The problem with web 2.0 crowds: imitative!

I attended a talk by David Sumpter on “How animal groups make decisions” (hosted by Max Reuter). David is a mathematician working on self-organisation and decision-making based on simple rules. His team looked at behavioral rules that explain, for example, how birds fly together.

My take-away from his talk: Group decision-making may be better than individual decision-making ONLY if each member of the group takes decision indepedently. Indeed, idependence is one of the four elements to form a wise crowd. Alas, I think that social prejudices make it impossible to reach independent decisions in our society. Does this suggest the end of the wisdom of crowds?

Few scribble notes:

>>> Model for two birds
They build a simple model of how two birds x and y change position. They discovered that two forces are at play – each bird moves towards its landmark and towards the other bird (if two birds are close by). They built a model of these two forces; the model nicely fits real data. They also found that birds do better (speed/accuracy) when they are in pairs opposed to when they are alone.

>>> Model for larger groups
They looked at sticklebacks. They tested Condorcet’s jury principle, which says that, if members of a jury take their decisions independently, the probability of the jury making the correct (majority) decision sharply increases with the jury’s size. They found this to be true for sticklebacks – the bigger the group size, the higher chance each individual stickleback follows the bigger member in (the leader of) the group. Plus, the bigger the group size, the higher chance a fish makes the ‘correct’ decision. They modified Condorcet’s principle and came up with an effective predicitive model.
They also looked at how ants migrate to nests of different quality. During the migration, each ant goes through 4 possible states: the ant 1) searches for sites; 2) assesses a site. If quorum threshold reached for the site, then the ant leads a 3) tandem run; otherwise 4) it goes into transport mode. A common feature of group decision-making is the use of quorum responses (individuals respond only when they see a threshold number of individuals perform a particular behavior – more). Even humans use quorums in decision-making.

2 Responses to “The problem with web 2.0 crowds: imitative!”

  1. mike says:

    Could you elaborate a bit more on what you take Sumpter to mean by ‘independence’? It sounds like birds, fish and ants all take the actions of other members of the group into account when making decisions, whereas Surowiecki seems to argue against doing so – is Sumpter saying it’s possible to consider the actions of others but still decide ‘independently’ in some sense?

    I wonder whether the difference between one-time and recurrent decisions might also be important: in groupthink, the order in which people declare their opinions is highly significant – it gets harder and harder to oppose the emerging ‘consensus’. A form of quorum behaviour, but very history-dependent. Perhaps this isn’t a problem for animal swarms because all members of the group continuously update their decisions? If so, then the question of synchronisation may be relevant – I’m thinking of cellular automata and the work Duncan Watts and Steve Strogatz did on crickets synchronising their chirps before they got distracted by small world networks. (Watts and Strogatz that is, not the crickets!)

  2. I haven’t understand it either :) Independence seems to be a requirement for ‘correct’ group decisions. Yet group decisions are often modeled by quorum, which excludes independence… I’m confused :)

    I’ve asked both David and Max to comment on this blog