Not signed in (Sign In)

Not signed in

Want to take part in these discussions? Sign in if you have an account, or apply for one below

  • Sign in using OpenID

Discussion Tag Cloud

Vanilla 1.1.10 is a product of Lussumo. More Information: Documentation, Community Support.

Welcome to Math2.0
If you want to take part in these discussions either sign in now (if you have an account), apply for one now (if you don't).
    • CommentRowNumber1.
    • CommentAuthorKevin Walker
    • CommentTimeFeb 22nd 2012
    • (edited Feb 22nd 2012)

    (This is intended to be a more narrowly focused offshoot of this thread.)

    Henry Cohn suggested (here and here, and perhaps elsewhere too) that it should be possible to have discussion threads on the proposed site which do not correspond to a single arXiv paper. For example, threads which discuss several related arXiv papers, or threads which discuss non-arXiv papers. Henry asked if there were any downsides to this more flexible organization scheme, and I didn’t notice any answers to his question.

    This (not restricting discussion topics) seems like a good idea to me. As Henry points out, it should be possible to keep track of which threads mention a given paper, so even if one’s only interest in the site is to accumulate commentary on specific papers, this proposal offers advantages.

    • CommentRowNumber2.
    • CommentAuthorJohn Iskra
    • CommentTimeFeb 22nd 2012
    I don't think that human conversation should be shoe-horned in this way. I don't think that it can be anyway. A certain amount of verbal and written 'wandering' is part of a prodcutive creative process. Still, I agree we do want some structure.
    One approach to this might be to utilize software such as discussed and show here: http://blog.contextdiscovery.com/websummarizer/map-wikipedia-summary-by-wikisummarizer/
    This would, to a certain extent, address the underlying problem which such rules are designed to remedy: How to organize information in a way which makes understanding the relationships among the data (opinions..) easier to see.
    • CommentRowNumber3.
    • CommentAuthorOlivier GERARD
    • CommentTimeFeb 22nd 2012

    @Kevin : it seems to me that even when focusing on a specific arxiv paper, there will be cases when the discussion will also involve its references (choice, content, interpretation, lack of), a part of them possibly on arxiv as well. So while it seems interesting to be able to find/search/tag/categorize discussions thread as refering mainly on a given article, there should be a natural (and automatic way) to find the discussion by mentioning one of its references or one of the articles that will quote it in the future.

    • CommentRowNumber4.
    • CommentAuthorKevin Walker
    • CommentTimeFeb 22nd 2012
    • (edited Feb 22nd 2012)

    Just to clarify, when I wrote “This seems like a good idea to me…”, the “This” refers to Henry’s idea of not restricting discussions to single arXiv papers. I guess I should edit my original comment.

    • CommentRowNumber5.
    • CommentAuthorMarc Harper
    • CommentTimeFeb 23rd 2012
    I don't see why discussion would have to be limited to a single paper. Each paper can have multiple discussions/threads, and one could be how another paper relates to this one. Surely this discussion could be linked to both papers, if such a linking is actually useful.
    • CommentRowNumber6.
    • CommentAuthorHenry Cohn
    • CommentTimeFeb 23rd 2012
    I guess the issue as I see it is that the underlying architecture should reflect how the system could or should be used. Right now, it sounds like discussions are subsidiary structures attached to a particular paper, perhaps with the possibility of linking them to multiple papers if necessary. Instead, I think discussions should be first-class objects in their own right, with no presumption that they have any special relevance to a single paper in the first place.

    For example, many comments will be inspired by a given paper but really not about it. Someone may ask whether different models for the phenomena in the paper have been studied, or what the precursors of a certain trick are. These discussions should certainly be linked with the paper that inspired them, but they will quickly move on from there.

    Maybe I am misinterpreting what people have been saying, but my impression is that under the current plan, a user who is looking at arXiv:xxxx.yyyy will be presented with (among other things) a list of discussions about that paper and the option of starting a new discussion to be filed there. Having started such a discussion, the user may have the option of cross-listing it under another paper as well.

    Instead, I would present it as a list of all discussions that mention the paper (with another interface elsewhere to get to all discussions). Instead of starting a new discussion about a paper, you would just start a new discussion, and it would automatically be linked to every paper that is mentioned in it (the same sort of way arXiv trackbacks work now). If no paper was explicitly mentioned, then the system would guess that the topic is the paper you were last looking at and would ask you to confirm, but it should accept starting with no paper at all. (For example, a reference request where you don't know yet what paper it should be linked to.) This seems better aligned with how people are likely to use it, without sacrificing anything or being much harder to implement.
  1. How do we come in to the site ?
    My understanding was the following - from arXiv's paper we have a link - and we come in
    precisely to the page which discussed this precise paper.
    So if discussions which "claimed" to be about paper will be of actually quite far from the paper's content it will be disappointing.

    My first negative reaction on Henry Cohn's proposal was based on "coming in" problem.
    If we have discussions which are NOT paper related how will somebody find it ?
    However, thinking a little more, I do not think it is problem.
    For example a) one starts discussion and just sends to his friends e-mails "let us discuss..."
    b) we might have "recomendation system" like on MO, which for each discussions hints
    other discussions which are related to it.
    c) may be some one can find some discussion by google search - however if site is not very popular small chance for this method of come in,
    c) finally we might have list "NOT-paper related discussion" - but this I do not think pretty useful,
    since if you have 100 such discussions , I am too lazy to look over all of them to find something intersting for me.

    PS
    @Kevin Walker I also thought to create special thread to discuss this subject :)
    PSPS
    How should front-page look like ? (may be special thread for this ?)
    • CommentRowNumber8.
    • CommentAuthorAndrew Stacey
    • CommentTimeFeb 23rd 2012

    How do we come in to the site ?

    This is the part that’s bothering me. Why would I visit this site? What’s the draw? I can believe that once I’ve used it a bit and found it useful then I’ll keep coming back, but that still seems to require quite an activation energy, and more so in the early days when there isn’t much content there.

    Broadening it out to general discussion on mathematics sounds good, except that it will be even harder to moderate and even harder to navigate. Moreover, for myself then I don’t need this: I already have the nForum and if I want to discuss something mathematical I can do it there.

    Indeed, I think that there’s a lesson to be learnt from the nForum. Its original purpose was not for general mathematical discussions. Its original purpose was as a place to record stuff about the nLab for others to easily find where someone had made changes. This naturally sparks discussion, but even when not much discussion is happening then people still have a reason to check the site: to see what changes have been made on the nLab.

    I’m not sure that a general discussion site, or even a per-arxiv paper discussion site, will have enough useful stuff to keep me checking it each day on the off-chance that there’s something of interest. I need something more from this site.

    What I think really is missing from the arXiv, and which could be integrated into this idea, is the ability to easily add meta-data. Making it easy for someone to bookmarks papers is part of it (which is part of the original proposal, but I think that the bookmarking part should be emphasised more and the fact that this data might be put to use should be de-emphasised), but also making it possible for users to tag papers with subject areas - let’s face it, neither the arXiv’s classification nor the AMS one is particularly useful for finding papers - and to link papers together in some fashion. In short, provide a portal to the arXiv that means that I would prefer to check the arXiv via this site instead of via the main arXiv site because of all the extras that the site lets me do. Then I’ll check it every day. Then I might happen to spot an interesting discussion, join in, and - hey presto! - the activation energy barrier comes crashing down.

    This is the sort of thing I was trying to get at in my Useful at the point of use post, and which I think Ben is trying to say in his post on selfishness. Give me a reason to check it every day, and you can sneak in a load of extra functionality without me noticing. Make it all about “benefiting the community” and I’ll cheer from the sidelines, but when it comes down to it the 101 other things I have to do each day will crowd it out.

    • CommentRowNumber9.
    • CommentAuthorMarc Harper
    • CommentTimeFeb 23rd 2012
    • (edited Feb 23rd 2012)
    @Henry What if discussions could be started in a particular area, say on the "homepage" for algebraic topology? On the page coresponding to papers mentioned in the thread there would be a link back to the thread. So a thread for a paper could be started directly for the paper (from it's page) or later be associated to the paper if mentioned at some point.

    @Andrew Thanks for your comments.I completely agree that the site needs to be useful at the point of use. We do intend for the site to organize around communities. Up until now I had thought that predefined fields would be better initially since an unrestricted tagging system has a host of problems (such as singular versus plural usage, misspellings, etc.) Do you think something like ArXiv classfication + AMS classification should be supplemented by a tagging system? It seems like the tags themselves would have to be aggregated somehow to be useful.

    The other draw is the recommendation system. Ideally you'd have a stream of papers that are either recommended to you specifically, recommeded by someone you follow, or recommended based on what the system thinks you are interested in. This should make it easier to stay up on the latest works without having to poll journal sites or rely on your peers to communicate interest via other mechanisms.
    • CommentRowNumber10.
    • CommentAuthorAndrew Stacey
    • CommentTimeFeb 24th 2012

    We do intend for the site to organize around communities.

    This is the problem. The communities won’t exist until there is a critical mass of users, but you won’t get a critical mass of users until there are existing communities demonstrating that it works.

    It seems like the tags themselves would have to be aggregated somehow to be useful.

    No. Tags are useful when a user can put their own tags on a paper. If you build a site where:

    1. I can check what’s new on the arxiv today, and easily scan back and forth through recent days - maybe a bookmark to remember the most recent day that I’ve checked.
    2. Save papers for later perusal.
    3. Tag papers with my own set of tags (autocompleting on my tags, not anyone else’s)
    4. Do searches and other useful stuff (via the arXiv API)
    5. Organise my saved papers by various criteria
    6. Import non-arXiv papers, say from MathSciNet, Google Scholar, or manually
    7. Export to bibtex (or biblatex, or other setup)
    8. Easily add notes to a paper

    Then I might use it. Moreover, I might decide to make a subset of my tags “public” for others to see, same with my comments. (However, it’ll have to be good because I already have a system that does pretty much all of that. But my system is a bit rough around the edges and I have no incentive to smooth those out to build it into a bigger project so I think that there is a place for a polished system doing all the above and if it’s good, I’ll swap from mine to it. But my point is that I do use that every day.)

    Of course, the public tags might need aggregating, or at least maybe there could be a family of public tags and I could decide that my private tag “comparative smootheology” corresponded to the public tag “generalised smooth spaces” (or even “generalized smooth spaces”) so the system would make it possible for me to set up an alias. Hierarchies would also be useful. I could designate “comparative smootheology” a sub-tag of “differential topology” and again, by making it public anyone who searched for “differential topology” would see all the papers that I’d tagged as “comparative smootheology”.

    Now, I’m the best person to decide the aliases here because I’m the only person who truly understands what my tags mean. So your job as the implementer would be to make it as easy as possible for me to make those aliases.

    The idea that I’m trying to push here is that the site is useful to me by myself. I gain from using it even if no-one else does. Then there’s some extra stuff involving interactions and these are really easy to do. So easy that the thought process is a bit like “I’m here anyway, I may as well do a bit for the community”. So the decision as to whether or not to help the community happens when I’m on the site, not when I’m deciding to go to the site.

    I think that my “time for the community” is somewhat limited: I also have to do my research and teaching and administration. So if using this site is purely “for the community” there’s always the danger it’ll get squeezed out. But if you can make it directly connect to one of those three (probably research!) then it won’t get squeezed out because it takes higher priority.

    • CommentRowNumber11.
    • CommentAuthorOlivier GERARD
    • CommentTimeFeb 24th 2012

    @Andrew :

    a) I second the requirement for personal tags. This is often overlooked. This is not like a general public site where you want to share/crowdsource the tags, correct/curate them systematically, or as in Stackexchange sites, where creating a new tag requires a sufficient experience and trust level so that the taxonomy stays tame and coherent for the communty.

    I would add : please make sure from the start to support multiword tags and accented characters.

    b) Some of your requirements are at the functional center of sites such as CiteYouLike and Mendeley. Also I am certainly not the only one having a few personal tools to keep track of what I read or intend to read.

    So importing into this site whole bibliographies and comment texts, from a variety of file formats and external websites could also be an incentive for users already hooked to other partial solutions. Better yet synchronization with some bibliographical websites.

    c) defining, saving, editing cumulative, multi-keyword searches on arxiv content to filter potentially interesting papers.

    But is not all this at once a bit too much ? We are a bit far from “useful at the point of use”.

    • CommentRowNumber12.
    • CommentAuthorAndrew Stacey
    • CommentTimeFeb 24th 2012

    But is not all this at once a bit too much ?

    Yes. It should be built up incrementally. By painting a bigger picture, I hoped to jolt the discussion off from commenting as the key feature and on to something else. I would go for personalised tagging (and I agree with your comments on practicalities), and then add other features gradually. The important thing, I think, is to identify one feature which would be useful even if only a handful of people use it. Then implement that and make it really slick. Once people are using it for that key feature, others can be added and each will draw a few more people in. But in the early stages, only a handful are going to seriously use it and it has to make a difference in actuality (as opposed to in potential) to that handful.

    • CommentRowNumber13.
    • CommentAuthorJohn Baez
    • CommentTimeFeb 25th 2012
    • (edited Feb 25th 2012)

    Andrew Stacey wrote:

    Why would I visit this site? What’s the draw?

    The draw is that it’s just like the arXiv, only better - i.e., with more functionality. Of course there are some functions, like uploading papers, that only the arXiv can do. But we can just make a link to the arXiv for those.

    Most importantly: if you want to download a paper, you can go to the arXiv, or go here - but here you can also see what people said about it. For example, maybe someone found a mistake in it, or maybe someone had trouble with a tricky step in an argument and asked about that, and maybe they got an answer. Maybe you have some things you want to say, or ask, about the paper you downloaded. So, going here instead of the arXiv shouldn’t hurt or cost much: it should mainly just help.

    The site may also eventually serve as your personal database of arXiv papers. I know you, Andrew, already have one, because you’ve taken the trouble to write software for this purpose… but most of us haven’t!

    We haven’t talked about this enough yet, but besides the advantages of being able to see what other people have said about arXiv papers, there are advantages to having an easy way to read what you yourself have said about papers - even if it’s as simple as a tick saying you want to remember to read it.

    (EDIT: I see you said similar things in more detail yourself, later in this thread.)

    This sort of functionality can start out simple, by giving everyone the ability to read what every other person has written - and thus, giving everyone the ability to read what they themselves have written. Later it could become searchable in more interesting ways.

    (We could also allow people to keep certain comments private, so only they themselves can read them. This would encourage using the place as a personal database.)

    Anyway, I’m digressing. My main point is that as long as most of the papers on this Selected Papers Network come from the arXiv, and as long as it’s brand new, we should make its format look similar to the arXiv, so it’s just like the arXiv, only better.

    • CommentRowNumber14.
    • CommentAuthorScott Morrison
    • CommentTimeFeb 26th 2012

    I just wanted to point out Laba’s post (a must read for everyone here!), in which she said:

    The main thing is, I like that the person who’s being asked to evaluate my paper is also being asked to actually read it.
    

    I’m really worried that the level of comments will be too low on any such system. Systems to discourage “low-level” participation are of course dangerous; they risk discouraging all participation, but I think they’re also necessary.

    I’d strongly encourage: * enforcing real names * tying identities to arxiv accounts (yes; if you don’t have a paper, you can’t participate) * not allowing retraction of comments (or rather, allowing access to edit histories)

    • CommentRowNumber15.
    • CommentAuthorIzabella Laba
    • CommentTimeFeb 27th 2012
    Thanks, Scott.

    Requiring a membership is certainly a good idea, but I'm afraid that more is needed. I guess you've read the sample comment exchanges in my post. I wrote them that way for a reason. Note that none of the comments use offensive language or are insulting in the usual sense. They're actually quite innocuous, as long as you only look at one at a time. It's the whole sequence that can drive you crazy. I don't know how to prevent that sort of thing, short of having the whole site monitored and moderated proactively by an actual human being who has no qualms about using the "delete" button.
    • CommentRowNumber16.
    • CommentAuthorScott Morrison
    • CommentTimeFeb 27th 2012
    • (edited Mar 6th 2012)

    Hi Izabella,

    I did like your example threads. I’ve certainly seen those dynamics in action. It’s going to be interesting for an article discussion site, because I think moderators, even if willing, are going to feel constrained about actively intervening. On MathOverflow, we were pretty comfortable deleting comments, outright banning people, and so on, because it didn’t feel like we had any particular obligation to host people who were damaging the site (or more particularly, its “community standards”). Even there we’ve had plenty of rudeness and unpleasant interactions. But on a site that aspires to become a general model for discussing papers, I think moderators are inevitably going to feel that “everyone has to be allowed to have their say”. I would certainly feel that way, and I’m by no means volunteering for any such moderating jobs!

    Maybe this is all to say that we do need to think hard about how moderating can work: efforts to automate keeping people to “professional” (maybe even better than that…) standards may be even more important, because it’s going to be harder to moderate on an ad hoc basis.

    • CommentRowNumber17.
    • CommentAuthorChris Lee
    • CommentTimeFeb 27th 2012
    • (edited Feb 27th 2012)
    @Andrew: all very good features. I'm curious to hear your thoughts on a basic question about tagging philosophy. Some people seem to think about tags as 'atomic' terms which they then combine to specify a particular field, whereas other people use a highly specific phrase as one tag. I was thinking in the latter terms, because I want people to be able to define their research interests specifically. This way, a large fraction of papers marked with their desired tag would be relevant to their research interest. This also aids searching, since most searches will just use a single tag and will return a relatively small set of results (whereas the former approach typically requires intersecting multiple tag searches, each of which returns a large set of results). It sounded like you too favor the 'well-specified tag' approach; is that so? And if we want this to be the default, how should we make this clear to people? It seems like we would want to avoid user confusion on this point.

    Second, most of the discussion about tags seems to focus on a specified field or topic. But of course tagging could be used any way people want to (e.g. define a tag 'papers that suck'). Do we want to keep a clear division between field-tagging vs. other tagging? (If the system does not know which tags are fields vs. not, it could end up offering nonsensical interface choices as 'Based on your papers, it looks like your research interest is: papers that suck. Do you want to join that research field?').
    • CommentRowNumber18.
    • CommentAuthorChris Lee
    • CommentTimeFeb 27th 2012
    @Scott: based on what you and John have said about moderation, it seems crucial to set expectations clearly, not as 'everyone must be allowed to say whatever they want' but rather 'everyone who wants to participate in a research forum must follow basic research communication guidelines.' All researchers know perfectly well (from publishing papers) what those guidelines are: e.g. stay on-topic; stick to the evidence; no ad-hominem attacks, etc. It's just a matter of enforcing that standard in a friendly but consistent way. For example, if anyone feels a comment violates the guidelines, they mark it and say specifically what they think violates the guidelines. The original poster receives this message, and has the option to respond either by accepting the critique and revising their comment to follow the guidelines (to be OK'd for reposting by the person who complained). Alternatively they could explain why they think it is not a violation; if the person who complained does not agree, it goes to moderation.
    • CommentRowNumber19.
    • CommentAuthorIzabella Laba
    • CommentTimeFeb 28th 2012
    I don't know how to solve the moderation problem. Guidelines would be good, but there is also the problem of "discussions" like the fictional sample in my post, where no single comment violates the guidelines (as in Chris's comment) but the whole exchange is still obnoxious. In my experience, though, that sort of thing is often a drive-by. So maybe a graduated system of permissions would help:

    1) It should be possible for authors to enable only parts of what the site offers. For example, the author might enable the metadata functionality, but disallow comments and/or decline to link to discussions mentioning the paper in other threads.

    2) It would also be useful for authors to have different "circles" of people who are allowed to comment on their papers. I might for example create a circle of harmonic analysts etc. who work in roughly the same area as I do. (In fact, I might create several circles for papers in different areas, and then use them in various combinations.) Not on my list, but want to comment? (For example, graduate students that I don't know about yet.) Send me a request, and I could either add you to my circles or give you a one-time permission.

    If an author allows all users to post comments, they might get more action, but they will also risk the drive-by's. Their choice.

    Hope that this helps. I'm trying to come up with suggestions that would minimize the need for a moderator's intervention. A moderator might be reluctant to delete comments that don't cross obvious boundaries (so a lot of obnoxious stuff could stay on the site), and in any case you probably don't want to hire several full-time employees to moderate everything.
    • CommentRowNumber20.
    • CommentAuthordarij grinberg
    • CommentTimeFeb 28th 2012
    • (edited Feb 28th 2012)

    @Izabella Laba 19:

    I am in favor of having the technique implemented first, and solving social problems as they emerge. It is rather clear that they are solvable, and this would probably be easier to do once the project is already running.

    As for the comments: I don’t think that they “don’t cross obvious boundaries”. Comments C, D and E_1 don’t, but E_2 obviously does, being both an insinuation and offtopic. Even on MO it would have been deleted by a moderator. It is a more difficult question what to do with a polemic like E_2 entangled with nontrivial mathematical content in one and the same comment, but in your example the situation is clear. Also, Comment D in the form given by you is nonconstructive enough (both for the author and for the readers) to be deletable without loss. And Comment E_1 is a duplicate, so the same goes for it. So, while these comments are a problem, they are a problem easily dealt with by moderators, at least in the case you mentioned.

    I don’t think an opt-in for having one’s papers discussed is a reasonable option - the project will be much harder to kickstart this way, since there will be only a small pool of discussable questions initially.

    • CommentRowNumber21.
    • CommentAuthorIzabella Laba
    • CommentTimeFeb 29th 2012
    @darij grinberg:

    I think it depends on what level of mass participation you expect. If you would like to start with a small site for like-minded individuals who are already keen on the idea, they you can probably work out the social problems as you go. If you expect mass participation on the level of arXiv, then you need to make the site attractive to busy, stressed out, and not especially well motivated people who don't need any more aggravations. I would think that having the option of allowing only a selected group to comment would remove one of the main objections and go a long way towards encouraging such people to sign up.

    Personally, if I were considering whether to sign up, I might (for example) like or dislike the way tags work, but it would not make or break the deal. But if I could not sign up without allowing unrestricted comments, then most likely I would not sign up.

    The kind of moderation you are describing is fairly aggressive, and yes, there would be the problem of comments that contain something valuable but are phrased impolitely or condescendingly. You will need to find someone who will be willing to do it and will have the time. Even so, once the comments have been posted and the author (and possibly others) have read them, some damage has already been done. The author has already been at the receiving end of unprofessional behaviour and had to go to the trouble of asking the moderator for intervention.
    • CommentRowNumber22.
    • CommentAuthordarij grinberg
    • CommentTimeFeb 29th 2012
    • (edited Feb 29th 2012)

    The idea as discussed right now involves commenting on every single paper out in the wild, no matter whether the author has signed up or not. At least this is how I understood it. And for all the collateral damage this is likely to create, it is the only way that will spare the site the fate of several other similar projects that have failed - namely, the fate of having not enough material to begin with. (Okay, to be honest, some of these similar projects have lots of other downsides as well.)

    At the moment there is already a possibility to discuss (as in: ask questions about) papers on MathOverflow. While it is not considered 100% proper use of MathOverflow, people (including me) are already using it, and this isn’t the end of the world. Into every question about a proof one can read a doubt about its validity, but anyone with some mathematical experience will not a priori interpret a question as a doubt unless it has several cues in that direction. So most of the comments will not cause damage, and this is reason enough for a commenting system to exist.

    The kind of moderation I describe is, I think, typical for MathOverflow. Or at least on MathOverflow the comments D, E_1 and definitely E_2 would receive enough downvotes to get negative weight in the eyes of most readers. Without voting this might not be that obvious, but one still shouldn’t be modelling the reader as an idiot who believes everything he reads on the internet.

    • CommentRowNumber23.
    • CommentAuthorIzabella Laba
    • CommentTimeFeb 29th 2012
    OK, I guess everyone can discuss anything they like on an internet forum. My impression was that the site was intended to have some kind of an "official" function, for example a paper might be required to spend some time on the site as part of the refereeing process (I think this was mentioned in one of Gowers's posts), or that the site might be coordinated with arXiv. I was writing from the assumption that this is where things might be going.

    But this is perhaps getting off topic in this thread. I responded here because this is where Scott linked to my blog post, but if more discussion is needed, I don't mind starting a new thread on moderation (if there isn't one already).
    • CommentRowNumber24.
    • CommentAuthorKevin Walker
    • CommentTimeFeb 29th 2012

    Re #23: I think there are two somewhat incompatible sets of assumptions people are making about the site. Some think of it as a substitute for or adjunct to the publishing and refereeing process. In this case it seems natural for authors to opt into the system, just as authors now choose whether and when to submit their papers to a journal. Others think of it as an annotation and discussion site, like Wikipedia or Math Reviews or MathOverflow, but (perhaps) with an emphasis on specific papers. In this case it seems more natural that all papers are potential topics, regardless of the author’s wishes. I think the second model is closer to what the implementers have in mind.

    • CommentRowNumber25.
    • CommentAuthorAndrew Stacey
    • CommentTimeMar 1st 2012

    Certainly for the Selected Papers Network (Izabella: please note that this discussion site is about far more than just that proposal!), I would be very uneasy if it were used for evaluation purposes so I vote very strongly for the second interpretation.

    Chris: re 17. In all honesty, I don’t know! I would say that the private and public tagging systems should be almost completely independent. I should be able to set up a correspondence between my private and public tags, and the system should only do things like make suggestions based on the public set. So I may well have “Down right wrong” as one of my private tags, but the system certainly should not suggest things like “Tim Gowers thought that paper X was also ’Down right wrong’, would you like to add it to your list?”.

    My own tagging system is completely organic: I choose whatever tags seem right when I bookmark a paper with no particular thought as to a strategy. It would be useful, when I do so, to have a list of the tags I’ve chosen before. You can see my list at http://ncatlab.org/lspace/list (they’re called “categories” there). Actually, let me repeat it here and comment on them:

    Tall-Wraith monoids, algebraic theories, algebraic topology, category theory, cohomology operations, cohomology theories, comparative smootheology, differential geometry, differential topology, dirac operators, elliptic cohomology, exposition, frölicher spaces, functional analysis, general relativity, generalised cohomology theories, generalised complex geometry, gerbes, higher category theory, index theory, infinite dimensional manifolds, knot theory, know theory, linear algebra, loop groups, loop spaces, mapping spaces, operads, rainy day, smootheology, string topology, teaching, topological field theory, tqft, two bundles, universal algebra

    In that list there are a lot of subject tags. However, they appear in a variety of gradations. “algebraic topology” is quite broad, but it gets refined down as one gets nearer my specific interests: “algebraic topology” → “cohomology theories” → “cohomology operations” → “Tall-Wraith monoids” (I see that I have “cohomology theories” and “generalised cohomology theories” a good system would let me make those two aliases of each other; also “know theory” is certainly a typo).

    There are also some tags which related to specific papers that I’ve written: “Tall-Wraith monoids”, “comparative smootheology”. There, I’m categorising by which part of my research it is most related to.

    Then there are some “meta” tags: “teaching”, “rainy day”, “exposition”.

    There are certain tags where I’d be happy for the system to make assumptions, and other tags where I wouldn’t. Problem (for the system) is that only I know which ones these are. “Comparative smootheology” and “Tall-Wraith monoids” are ones that would be fine: anything tagged with either of those two is highly likely to be referring to one of my articles (or if not, I want to know about it!). On the other hand, “cohomology operations” is probably a bit too broad for the system to know which part I’m particularly interested in. But at the other end of the scale “algebraic topology” would be fine because it’s a “top level” tag that’s fairly indisputable as to what it’s about.

    I guess that my overall point is that tagging is complicated so a certain amount is going to have to be “see what works”. But the more I think about public-private tags, the more I like it. It lets me retain control over what information the system can use in its “social” aspect.