Evolutionary Learning of Graphical Causal Models from Behavior

H. Van Dyke Parunak

Keywords: Graphical Causal Models, Agent-Based Models, Genetic Algorithms, Genetic Programming, Model Learning

TL;DR: Reports previous work on evolutionary learning of elements of a specific Graphical Causal Model (GCM), and suggests how these can be generalized to other GCMs.

Abstract: Many causal formalisms take the form of directed graphs, in which "causality" is implicitly defined as the relation between two nodes joined by an edge from the "cause" to the "effect," and modulated by parameters stored on the nodes (and sometimes on the agents that process the graph). Parunak has recently described a generalized formalism that embraces many varieties of this kind of structure, and also demonstrated the abilities, and limitations, of synthetic evolution to learn the required structures and parameters from specified behavior (sequence of states or actions) in one specific formalism. We review the generalized formalism, describe the specific variant in which evolutionary mechanisms have been explored, summarize the results obtained, and suggest how these methods may be extended to other causal formalisms.

Paper Type: Full (minimum of 10 pages and a maximum of 16 excluding references)

Poster Opt In: Yes, I'm open to have my submission accepted as a poster

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Submission Number: 2