the percentage of women in Algebraic Geometry and Number Theory? I have no stats on this but my personal experience would lend me to believe that the percentage of women in these two areas is less than the percentage of women in the general math populace (does anybody have numbers on this?). This could explain the bias in MO to some extent, if true. Is the representation of topics on MO similar to the representation of topics in top general journals like Annals, Duke, or Inventionnes? I suspect that again algebraic geometry and number theory enjoy some favor in these journals beyond their representation in the populace, while most likely applied math is underrepresented. I think Igor Pak had on his blog some figures on this for Annals.

There are two structural things that could be done, perhaps, that might make MO a bit friendlier to underrepresented groups and areas but it comes with a price. One thing would be to increase the threshold of votes needed to close. Given how many people can close at this point, I think this wouldn't help HW/spam questions to survive but it would give reasonable questions from outside the mainstream a better shot at survival. If nothing else it could give more time for an expert to leave a comment "hey, this is a reasonable questions, why all the votes to close?" I know countless examples (like the one quid linked to above) where questions were closed by people outside the area because it somehow hit their radar as being below the level of the site. Some how a string of differential equations whose notation is not explained can seem below level to pure mathematicians while equally unexplained terminology used in questions on (infinity,1)-toposes or Quillen model categories would never seem below level. The other structural thing, but this would be a major change, would be to have upvoting/downvoting on comments and have this affect reputation. If people could loose a fair amount of rep for dismissive (but not blatantly offensive) comments, then you might get less of them. I think such comments can discourage people.

WillieWong comments on "Structural bias on MathOverflow" (20788)

2012-12-12T09:10:02-08:00

@Donu: two divided by a finite number is not infinitesimally small. Or if you really mean infinitesimally small: where do you work? I'm sure after hiring infinitely many mathematicians they can fit another one in!

Donu Arapura comments on "Structural bias on MathOverflow" (20787)

2012-12-12T07:23:55-08:00

Well, I don't have any deep insight into this. But it is clear that an infinitesimally small percentage of all mathematicians participate in Mathoverflow, so this no doubt magnifies "historical" imbalances. As far as I can tell, in my department, which is fairly large, there is only one other person who regularly participates on Mathoverflow. He's an analyst, and I am, unfortunately,
yet another algebraic geometer.

Steve Huntsman comments on "Structural bias on MathOverflow" (20786)

2012-12-12T06:56:48-08:00

@ Alexander -- Users are not interested in it because there is not a critical mass of interested users. I editorially quote Nilima from http://tea.mathoverflow.net/discussion/1052/ (the meta discussion of the question cited in quid's previous comment): "At any rate, given the relative paucity of numerical analysis on MO...compared to, say, algebraic geometry or number theory...[a sophisticated but apparently unpolished question] is likely to get closed, [and] one encounters a circular problem, since the numerical-PDE folks are less likely to use MO as a forum. This would be a shame in a philosophical sense, but not a disaster. I personally have no energy to address this."

As far as I'm concerned, "research-level" means after the first year of grad school (i.e., when one usually begins to do research). I think that questions at or above that level in any discipline of mathematics belong on MO as much or more than they belong on MSE or scicomp.SE. And I think since our collective quick trigger finger w/r/t closing many such questions has and will continue to hurt MO and by extension the mathematical community. I think we can fix this perception by actively reaching out. If we fix this subject bias then MO can evolve into a fixture of mathematical culture over the long term; otherwise it will degenerate.

quid comments on "Structural bias on MathOverflow" (20785)

2012-12-12T02:06:51-08:00

@Todd Trimble: Mass spring model for hair is another example; see in particular also the meta of that question mentioned in a comment there for general thoughts on the matter.

Personally, I would be even more interested to see example from geraldedgar, since here what was thought of is quite a bit less clear for me than for Steve Huntsman.

Alexander Chervov comments on "Structural bias on MathOverflow" (20784)

2012-12-11T23:05:39-08:00

@Todd and Steve, I would not agree with Steve, I do not see big attitude problem with applied math question, just I think not so many users are interested in it. (Although there are some examples of attitude which Steve is talking about in my practice here e.g. http://mathoverflow.net/questions/79290/how-to-do-mgram-schmidt-orthogonalization-with-integers-real-life-problem this one was almost closed - but that the only case I can remember - I think it is exception). In general in industry you are sometimes in a hurry and do not have time to polish neither your programs, neither questions...

Todd Trimble comments on "Structural bias on MathOverflow" (20783)

2012-12-11T19:25:14-08:00

Frank Thorne may be on to something when he mentioned "prose". One type of answer that is often greatly celebrated, often very disproportionately so in my opinion, is the short, pithy, conceptual answer. Answers that require a hard-core calculation or technique are rarely upvoted much, AFAICT. So perhaps "hard analysis", which requires great command of technique and subtle estimates, might not capture the public fancy so much. Might that tie in to what you're observing, Steve?

Todd Trimble comments on "Structural bias on MathOverflow" (20782)

2012-12-11T19:16:54-08:00

And yet many applied questions that require great sophistication to address but don't appear to have the same level of polish are considered to be unworthy of MO.

I'm genuinely interested in seeing in any examples you have of this, so that I might better recognize what you've been observing.

Steve Huntsman comments on "Structural bias on MathOverflow" (20781)

2012-12-11T18:30:26-08:00

@Todd: my point is that MO is wildly uneven in its approach to different subfields of mathematics because of the wildly uneven background of its membership. A question like "what is the structure describing the zeros of a set of polynomials" would be instantly closed before you could even say "variety". And yet many applied questions that require great sophistication to address but don't appear to have the same level of polish are considered to be unworthy of MO.

The net effect is that people (like me) are loath to invest the effort into asking or answering questions on a site that often doesn't do what it purports or nominally aspires to. I just looked at my own involvement over time, and I can literally see the horizontal asymptote. If MO is for research-level questions in mathematics, then it should welcome research-level questions in mathematics--all of mathematics--while avoiding the complement of that set.

Frank Thorne comments on "Structural bias on MathOverflow" (20780)

2012-12-11T18:14:57-08:00

Unfortunately it seems to me that analytic number theory is somewhat underrepresented. But it seems that there is a natural explanation: answers typically require some elaborate formulas, which are at least slightly painful to TeX up. In contrast, algebraic geometry questions (say) can be answered mostly in prose.

Presumably the same is true of, e.g, PDE.

Not sure this is the explanation, but it's a theory, anyway.

Todd Trimble comments on "Structural bias on MathOverflow" (20779)

2012-12-11T18:07:48-08:00

I really appreciate Steve Huntsman's bringing this bias to our attention, but (Steve) I'm a little confused about question 33230. It seems users like David Speyer and Victor Protsak (who presumably represent some of the more dominant MO fields) were in support of keeping this question open, while you said you entered a vote to close because it had a trivial answer. This runs counter to the notion that the MO mainline is dismissive toward (say) stochastic mathematics. Did I misunderstand your point?

I would be interested to see cases where you feel the MO mainline wrongly shut down questions.

quid comments on "Structural bias on MathOverflow" (20778)

2012-12-11T17:22:45-08:00

@Steve Huntsman: first, to avoid a possible misconception, I certainly had nothing against MO having more activity in these areas. Second, still, this site exists and might contribute to the fact that there is little activity on MO in these areas; so it might be part of an explanantion for the bias.

Third, it is in a true sense unclear to me if a specific "recrutement activity" in this area has any merit (globally) in view of the existence of that site, which is (at 1k questions over a year or so) not large.

Finally, what could IMO be interesting is to start some dialog with people, specifically the mathematicians working in academia, there; I don't know on whether they know/how they see MO, whether they would prefer to be on MO but feel unwelcome, or by contrast are glad not to be bother by all these pure things flaoting around on MO.

Added: regarding your final paragraph, which I either missed or you added later: what is 'belong on MO' supposed to mean? Okay, they might be on-topic here as well. Somebody asked them elsewhere. So, what? A quite agressive discussion on meta.math.SE developped following the presentation of the idea that after migration of MO a certain type of questions should be moved from math.SE to MO.

Added 2: But let me add that I agree, and even said on accassion I think, that MO is also in my opinion somewhat dismissive towards certain subjects, and it should not. It will however be hard to change this. Which does not mean on should not try.

Steve Huntsman comments on "Structural bias on MathOverflow" (20777)

2012-12-11T17:00:55-08:00

I would bet $.50 and a Sprite that scicomp has more activity than MO because MO is dismissive of the topics it covers.

That is not a reason for MO to continue to be dismissive of it.

If it's "all AMS and no SIAM" then we're all losing something. I recall a comment I made on http://mathoverflow.net/questions/33230/matrices-whose-exponential-is-stochastic : "The grumpy old man in me can't resist: an order of magnitude more people surely learn about Markov processes than coherent sheaves or motivic homotopy theory or whatever "most" mathematicians supposedly study at any level." IMHO this was an example where the MO community was ill-placed to recognize what qualified as research-level mathematics because of its biases--and not in the direction that indicated an overabundance of awareness on the topic.

We see questions get closed and/or shunted to MSE all the time. But I see MSE (and if I went more than once every six months would probably see scicomp) questions that belong on MO a lot.

quid comments on "Structural bias on MathOverflow" (20776)

2012-12-11T16:40:34-08:00

While Nilima Nigam seems (based on a quick search) not active there she said Jan 14th, 2012, in response to David Ketcheson advertesing here on meta the site scicomp.SE, mentioned above:

I agree with David. This SE seems actually a very good site for numerical analysis/scientific computing. Within a few months of its inception it has more substantive questions, and more answers, in these areas than in the equivalent time span on MO.

Will Jagy comments on "Structural bias on MathOverflow" (20775)

2012-12-11T15:48:34-08:00

Francois, I don't know, I sent you her email. I seems one of your addresses worked, from a blog, the older bounced.

François G. Dorais comments on "Structural bias on MathOverflow" (20774)

2012-12-11T15:35:52-08:00

I'm curious what those other places are?

Will Jagy comments on "Structural bias on MathOverflow" (20773)

2012-12-11T15:31:03-08:00

Francois, yes, I wound up writing to Nilima Nigam a few times. I'm going to call what she does numerical PDE because I do not remember ny better than that. She is not someone who needed MO for answers. She told me that if she did have quick questions, the were better places to ask. And she did not like the quick closure and generally dismissive attitude towards numerical questions. So that is mostly on the supply side. On the demand side, well, Nilima left, J Sethian never got interested in MO (he is my age), so maybe this is not a productive place to ask.

Bootstrapping is about the right term.

François G. Dorais comments on "Structural bias on MathOverflow" (20772)

2012-12-11T15:18:51-08:00

I'm a bit puzzled by one aspect of this lack of representation of certain fields. One saying I hear a lot these days (out of context) is: You can't ask for what you don't know. Presumably, if some people take the time to say that a certain field X is underrepresented that means that we already do have qualified and interested people in field X. If that is correct, then it's some kind of bootstrapping issue that is going on. Does anyone have thoughts about this?

Will Jagy comments on "Structural bias on MathOverflow" (20771)

2012-12-11T15:01:33-08:00

Well, just to repeat some ideas of mine that were never popular...Computer science departments have not done well in attracting women. Evidently there were a couple of decades of growth, then female enrolments dropped off again. Something about industry jobs remaining a boys club. Mathematics departments are making some headway, partly through government programs. Similarly, topics that are underrepresented on MO are often doing fairly well in the academic setting.

So, my cluster of ideas, and i know i may be the only one, so please be nice, amounts to less isolationism of MO. In the first couple of years, i would regularly email people about MO questions, sometimes to link two other parties who did not know how to contact each other. Very few people ever replied to me. The part about two other people was this: once you find the two or three people who actually care about a topic, surely the discussion can go much faster and deeper off site, with occasional summaries posted on MO? There seemed to be strong objections based on transparency, which I really do not get. Are your own notes in preparing a manuscript public?

Meanwhile, underrepresented topics. If there is some perception that posting on MO will do you some kind of good, perhaps just a contact, that will somewhat weigh against the perception that there is currently no interest on MO in topic ___.

The reason I thought to do this is that there is a question on MSE, some poor slob whose wife has been diagnosed with lupus, he wants to know the probability that she actually does have lupus. He gave nowhere near enough information. The arguments for keeping the question open (and now for re-opening) amount to a jingoistic MSE for MSE attitude that makes little sense to me. It is part of the larger world, and you should not give an incompetent half-baked probability calculations when he should be talking,in person, to other kinds of experts.

And i have often felt that MO has some of that MO for MO feel. But as long as it remains a closed system, inequities are likely to continue. A person in field X says look, nobody on MO has posted in my field since___. The one expert in filed X on MO eventually leaves, there are no interesting questions.

My own thing is communication expanding offsite. Sometimes i think it is just a case of person Y saying, what are you talking about, i telephone A,B once a week, I see C,D,E in conferences at least twice a year, communication is fine. But then, there are people F,G,H who are just outside that loop. And, of course, I've had my share of disappointments along these lines.

I forgot this part: for MO the larger world could be described as traditional mathematics publishing. There was a feel when i started here that this was a grand alternative to traditions, so I automatically started calling it MathOverthrow. The thing is, an awful lot of government money and people's energy have gone into, well, diversity, in mathematics, both subject matter and personnel. MO lacks such resources and always will. Larger ties of some sort to the traditional structures of mathematics life would be good.

quid comments on "Structural bias on MathOverflow" (20769)

2012-12-11T07:37:14-08:00

For the discussion of fields, in particular if we come to certain parts at the applied/computational end it is perhaps worth keeping in mind the existence of http://scicomp.stackexchange.com/ , several of the top-rep users being mathematicians; and for (applied) statistics there is http://stats.stackexchange.com/ though this might already be away a bit. To me it is genuinely unclear if there is any merit in trying to attract contributions that seem a very good fit for the former to MO, even if they would also fit here. Even more so as likely the sites will soon be part of the same network.

Angelo comments on "Structural bias on MathOverflow" (20768)

2012-12-11T07:28:08-08:00

To François: Oops, sorry, I had misinterpreted what you where saying.

Steve Huntsman comments on "Structural bias on MathOverflow" (20767)

2012-12-11T06:21:31-08:00

My original idea was to actively solicit luminaries from under-represented disciplines to post appropriate questions, and publicize these questions elsewhere. Because AG etc. are already so well represented, including all disciplines seems unlikely to reinforce the bias. Something like having featured questions referenced (perhaps even having outstanding answers briefly discussed) in a community journal like NAMS might be worth considering.

We might also ask Yla Tausczik if she has any ideas/suggestions/comments on the thread topic.

François G. Dorais comments on "Structural bias on MathOverflow" (20766)

2012-12-11T06:18:15-08:00

Angelo, those were MO question counts. The number of papers in each category can be found here.

geraldedgar comments on "Structural bias on MathOverflow" (20765)

2012-12-11T05:40:24-08:00

Steve said: MO has a pronounced bias in favor of [certain branches of mathematics] and against [other branches]

When I read the title of this thread, I thought that is what it would be about. I have the same impression as Steve. But what can be cone about it? I don't know. (Asking our eager closers to make small changes in their habits has not worked.)

Angelo comments on "Structural bias on MathOverflow" (20763)

2012-12-10T22:05:13-08:00

Going back to the bias against analysts and applied mathematicians, the number of papers in arxiv does not give a fair estimate of the percentage of mathematicians who are active in each discipline. Obviously, no one can believe that there are 10 times more algebraic geometers than mathematical physicists, or PDE people. To get an idea, MathSciNet lists 37,030 papers whose primary classification is algebraic geometry, 58,527 in number theory, 133,995 in PDE, 72,942 in fluid mechanics, 106,796 in operations research and mathematical programming. Of course, this does not give an completely fair picture either, because algebraic geometers tend to publish less than analysts, mathematical physicists or applied mathematicians; still, it seems clear that the distribution of questions in MO does not reflect the demographics of the mathematical population.

What may be the reason, or what could be done, I don't know. It may be that the fact that there are more people active in a certain area tends to attract people in the same area, because it is easier to get good answers, and because people hear more about MO.

Is the bias related with the fact that algebraic geometers use the arxiv far more than people in most other areas? Is there a cultural difference at work? I have no idea.

Todd Trimble comments on "Structural bias on MathOverflow" (20761)

2012-12-10T17:30:28-08:00

Thanks very much, quid. I had guessed the anon/pseudonymous numbered higher than the rough estimates you report; that's interesting.

quid comments on "Structural bias on MathOverflow" (20760)

2012-12-10T17:10:03-08:00

@Todd Trimble: Technically, it certainly depends what counts as 'user' and in addition what counts as anon/pseudonymous [people just whose display name is not the full real name are never counted in here, if they 'identify' further on the userpage via a link to a webpage for example]. What I meant, and thought to write but did not quite arrive at making precise, but thought to be understood were in some sense 'regular' users.

In any case, at the top-end (100) the anon/pseudonymous users are about 5 percent, then they increase (slowly) but if you flip through the pages they stay a few quite long. At around Top 300 (this is about the voting threshold it seems like 15percent, roughly). If you want to get a rough idead further down I'd suggest pick in the userlist orderer by rep around page 30, that is top1000 or 700rep. They seem still fewer there.

Todd Trimble comments on "Structural bias on MathOverflow" (20759)

2012-12-10T16:49:01-08:00

Andres, I don't know. If I absolutely had to get some idea and do it myself, I'd go through the long list of users where in many cases (such as 'quid') it would be obvious that it's a pseudonym or anonym, in many cases (e.g., Andreas Blass) obvious that it's a real name. Names like 'Agol' or 'gowers' not exactly pseudonyms and would not be counted as such. Also cases like HW and VA are not too hard to decipher. I think that most cases would be either easy to associate with a real person or pseudonyms which are hard or impossible, although proceeding in this manner would obviously be a tedious and time-consuming task.

In case you're saying that it's a stupid question or unanswerable without putting in a lot of work, you might be right, but mainly I was after a rough estimate and wondered if anyone had any idea.

Andres Caicedo comments on "Structural bias on MathOverflow" (20757)

2012-12-10T16:13:11-08:00

How would you determine such a thing, Todd?

Todd Trimble comments on "Structural bias on MathOverflow" (20756)

2012-12-10T16:08:07-08:00

Actually, what is the percentage of anonymous/pseudonymous users amongst all users?

quid comments on "Structural bias on MathOverflow" (20755)

2012-12-10T15:26:44-08:00

Since it is about any kind of bias:

I think there is a bias against anon/pseodonymous users. This is not a personal complaint, and in some sense by-design as there is a suggestion in the FAQs. Still, I wanted to bring it up, since to me it falls well into this microcosmos or toy-model of the real world idea, and I though about this before, but I really do not want the following to be overinterpreted. (If somebody sees a problem of whatever kind with this I remove it, it is not my intention to be provocative here.)

It is in some sense an interesting experience for me (as somebody who falls in real-life along most, perhaps all, typical criteria in, how to say this neutrally 'the local standard group' perhaps, to the extent this makes sense) to be a member of some group seen with certain reservations by 'the local standard group'.

And, I believe to notice certain analogs of patterns of behaviour towards this group that are AFAIK somewhat typical in such a situation. For example, that memebership in this group is brought up in situation of conflict somewhat arbitrarily (that is, as some kind of ad hominem as opposed to there being any direct connection of membership and conflict-at-hand).

François G. Dorais comments on "Structural bias on MathOverflow" (20753)

2012-12-10T12:55:28-08:00

Any kind of bias... We shouldn't be biased against certain biases!

Angelo comments on "Structural bias on MathOverflow" (20752)

2012-12-10T12:43:10-08:00

It's not clear to me what the topic of the thread is supposed to be. Is it bias against women, or any kind of bias?

François G. Dorais comments on "Structural bias on MathOverflow" (20751)

2012-12-10T12:40:05-08:00

(Yes, there are a few other problematic categories too. I hacked a clustering program last year that I might dig up to further analyze this if I have time.)

Andres Caicedo comments on "Structural bias on MathOverflow" (20750)

2012-12-10T12:27:09-08:00

(There are quite a few questions in set-theory not cross-listed with lo.logic)

François G. Dorais comments on "Structural bias on MathOverflow" (20749)

2012-12-10T12:02:35-08:00

Thanks for reminding us of that, Steve. In case we look back here in the far future, here is a snapshot of the current counts for each arxiv category:

ag.algebraic-geometry 5443
nt.number-theory 3408
co.combinatorics 2043
at.algebraic-topology 2026
gr.group-theory 2000
dg.differential-geometry 1806
pr.probability 1777
ct.category-theory 1522
rt.representation-theory 1509
fa.functional-analysis 1392
ac.commutative-algebra 1373
lo.logic 1194
ca.analysis-and-odes 1069
gt.geometric-topology 928
gn.general-topology 907
ra.rings-and-algebras 771
st.statistics 657
mg.metric-geometry 547
mp.mathematical-physics 533
ap.analysis-of-pdes 531
ds.dynamical-systems 498
cv.complex-variables 487
oa.operator-algebras 416
ho.history-overview 413
na.numerical-analysis 409
oc.optimization-control 359
sg.symplectic-geometry 324
qa.quantum-algebra 293
kt.k-theory-homology 214
it.information-theory 173

Steve Huntsman comments on "Structural bias on MathOverflow" (20748)

2012-12-10T11:31:40-08:00

I would like to remind folks of my long-held position that MO has a pronounced bias in favor of algebraic geometers, number theorists, etc. and against applied mathematicians, analysts, etc:

http://mathoverflow.net/questions/6292/why-is-algebraic-geometry-so-over-represented-on-this-site-closed

The fact that this bias has persisted for so long suggests to me that it is structural, and not merely due to "founder effect". My personal impression is that applied questions are far more likely to be closed even when accounting for the level and quality of the question.

gilkalai comments on "Structural bias on MathOverflow" (20746)

2012-12-10T09:19:57-08:00

Well, I did not intend to start a discussion about women at MO, and certainly not about action to "improve the situation." MO and other Internet activities are quite important in my opinion, but they are fairly small in the overall mathematical scene, and they are not (with very few exeptions) where the cutting-edge part of math research is happenning (while sometimes being reported). It is still not clear if spending much effort on MO or similar activities is a good idea, and probably it is good for some and bad for others. I tend to think that this is (within limits) a good idea for me, but I may be wrong even on this very specific and personal issue.

I think that the MO example is potentially interesting, because the ratio of women at MO is much smaller than the ratio of women in any relevant population that we can use for comparison: in the population of math professors ,and in the population of math graduate students, or in the population of prominent living mathematicians, etc. Understanding the MO situation may give us some hints about issues which are quite orthogonal to mathematics that discourage women (and others), and understanding it may be more important in contexts other than MO. But I dont have much to say beyond it, I did not study carefully this issue; I remember following with interest (and sometimes with surprise) the earlier thread on the matter and Laba's post on it.

Jon Bannon comments on "Structural bias on MathOverflow" (20745)

2012-12-10T09:07:03-08:00

I think it is a very good idea to hide user reputation. Looking at a (I think it was Ben Webster's) comment following I. Laba's blog post, one is led to believe that the reputation system is mainly for community moderation purposes. I don't think this is true, primarily because rep. is very publicly visible. Perhaps there is some worry that the market will lose liquidity if players who only dabble in order to increase their rep. are not involved, but the value of drawing more women to participate in the discussion ought to offset any such effect. Is there some other major reason not to make rep. less visible (for that matter, the total linear ordering on users)?

Or perhaps it is simply true that we are trying to foster competition here and if so it is tasteless to pretend otherwise.

Keep in mind that if rep. were less visible, asking a bunch of horrible questions and giving a bunch of horrible answers would still yield rep. changes, but the net rep. changes wouldn't be the first thing you see after someone's name.

deane.yang comments on "Structural bias on MathOverflow" (20744)

2012-12-10T08:11:35-08:00

First, despite the appearance of my first name, I'm male. I'm definitely interested in learning more about this, and finding ways to improve the situation. Does anyone know if the other stackexchange sites have suffered from the same issue or not? And whether they have tried to address it or not?

rita comments on "Structural bias on MathOverflow" (20743)

2012-12-10T02:37:35-08:00

@Francois:
I don't think I was especially targeted, either.
I meant that women (at least I) may tend to be more easily affected by this kind of things, and so more easily discouraged from using the site.

François G. Dorais comments on "Structural bias on MathOverflow" (20740)

2012-12-10T02:11:24-08:00

Rita, we were typing at the same time, that wasn't a response to your post. I was just trying to clarify the topic since "bias" can apparently be used in ways I wasn't aware of. In any case, I'm glad your MO experience has been (mostly) positive. I think we all get spurious downvotes, which could be due to inane reasons like some users poorly aiming at the up arrow... (I just checked and it doesn't look like you're the victim of a personal vendetta or anything like that.)

rita comments on "Structural bias on MathOverflow" (20739)

2012-12-10T01:48:18-08:00

@francois: you are right, I missed the point.

My impression is that to change the situation what one need is that more women decide become visible on MO, so that in turn other women don't feel "strange " there. I do not think the reputation score is the real issue: everybody starts at 1 and there are lots of users (probably the younger ones) who seem quite happy to participate even if they do not have a high score.
On the other hand, sometimes the feedback from other users has sounded to me sort of aggressive and I was set back by it. And I got some downvotes to perfectly reasonable answers that I could not account for.

Unfortunately I have no idea what to suggest: the occasional comments that I considered aggressive definitely did not deserve flagging as offensive and it is important to be also able to downvote.

My original question was motivated by the fact that I had the impression that some bias against women existed on MO (as well as anywhere else), but I had no evidence and I wanted to learn from Gil what evidence he had.

François G. Dorais comments on "Structural bias on MathOverflow" (20738)

2012-12-10T01:16:25-08:00

I guess it wasn't clear from the original question that we want feedback regarding the site itself. Here is an example from Izabella Laba's blog post http://ilaba.wordpress.com/2011/03/28/why-im-not-on-mathoverflow/

To me, the “reputation” system looks like a formalized version of the same informal evaluation systems that social groups have been using forever. If a woman has noticed in the past that she is being taken less seriously than her male colleagues, she’ll expect the same on MO, for example that she’ll get fewer points than a male colleague for the same knowledgeable answer. The point system encourages a competitive mindset, and I think that mathematicians tend to be competitive by nature. If you’re female and expect to start with a huge disadvantage just for this reason, you may well be discouraged from participation.

This of course is based on general life experience, not on anything that has actually happened on MO. I’d love to see MO develop a reputation (heh) for treating women fairly.

As a partial remedy for this, we added two preference options:

Show symbols !, ⋅, ∘, ⊡, ⊟, ⊞ for reputations above 0, 15, 100, 2000, 3000, 10000, respectively.
Do not show user reputations.

Is there anything else we could do to make the reputation system more palatable?

rita comments on "Structural bias on MathOverflow" (20737)

2012-12-10T00:58:25-08:00

I've just had a look at the previous MO discussion on women participation. When I asked Gil to say more on the topic I did not know about that discussion and if I had I would have refrained from asking, for fear of reopening a similar one. However, now I feel responsible for the opening of the thread and I'm going to share some of my impressions.

Personally, I've never had the impression of being treated differently on MO because my username (=my given name!) clearly says I'm a woman. Or rather, I have sometimes suspected that the fact that I am a woman makes a difference to the attitude of some people on MO, but exactly in the same way as it happens in my offline mathematical life. I can live with that, actually I hardly notice it.

On the other hand, I could not help noticing that there seem to be hardly any women around, at least in algebraic geometry. Since I enjoy MO very much, I've been wondering why.
My guess is that there a few women that participate under a gender-neutral pseudonym, and that the reason for this is not (or not only) the fear of being harassed but more likely the fear of being judged negatively. I sometimes have that fear, too, although I try to get over it. I think it is related to gender in the sense that I mostly see it in women, but I don't know where it comes from, whether it's innate, or induced by education and society, or a mixture of these.

I will close with an anecdote. It's language-dependent, but I'll try to translate. A couple of weeks ago, I was in my office, sitting at the only (!) desk and somebody knocked on the door. It was an insurance salesman, as I learned later. Here's the conversation we had: "May I come in?" "Yes, good afternoon. What can I do for you?" "(In a very self-assured tone) I need to talk to male-professor-My_last_name." "Actually, it's me..."

At that point the poor guy would have liked to vanish in the air. The reason for the mistake was, of course, that the label on my door says "R. My_family_name". He had probably already knocked on all doors in the corridor and had seen only men. However I don't think he'll make the same mistake again if he wants to make a living by selling.

And, to be honest, I sometimes exploit stereotypes: when I want to get rid of salesman and the like, I often go: "uh, I don't know... I think I'd better discuss this with my husband first". At that point they let go. Always.

On the whole I feel quite lucky. I wished for a career in math and I've had it. If I had been born 50, or even 30 years earlier, I would not be sitting behind a desk in math department. And I'm sure the next generation will fare even better.

François G. Dorais comments on "Structural bias on MathOverflow" (20736)

2012-12-10T00:58:17-08:00

Bias is very specific and was chosen deliberately. There is little hope that we can change the community and individuals. We can, however, change the site (well, pretty soon we will) to remove some structural bias. That said, I have yet to hear anything we could do.

Edit: I added "structural" to the title to make the topic clearer.

Neil Strickland comments on "Structural bias on MathOverflow" (20735)

2012-12-10T00:47:06-08:00

While I don't agree with everything that Fedja wrote, I do agree that the title of this thread is very unhelpful. I suggest that it should be changed to 'gender imbalance on Mathoverflow'.

fedja comments on "Structural bias on MathOverflow" (20734)

2012-12-09T23:43:22-08:00

Well, Francois, if we start with using the word "bias" in the title of this thread, which, according to Wikipedia, means

an inclination of temperament or outlook to present or hold a partial perspective at the expense of (possibly equally valid) alternatives in reference to objects, people, or groups. Anything biased generally is one-sided and therefore lacks a neutral point of view. Bias can come in many forms and is often considered to be synonymous with prejudice or bigotry.

I don't expect anything good as the result. The ones who view the lack of women in academia or on MO as a real problem will resort to the usual strategy of labeling their opponents as "male chauvinist pigs", "hidden sexists", etc., and the ones who don't (like myself) will dig their heels in and say things they don't even really mean just to counterbalance all that liberal nonsense. ;) I guess that unless we all agree that academia is neither a bath house, as Hilbert once said, nor a zoo where one has to have a pair of each creature and discuss the issue in a straight way trying to figure out what exactly repels women from MO and academia, if anything (not from some strange gaming sites, dark streets in Chicago, etc.) and whether each particular problem is best solved by changing the routine behavior of the male population or changing the routine ways of thinking of the female one, this thread is also doomed to get closed soon. Note that I said "what repels" and not "how to attract". I firmly believe that you should try (within reasonable limits) not to repel anyone worth having (and the same Izabella, say, would be definite worth having here) but I personally have no desire to go out of my way to "attract" anybody to anything and do not expect anyone here or elsewhere to have any desire to go out of their ways to attract me. The choice whether to join or not to join MO (or any other forum or group) is free and it has to be made not based upon whether there is no single minus in that (this way we'll never satisfy anybody, not to say everybody) but on whether the pluses outweigh the minuses. With all this said, I'd like to hear what Gil wants to say on the issue :).

François G. Dorais comments on "Structural bias on MathOverflow" (20731)

2012-12-09T21:17:19-08:00

Gil brought up bias against women in another thread. We had a lengthy discussion about this some time ago (which eventually degenerated). At a recent meeting where I was asked to talk briefly about the main issues that MathOverflow has had, I mentioned "demographic bias" as one of them. Although we recognize the importance of this issue, we have not found much that we (the moderators) could do about this. Of course, we welcome feedback on this important issue.

Warning: This is a sensitive topic! Please stay on topic and be respectful...