Questions

Questions

Luisa Andreuccioli -
Number of replies: 1

1) Reflect on the use and possible misuse of the model
From my understanding of Rational Choice theory, I think the model is useful precisely in identifying and shedding lights on biases that humans are susceptible to. At the same time, the most immediate example that comes to my mind for a misuse of this model, would be to assume or to try to explain humans' behavior merely according to rational choice theory. 

2) How does Rational Choice Theory compare to Bayesian models?
When trying to compare Bayesian models to Rational Choice Theory, the first thing that comes to my mind, is that Bayesian frameworks are likely involved when an agent tries to act in a way that would be predictable by Rational Choice theory. However, and at the same time, one way that these two models differ, is that Bayesian reasoning would also underlie the kind of reasoning that would lead to choices that would not be predicted by Rational Choice Theory.

3) Could the rational agent model be useful for your research? How?
I am mostly thinking about cognitive development in general to reflect on this question. I think that the extent to which the concept of rational agent can be applied to that of an infant learning about the world around them, is that, in the context of learning, we also have to maximise benefit and reduce costs (first of all, within development, the most efficient strategy is not to learn everything that there is to learn, because that would be too costly and not pay off, i.e. it would not be representative of what one will need later; in other words, the best way to maximise learning while reducing the cost is to just 'learn' about what the specific environmnt requires you to learn. A good example of this is that if an infant grows up with humans they will become adept at distinguishing human faces; but if they were raised by monkeys and had no humans around, they will become adept at distinguish monkey - but not human - faces.

There are other resemblances too between the rational agent model and cognitive development. Specifically, trying to maximise benefit in one's own learning might mean very different things depending on the circumstances (e.g. investing a lot of effort in understanding and learning about how objects interact with each other in the physical environemnt, gravity, causality, etc. at the expense of, for example, learning about the functions of those objects. In an identical context, an older infant, who has had significantly more experience with objects, might turn their focus on the functions of those specific objects, or towards extrapolating other inductive rules). 

In reply to Luisa Andreuccioli

Re: Questions

Christophe Heintz -
1) 100% in agreement with what you are saying. This corresponds also with my suggestion that Rational Choice Theory can help formulating the null hypothesis, and then we can obtain signatures of the actual algorithm with observed systematic departures from the null hypothesis.

2) I hope I specified what I think the relations are during the session. They don't seem to correspond to what you write, but I am not sure.
What I said:
a. To the extent that RCT is a form of belief-desire psychology, it involves the "rational" formation of belief. Rational belief formation is best describe by Bayesian processes (at least when there is uncertainty). Then RCT specifies "instrumental rationality": selection of the most adequate means to attain one's goal/satisfy one's desires (in view of one's beliefs).
b. To the extent that cognition is just one type of action, it will be constrained by the costs and benefit of computing this or that information. For analysing this aspect, you need to analyse expected utility ... i.e. use the tools of RCT.

3) You've restated the point 2.b above in a much better and more precise way than I did. I fully agree with what you say. If you want more information about utility calculation of learning, just come to see me.