Monday, March 24, 2008

The science of interaction

Our guest blogger, Tirta, raises the issue on the individual vs. collective behavior, and the importance of taking the distinction into account.
- Manager
The science of interaction
by Tirta

I have the following issue with Milton Friedman's dictum that economic assumptions are only as good as the predictions they make.

The dictum implies that in any economic model, what really happens in individual minds doesn't really matter, as long as the collective phenomenon of interest can be predicted. In fact, one of the criticisms against the recent interest in behavioral economics is that at the aggregate level, there's no such thing as irrationality. Individuals may be irrational, but society as a whole is rational – as shown by the success of rational models in their predictions of many collective actions.

Now this is all fine, if there is only one explanation behind each phenomenon of interest. But what if the phenomenon of interest can be predicted by two differing models built upon two dissociable assumptions behind the action of the individual – one being more psychologically realistic than the other? Should we choose the simpler assumption (i.e. homo economicus) for the sake of neat predictions? Or should we pick the more realistic one (e.g. homo behavioral-economicus, homo neuro-economicus) at the expense of perhaps less clear and less fruitful predictions?

I think there is a danger in the tendency to prefer simple and neat models to complex and more complicated ones. Parsimony as a scientific criterion doesn't always apply, at least when it comes to explaining how the human mind works. A century of psychological explorations have shown that the human mind and brain are necessarily complicated and, as far as reality goes, cannot be further simplified.

So it seems that we are left with either making good-collective predictions based on bad-individual assumptions, or making bad-collective predictions based on good-individual assumptions.

Can we make good-collective predictions based on good-individual assumptions?

In principle yes, and I think the key lies in understanding how the actions of individuals result in collective phenomena. The more we understand about the science of interaction, the less we have to rely on making unrealistic assumptions about individuals, while at the same time maintaining our accuracy in predicting the emergence of collective actions.

Now I personally don't know how far we've come with the science of interaction. But I sense that, if anything, the end result of understanding how different individuals interact with one another in collective settings would be better and more fruitful economic predictions.


  1. First, I must say that Network science is by far the most systematic and comprehensive attempt for the science of interaction.

    Yet, there is a division here: One camp believes that there is a general theory of network science that can be applied in many different fields e.g., social, physical, biological, or culture, linguistic, history etc.

    Another camp thinks that universal theory is impossible. For example, social network theory should be different from gene-expression network theory (if there are such things).

    I fall to the later camp, i.e., network means almost nothing without paying attention to the process that we're studying.

    Second, another way to frame the problem is aggregation vs reduction. Interaction is crucial for explaining aggregation process.

    Thus, the problem is not whether the assumption used is realistic or not, but whether we can understand collective behavior completely by only focusing on the micro/individual level (even with very realistic assumption).

  2. roby, sometimes, i just really wish i could understand what you're talking about.

    it sounds really smart

  3. Tree,

    Yeah, those posts take a loong time for me to read. Let see if I can get them right.

    Tirta and Roby,

    If I get it right, you believe that this interaction science can explain how interactions between irrational individuals cause the collective (as a group) to act rationally. It's like explaining how interactions between human right advocates can make the collective (the group they belong to) to be at advantage.

    Note here that I believe you define rational action to be action that put the actors into advantage.


    Anyway, IMHO, I think due to natural selection (of groups consisting of various individuals type), all collective behaviours of all (naturally) existing group will almost certainly be rational, because all the irrationals groupings won't live long (because the disadvantage will slowly put the members at disadvantage till the end). The member of irrational groups will either (unlikely) die or (more likely) disperse themselves.

    All surviving grouping will exist commonly and long enough to be studied, and the result will definitely show how they act rationally. All the other groups won't live long enough to be of study interest.

    That's why all existing study of collective behaviour will often show how rational they are. This phenomenon can probably make the scope of possibilities in interaction science smaller.

  4. treespotter: if it sounds really smart, it is really smart.

    amitz sekali: i'm not talking about rationality. network modeling can use rational or irrational agents; doesn't matter.

  5. Roby,

    Then is what you're talking about boils down to: understanding of individuals + understanding of interaction between individuals (network science) = understanding of collective?

  6. proof that roby is right: the discussion inexplicably revolves around resolving what exactly what he was talking about.

    Socrates had allegedly done this exactly once.

    I'll let you two bond in peace.