Search Unity

  1. Unity support for visionOS is now available. Learn more in our blog post.
    Dismiss Notice

thinking about games to explore group intelligence

Discussion in 'Game Design' started by JoeStrout, Mar 19, 2016.

  1. JoeStrout

    JoeStrout

    Joined:
    Jan 14, 2011
    Posts:
    9,859
    I've been thinking a lot lately about group intelligence, sometimes called "the wisdom of the crowd." This is where a (possibly large) group of people come are collectively "smarter" — able to make more accurate estimates, predictions, etc. — than most (or maybe all) of the individual members.

    But it matters very much how information or judgements from the individual members are combined. For example, in estimation tasks (How much does that heffer weigh? How many jelly beans in this jar? etc.), you get a much better aggregate result if you don't allow the participants to discuss it. Too much information-sharing, apparently, often results in people with wrong-headed but strong opinions having too much influence on the answer.

    Examples of some collective intelligence (CI) systems might include:
    • Predictive markets, such as PredictIt, where players bet money on the outcome of big events.
    • Um, pretty much any sports bookie ever, which amounts to the same thing for sporting events.
    • UNU, which has a unique "swarming" simulator where players tug a cursor around in a decision space in real-time.
    • Wikipedia (not even going to bother to hotlink that one for you!), which isn't intelligence, exactly, but is a pretty amazing example of collective knowledge.
    But most of these are quite slow and don't feel terribly interactive. The exception is UNU, which is quite quick and looks fun, but I have some doubts about whether it is an optimal algorithm.

    So I've been toying with the idea of building a platform in Unity for exploring different ways of bringing large groups of people together to form collective intelligence. I'd like to apply it to things that are normally much too quick for CI, such as playing a board game against a human player in real time, or even carrying on a conversation. We could experiment with different UIs, information-sharing methods, etc., and see which ones produce the smartest behavior.

    And if it's to be successful, there needs to be motivation for people to participate. That's why I'm thinking it might be best if it were designed as a game. You score points based on how well your inputs contribute to the group decision, or correctly predict outcomes, or something. If we get fancy and have tasks which can be broken down into subtasks, then we need a way for credit (points) to flow all the way to the lowest-level contributors. And maybe there are other ways to make it fun — ways to give folks a sense of accomplishment, even though they are just one part of a much larger group mind.

    What do you think? I realize this is pretty far out there, but does this inspire any thoughts in anybody?
     
  2. TonyLi

    TonyLi

    Joined:
    Apr 10, 2012
    Posts:
    12,628
    Twitch Plays? (BTW, twitchers somehow beat Dark Souls collectively last autumn.)

    What about a game that displays a 5-second timed menu of what the protagonist(s) should do? When the timer ends, it tallies the menu choices and takes the highest-voted action.