Artificial organisms
“What I cannot create, I do not understand” (Richard Feynman). One way we test how behavioral principles integrate is to build them. We construct artificial organisms that combine principles and processes under a set of assumptions, then watch what reproduces real behavior and where things break down. Where the lab studies how processes interact in living organisms, this program asks a related question: can the same processes, assembled from first principles, generate the same patterns?
What each approach reproduces
Every approach to building an artificial organism is an attempt to regenerate behavior we already measure. The principles themselves are long established, so the grid asks only one thing: who has reproduced each one with that specific AO approach? Each row is a behavioral process or phenomenon, each column is an AO approach. Purple means our lab did it with that AO; blue means another researcher did; a split cell means both have; an empty cell means no one has yet. Click a cell for the work behind it.
Every column is sourced: the Molecular dynamics and Dynamical Integrated Principles columns from our own repos (ours are the only such organisms, so no other-researcher layer), and the ETBD, MPR, and Q-learning columns from our lab repos (purple) plus a review of other groups' AOs of that approach (blue). Some reinforcement-learning cells hold only with augmented variants (latent-cause or distributional RL); hover for the specifics.
Artificial organisms (2)
RL in behavior science and AI
Cox et al · 2026 · DOI