ActInf ModelStream 015.0: Andrew Pashea "Active Agents: Agent-Based Modeling in the Social Sciences"

"Active Agents" An Active Inference Approach to Agent-Based Modeling in the Social Sciences"
Andrew Pashea

This talk presents the construction of Active Inference agents as discrete-time, discrete state-space, partially-observable Markov decision-making processes (POMDPs) using the pymdp library in Python. This is part of an initiative to develop and provide further learning resources for learning and applying Active Inference. While learning Active Inference, as with any other complex field, can initially seem daunting, learning to apply Active Inference 'in silico' can be enlightening for non-programmers and programmers alike as to the mechanics involved and spur inspiration for designing one's own models and simulations.
Following both the code and verbal explanations, we look at defining states, observations, and actions. Then we relate these three components together via probability distributions representing our agents beliefs. Then we cover some simple code where our agent can receive observations and use them to infer its posteriors over hidden states, as well as infer its posteriors over policies and samples an action, which is sent to the environment to generate a new round of observations. Finally, the agent learns its likelihood model given the new observation. This process defines a simple action-perception (or, perception-action) loop.

The presentation is adapted from materials for three-hour workshop at the International Conference for Computational Social Science held at University of Pennsylvania in July, 2024.

Link to all materials (slides + code) and other growing learning resources:
https://github.com/apashea/IC2S2-Active-Inference-Tutorial
Tutorial script with Markdown explanations:
https://colab.research.google.com/drive/1VWvwZFzQlNdL8w1X8UQ1VWNBESZ32_AW?usp=sharing
Original tutorial description:
https://ic2s2-2024.org/tutorials

----

Active Inference Institute information:
Website: https://activeinference.org/
Twitter: https://twitter.com/InferenceActive
Discord: https://discord.gg/8VNKNp4jtx
YouTube: https://www.youtube.com/c/ActiveInference/
Active Inference Livestreams: https://coda.io/@active-inference-institute/livestreams