Rational Choice Theory and Bayesianism

Rational Choice Theory and Bayesianism

Shubhamkar Ayare -
Number of replies: 0

In the broader picture, I'm more interested in finding a [possibly impure/stateful] function called "mind" that takes in an "environment" as an argument, where the environment could be as general as possible, and the same "mind" function can compute the behavior over the lifespan of the organism. Towards this, two of the issues I have in common with both Bayesianism as well as Rational Choice Theory include:

1. Most often, I have seen both frameworks being applied to forced choice environments. But the real world seems fairly complex.

2. In both cases, the individual is assumed to have a fixed motivation in the context of modeling. That works well for the course of an experiment. But in reality, our motivations change, eg. when we are hungry vs full, tired vs energetic (which affects altruistic behavior?), lonely vs socially-full, and just waking up on the wrong side of the bed! 

That's my popular impression of the two frameworks, and I'd love to be told, "No, you haven't read XYZ papers. These models have also been applied in settings where the choice-options themselves were very unclear." or "No, the real world isn't really complex. At any given moment, we have a small number of choices that are determined by ABC factors particularly due to constraints of our working memory." I was also happy that both the lecture and the Al-Shawaf et al.'s paper addressed concerns about just-so stories, that so long as you use the information derived from observations to make predictions, what you are doing goes beyond just-so stories. (Although, the notion of prediction is a separate topic itself I want to dive in later.)

I'm gonna read Braitenberg's book on Vehicles sometime! Thank you Luisa and others for the recommendation earlier!

Coming to differences between the two, I think the critical difference is Bayesian modeling involves updation of priors-posteriors on the basis of evidence, while RCT makes no such assumption, unless perhaps that is encoded in a time-and-data evolving utility function itself.

---

Tying to my research work, I have seen (but yet to analyze :() Bayesian frameworks being applied to causal cognition. I want to address the first problem above in the context of causal cognition. The Bayesian model is already given variables handed down to it by the human researcher. I'm more interested in finding where the child gets those variables come from.