Not signed in (Sign In)

Vanilla 1.1.9 is a product of Lussumo. More Information: Documentation, Community Support.

  1.  
    This is a set of questions rather than a comment. Basically, how do I decide what to vote up? Should I vote for every question that catches my attention enough for me to click on it? No---particularly if I read it and find it poor. But what if I read it and find it outside my range of understanding? Still no, although its failure to get my vote reflects my ignorance rather than its merit. Suppose that I understand the question at least enough to know that it's well posed, but I don't know the field enough to judge whether it's trivial or absurdly difficult. Perhaps I should leave it to the experts to judge. What if the question is sound, to my knowledge, but too technical or specialized to interest me when I read the details? I guess that most people would not vote up in such a case, and so I tend follow them, but I might vote if there is something about the question that appeals to me. At the other end of the spectrum are the "pop" questions that everyone likes, including me. should I toss my vote onto the already towering pile of votes? What are your criteria for voting?
  2.  
    I don't over think it -- I vote up every every question I read that sounds at all interesting.
  3.  

    Vote for questions that you wish there were more of!

  4.  
    There are other aspects of voting up, other than the particular merits of a question or the assessment of one's ability to judge it. If you vote liberally, are you inflating (i.e. debasing) the currency of votes? On the other hand, it seems churlish to be too sparing with your votes---particularly when others have been kind enough to vote for your questions! Relative to the questions that one would not rule out of consideration, roughly what proportion should get one's vote?
  5.  

    Scott's rule for voting is ultimately the correct one. Some part of me is still very stingy, so I have to make other little rules to remind me to vote on things.

    • If I find myself thinking for a few minutes about the content of (even part of) a question, I vote it up. If after parsing it, I found it interesting enough to think about, it's worth voting up.
    • With answers, I try to check that they're right before voting them up, but even if it simply makes good progress on the question (like this), it's worth an upvote.
    • If I want to see more posts like this, I vote it up. Even if I don't completely understand a technical post, if it's clear to me that it's generating interest among the right people, I'm likely to vote it up. MO is really intended for technical questions, but they don't get as much love as they deserve.

    If you vote liberally, are you inflating (i.e. debasing) the currency of votes?

    Votes are not a currency. You can't really buy anything with them, you can't collect them, and when you give somebody else a vote, you don't lose anything. They are not meant as representatives of something with intrinsic value; they are a way of communicating appreciation. When you vote something up, you are saying to everybody (including the person who posted it) "Yes, I want to see more of that here." Aside from this information, you're not giving anybody anything.

    By using your votes, you are providing information that helps the system run better. So long as people are not voting indiscriminantly (i.e. so long as the information isn't noise), voting liberally is a good thing.

  6.  
    It would be interesting to know how the number of votes cast correlates to arxiv subject areas. I would guess for instance that ap votes are harder to come by than ag votes per unit question. This is a reason for folks interested in under-represented areas to take Scott's approach.
  7.  

    @Steve: that would be interesting indeed! For each tag, I counted the number of votes of every type on questions or answers within that tag. Here are just the arxiv tags, sorted in descending order by upvotes per post:

    Edit: I've updated this list to today's database dump

     upvotes per post, questions, answers, upvotes, downvotes, tag-name
     7.782364    146   920  8296   410  ov.overview
     6.310345      8    21   183    12  cs.theory
     5.887805     64   141  1207     8  kt.k-theory-homology
     5.790742    382  1195  9132   401  lo.logic
     5.663679    987  2248 18322   450  nt.number-theory
     5.632411    468  1050  8550   180  gr.group-theory
     5.508371    609  1362 10857   155  at.algebraic-topology
     5.464115    264   572  4568    72  gt.geometric-topology
     5.084048   1548  3235 24317   380  ag.algebraic-geometry
     5.041344    121   266  1951    63  cv.complex-variables
     4.922099    164   465  3096   121  mp.mathematical-physics
     4.845298    538  1259  8707   195  ct.category-theory
     4.829582    402   842  6008   135  ac.commutative-algebra
     4.808602    141   324  2236    89  cs.cc.complexity-theory
     4.807522    144   308  2173    48  mg.metric-geometry
     4.785567    579  1361  9284   247  co.combinatorics
     4.731004    326   819  5417   155  gt.general-topology
     4.659259    183   357  2516    76  ra.rings-and-algebras
     4.539792    434  1011  6560   151  dg.differential-geometry
     4.452418    412   870  5708    65  rt.representation-theory
     4.342857     72   138   912    29  sg.symplectic-geometry
     4.311688     94   214  1328    16  qa.quantum-algebra
     4.090476     68   142   859    14  ap.analysis-of-pdes
     4.077088    155   312  1904    68  ca.classical-analysis
     3.985075     29    38   267     1  sp.spectral-theory
     3.943221    291   572  3403    87  fa.functional-analysis
     3.893382    102   170  1059    20  oa.operator-algebras
     3.833333      2     4    23     1  transcend.-number-theory
     3.500000      1     1     7     0  cs.cg
     3.416011    481   918  4779   332  pr.probability
     3.031008     39    90   391    50  it.information-theory
     2.815789    101   203   856    36  na.numerical-analysis
     2.429907     44    63   260    10  oc.optimization-control
     1.908800    232   393  1193   131  st.statistics
     1.625000      3     5    13     0  cs.lg.learning
    
  8.  

    Here's the python code used to generate the data above from the public dumps:

    from lxml import etree
    
    posts = etree.parse(open("posts.xml","r")).getroot()
    votes = etree.parse(open("votes.xml","r")).getroot()
    
    def decodeTag(tag): # undo SE's weird encoding
        tag=tag.replace(u'\xe9','')  # begining char
        tag=tag.replace(u'\xe0','')  # ending char
        tag=tag.replace(u'\xf6','-') # dashes
        tag=tag.replace(u'\xfb','.') # dots
        return tag.encode('utf-8')
    
    # parent['123'] is '123' if post 123 is a question
    # or the PostId of the parent question if it's an answer
    parent = {}
    # tagsof['123'] is the list of tags of question with postId 123
    tagsof = {}
    # questionsInTag['tag-name'] is the number of questions in a given tag
    # answersInTag['tag-name'] is the number of answers in a given tag
    questionsInTag = {}
    answersInTag = {}
    
    for post in posts:
        postId = post.get("Id")
        if post.get("PostTypeId")=="1":
            parent[postId]=postId
            tagsof[postId] = [decodeTag(tag) for tag in post.get("Tags")[1:-1].split("  ")]
            for tag in tagsof[postId]:
                if tag not in questionsInTag.keys():
                    questionsInTag[tag]=0
                    answersInTag[tag]  =0
                questionsInTag[tag]+=1
    
    for post in posts: # with merged questions, the postId of answer can be lower than that of the question!
        postId = post.get("Id")
        if post.get("PostTypeId")=="2":
            parent[postId]=post.get("ParentId")
            for tag in tagsof[parent[postId]]:
                answersInTag[tag]+=1
    
    # votesOnTag['tag-name'] is a count of kinds of votes on
    # questions/answers in a particular tag
    votesOnTag = {}
    for vote in votes:
        voteType = int(vote.get("VoteTypeId"))
        for tag in tagsof[parent[vote.get("PostId")]]:
            if tag not in votesOnTag.keys(): votesOnTag[tag] = [0]*14
            votesOnTag[tag][voteType]+=1
    
    # record upvotes, downvotes, questions, answers
    updownQAtag = [(float(v[2])/(questionsInTag[tag]+answersInTag[tag]),
                  questionsInTag[tag],answersInTag[tag],v[2],v[3],tag) for tag,v in votesOnTag.items()]
    # sort by number of upvotes per post in descending order
    updownQAtag.sort( key = lambda a:a[0], reverse=True)
    
    for t in updownQAtag:
        if '.' in t[5]: #only take arxiv tags
            print '    '+' %3f' %t[0], ''.join(['%6d'% x for x in t[1:-1]])+'  '+t[-1]
    
  9.  

    Merged [ov.overview] into [ho.history-overview] and [gt.general-topology] into [gn.general-topology].

    • CommentAuthorVP
    • CommentTimeSep 1st 2010
     

    Anton, this is very interesting! Is there an easy way to supplement that table with the numbers of questions and answers within each tag?

  10.  

    I'd also be interested in seeing the total number of views per tag as well.

  11.  
    @VP: Use the tags window to get the numbers of questions. I note that tags for ds. and ho. have been tweaked, which along with gn. gives the lacunae in the following table of average votes/favorites per question in each subject area:

    up down fav num subject area
    13.1791 0.2127 2.2754 1547 'ag.algebraic-geometry'
    14.7211 0.3489 1.9493 986 'nt.number-theory'
    15.2582 0.2303 2.3931 608 'at.algebraic-topology'
    14.1732 0.3277 2.3948 537 'ct.category-theory'
    13.1367 0.3633 1.808 578 'co.combinatorics'
    15.1071 0.3383 1.8929 467 'gr.group-theory'
    17.7244 0.8556 2.1732 381 'lo.logic'
    12.6452 0.3018 2.341 434 'dg.differential-geometry'
    12.2488 0.291 1.8234 402 'ac.commutative-algebra'
    11.5364 0.1383 1.966 412 'rt.representation-theory'
    7.6965 0.5281 0.8815 481 'pr.probability'
    12.9621 0.2424 1.8902 264 'gt.geometric-topology'
    9.488 0.1856 1.3093 291 'fa.functional-analysis'
    14.2454 0.4601 2.9264 163 'mp.mathematical-physics'
    11.5764 0.2778 1.5833 144 'mg.metric-geometry'
    12.7438 0.4298 1.4298 121 'cv.complex-variables'
    7.4153 0.1803 1.0546 183 'ra.rings-and-algebras'
    12.7234 0.0745 2.3191 94 'qa.quantum-algebra'
    7.4581 0.2387 1.0774 155 'ca.classical-analysis'
    4.303 0.4675 0.5195 231 'st.statistics'
    9.0463 0.1944 1.3426 108 'ds.dynamical-systems'
    14.5156 0.0781 3.4375 64 'kt.k-theory-homology'
    11.375 0.375 2.3889 72 'sg.symplectic-geometry'
    7.6471 0.1667 1.3725 102 'oa.operator-algebras'
    10.4853 0.1029 2.0588 68 'ap.analysis-of-pdes'
    6.3663 0.2772 0.6139 101 'na.numerical-analysis'
    7.3077 1.1538 0.7179 39 'it.information-theory'
    5.1591 0.2273 0.7955 44 'oc.optimization-control'
    5.8276 0.0345 1.3793 29 'sp.spectral-theory'
  12.  
    The numbers seem high, not sure why.
  13.  

    The numbers are high because the vote tallies also included votes on answers within a given tag.

  14.  
    I see a pretty clear bias. Per-question numbers on non-analysis subjects other than cv (pr, fa, ca, ap, sp), algebras (ra, oa) and "real-world" applied subjects (pr, st, na, it, oc) are all pretty low compared to others, as are those on dynamical systems.