Bayesian Mechanics of Economic Choice: Computational Foundations of Economic Behavior

Hello, welcome. It is June 26th, 2025 and we are live in active inference guest stream 113.1. Welcome Jack as well. We are with Sergio Jacis and we will be discussing basing mechanics of economic choice computational foundations of economic behavior. There will be a presentation followed by a discussion. So thank you again for joining and looking forward to the presentation. Thank you for joining Jack and thank you for the active inference community for inviting me to share my my current research. Uh as Daniel said I am Samuel or Sergio I have both names. Uh I am residing in Mexico City. I am a secondyear PhD candidate in artificial intelligence and my current research basically focuses on applying the active inference framework to understand economic choice and financial decisions. Uh just to notice that this is joint work with Ernesto Moya, PhD, my thesis supervisor and Luis Alberto Casada, PhD as well, which are two Mexican academics as well. Well, let's begin. So basically I'm going to be speaking a lot about this paper that we recently published but I'm also going to give you a brief introduction about the main themes that we are trying to bring together which are basically topics from neurokconomics topics naturally from basin mechanics and the active inference framework and topics from classical or neocclassical economics and psychology. So we're given that we are trying to merge all of those fields and as you probably know here in the community the free energy principle is very interdisciplinary itself. So I think I would take some time to explain the intellectual journey of these different fields and how we think they nicely come together. So we'll start with economics. Then we would move to how psychology meets economics. Then a little bit of the thermodynamics of information both for neuroeconomics and active inference and then how the divisive normalization model from neuroeconomics can be derived formally from the free energy principle. Uh so that's the basic or the key insight from our per our paper. And finally I'll I'll share some of the future directions or the future work that we intend to pro to do for this year or the end of this year. So let's start. So I would like to start by by sharing the let's say the intellectual journey of how economist usually think about valuation and about encoding value. So let's start with the Swiss mathematician uh Daniel Bernuli and as you probably know he's very well known for creating uh classical utility theory or expect or classical expected utility theory which basically said that that economic agents or people don't encode does don't encode absolute monetary values. We don't encode absolute monetary values. we encode what he called moral value and moral value was just a way of saying subjective value. So monetary values are are I mean are subjective on how much we want something on the time we are on the on the place we are. So basically that's what he formalized and he formalized it as a weighted sum of expected probabilities and payoffs. Now later on very famously von Noyman and Morgan stem published their famous work theory of games and economic behavior where many of you listening uh probably have heard it because it that was the the foundation of game theory but one key insight and was that when they published this work they just expanded Berni's expected utility theory to account for the specific axioms that economic agents should abide to in order to be what we usually came to to refer to as rational to make rational decisions or rational rational choices. So these uh these axiums were that our decisions would be should be complete they should be transitive and this this is very very important. So when I prefer A to B and B to C uh we should follow this this this ordinal uh property and our choices should be value independent independent of the other choices independent of past choices. So that was another another choice and if we if our rational agent or our…