Games With A Cause

Some time ago I wrote about various attempts to gather common sense, the lack of which is believed to be one of the main hurdles to creating successful AI systems capable of human-like interaction.

A few weeks later I wrote about people as parts of computer systems to make them cheaper or more intelligent. Some of the best examples are when you can tap into the normal usage of Internet users to create something valuable as in the evolving banners example.

I began wondering if there was a way to make an incentive for Internet users to build a common sense database and actually giving them something in return. My suggestion: “Games With A Cause” – you play, we get a bit of common sense.

There are two ways to get Internet users to work with you. Either the work must be a part of their normal usage so they don’t really notice that they are doing any work (the evolving banners would be an example) or you have to give them something in return: a good service, monetary value or simply a good time while at it.

With that in mind I came up with an idea of making a series of simple Internet games where users could play and while at it, they would be building a database of common sense “knowledge bits”.

The first game is ready. The game is called Norm, and vaguely resembles the TV show “Family Feud” (however without any association with it whatsoever). The objective is to guess other user’s most common answers to simple questions. Norm is totally real time, so your answers will have changed the “correct” answers when the next player comes across the same question.

Now I’m certainly not claiming to be any kind of an AI expert and I probably will not make any kind of AI system from the gathered data myself. The data gathered in the current set of questions also has limited value, but I see this now more as a proof of concept. If it turns out to generate any traffic at all, the set of questions in Norm will be updated and it will encourage me to create a few more Games With A Cause (GWACs).

I have a few concepts in development, notably a search game where the player is asked a non-trivial question (e.g. “How many people have traveled to the moon?”) and he or she supposed to search the Web for an answer (“12” in this case). The gathered data in that case might help an AI system to form effective queries to search the web or at least give clues to how to structure such queries – and then we might be talking about some real value.

Or maybe all of this was just a long excuse to get the grips of PHP and MySQL 😉 But – hey – I haven’t really programmed anything for almost a year, so I was feeling a bit of an urge. In any case I hope Norm is fun to play.

If you have any comments or suggestions, please contact me. I will also be happy to send anyone the data that has been gathered in the GWACs at any time.


  1. Interesting thoughts, which blends several interesting threads.

    First, for an obscure question “number of phones in Uzbekhistan” the group average tends to converge to the best estimate, even though high variance. (Similar to concept of new exchange to trade “terror futures.”)

    Second, this “Delphi” related effect can improve operations when information is missing, in a “neo-Baysian approach to risk and uncertainty.

    Third can information-exchanges be structured to manage panics? That is, stocks are priced efficiently, EXCEPT on 10/7/83 etc.

    Fourth, how do we avoid trap of locally-optimal solutions, trapped in a well, which ignore better strategies, somewhat further away?

    Like to share thoughts on this chain, and natural application of game theory to business.

  2. Good points.

    Information about the “terrorism futures” Robert mentions can be found in this Wired article.

    Another similar project is Technology Review’s also has Innovation Futures which implements a market space for various trends, with the aim to predict the most likely outcome (e.g. “Will SCO be awarded damages?” and “When will Google have an IPO?“).

    The crisis issue you mention is a very interesting one. Based on recent research one might however doubt how effectively stocks are priced even when there is no panic as stock market decisions are no different from random decisions. That aside the stock markets already have their ways of dealing with panic situations, e.g. with halts and delays of trading in case of significant events or “where there is a significant order imbalance between buyers and sellers in a security”.

Comments are closed.