Applied AI research / Agenda in development
Research at the point where human judgment, machine-generated evidence and autonomous action meet. This is an active direction, not a claim of completed findings.
01 Core research question
How can humans safely delegate increasingly complex decisions and actions to autonomous AI systems?
From interpretation to action
Reliability is not only a property of a model. It emerges through the relationship between what a person understands, what a machine can substantiate and what a complete system is allowed to do.
How communication shapes perceived authority, trust and the decision to accept, verify, challenge or reject an AI-generated answer.
How a model’s claims can be grounded, calibrated and traced—and when a system should retrieve, clarify, abstain or escalate.
How agents plan, use tools and recover across long-running tasks—and how reliability, security and safe action can be evaluated at system level.
03 Communication → computation
Interactive communication asks how a system constructs authority and how a person interprets certainty. Dependable AI asks how that authority and certainty can be justified technically.
Questions about conversational evidence and user trust lead directly to grounding, provenance, calibration and verification. As systems become more autonomous, those same questions become operational: when should a system act, seek more evidence, ask for help or refuse?
Work in progress
Research directionPublic record
Research in progress. Working papers and publications will appear here.
No publications listedReproducible work
Datasets, evaluation harnesses, simulations, reproducible experiments and research prototypes will be kept separate from general engineering projects.
Research artifacts will appear as the agenda moves into experiments.
No artifacts listed