Given the rate at which Stack Overflow accrues questions, would it have the capacity to look for patterns of questions asked about programming languages or implementations or patterns in solutions/implementations.

If so, would it be capable of categorising posts (q/a) in terms of things you might be expected to ask 'at your observed stage of questioning/answering' within some kind of normalised distribution?

It might be helpful to me to know that if I'm asking about something now, that I'm likely to want to find out about some related (but unknown, to me) topic later.

I'm wondering what the capacity is for something like this. I mention psychohistory because recognising patterns relating to human behaviour to work out what small changes can be introduced to result in a statistically-knowable future outcome would seem to fit the kind of situation Stack Overflow is in. It has direct question/answer situations occurring regularly, with patterns of thought in terms of problem solving that it might be able to influence by introducing programmers to concepts they don't yet know they need, making it likely that they can cut the wasted time at a later date to learn something when they could learn it sooner.

Any ideas on whether or not SO can / should implement something like this?

Suggested Heading:

"Topics you might find interesting"


On SO: Machine learning how to use the Facebook interest of a users to give a decision this question references an implementation suggestion for machine learning.

Also on SO: What machine learning algorithm would be best in this scenario? a comment has an interesting and potentially related problem (what movies you might like) that could relate to 'What questions you might ask'.


I would like to see something like the following, assuming I ask a question tagged as PHP and MySQL and asked about "how" to "connect" to a "database":

Topics you might find interesting:

Abstracting DB away from applications Tagged as Database and Database Abstraction

What's the best way to abstract the database from a PHP application? Also tagged with ORM and design patterns

Something like this can encourage my learning from basic database access learning to patterns of usage I may (should?) come across later.


If anyone is interested: http://www3.imperial.ac.uk/newsandeventspggrp/imperialcollege/eventssummary/event_11-12-2012-13-34-29 there is a talk on this topic, which may be of interest to any SO staff in London at the time (just found this).

  • 3
    Psychohistory? As in Asimov's Foundation? – yannis Feb 5 '13 at 0:32
  • Yup :) the same one - You could say "predicted topics of interest", but it didn't fit the title space - I ran out of ideas for a phrase. – MyStream Feb 5 '13 at 0:33
  • @MyStream I downvoted your question, but changed my mind about the downvote. Your question is relevant and I cannot see any flaws in your question that can be corrected/improved; my downvote was simply because I do not think this is a feasible idea. For one thing, there are no precedents that I know of, and for another, it would involve major changes to the engine that would cost months of developer time. If you know of an existing implementation that inspiration could be drawn from, this might be a good idea. – Asad Saeeduddin Feb 5 '13 at 0:36
  • 1
    Ah - Thanks for the explanation - perhaps I should adjust the question to be more implementation oriented? e.g. If someone asks a question about 'x', similar questions about 'y' exist in the same domain asked by the people who asked 'x' earlier... or something better. But it was that kind of relationship I was thinking about - standardised in some way. – MyStream Feb 5 '13 at 0:38
  • @Asad I've added 2 references on SO to potentially related problem/solution situations. – MyStream Feb 5 '13 at 1:00
  • 3
    No. There are waaaay too many, uh, mules around. – Shog9 Feb 5 '13 at 3:43
  • A Mule being what, in this case, the person asking this question or are you suggesting that implementing anything remotely like a recommendation engine would be impossible on SO due to some Quality factor that can't be accounted for in some practical way? – MyStream Feb 5 '13 at 3:52
  • 1
    @Asad, of course there are precedents: on Google, if you misspell a word, it says "did you mean xyz?", and on Amazon, when you look at an article, it says "other people who bought x also bought y". It is not inconceivable on SO to say "other people who asked about x went on to ask about y...". I'm not saying it would be feasible, nor that the well isn't so polluted that you'd just get, uh, stules in return... – Benjol Feb 5 '13 at 6:14
  • This looks like a nice task for the Stack Exchange Data Scientist, if one ever got hired as mentioned here – AakashM Feb 5 '13 at 8:30
  • 1
    @benjol google does not analyze the concepts underlying the search terms, which is what this idea would entail. Additionally, for obvious reasons, "people who have asked this question before" is not a good idea on SO. Those are not precedents. – Asad Saeeduddin Feb 5 '13 at 15:22
  • In its simplest case, sequential questions with similar tags by the same people averaged over their word density, removing stop words could give you a first approximation to see if there is any value it taking it further at all, especially when taking into account the relationship between keywords and tags. – MyStream Feb 5 '13 at 17:28

You must log in to answer this question.

Browse other questions tagged .