ACHT-World: Causal World Models for Closed-Loop Self-Driving Laboratories

David Scott Lewis, Enrique Zueco

Submission Type: I want my submission to be considered for both oral and poster presentation.

Keywords: self-driving labs, agents and autonomous AI researchers, causal world models, autonomous science, closed-loop experimentation, Bayesian optimal experimental design, active causal hypothesis testing, mechanistic discovery, information gain, auditable discovery, replayable decision logs, trust layer, causal discovery, experimental design, lab automation, AI for chemistry, AI for biology, AI for materials science, protein association graphs, knowledge graphs, scientific workflows, benchmarks and software, governance, human-robot experimentation

TL;DR: ACHT-World turns self-driving labs from optimize-this engines into causal discovery systems by maintaining an explicit mechanistic belief state, choosing experiments for information gain, and logging replayable, auditable decision traces.

Confirmation Of Submission Requirements: I submit an abstract. It uses the template provided on the submission page and is no longer than 2 pages.

Submission Number: 401