The explainer
Talk of AI "deciding" often smuggles in a folk theory of free will. This explainer separates felt choice, causal models, and goal-directed systems so readers can ask better questions about agency without needing the night's seating chart.
Felt choice is not a metaphysics
Hard determinism says every mental state has prior causes. Compatibilism keeps a practical sense of acting on reasons. Stronger accounts treat novelty, time, or knowledge-creation as room enough for real choice. None of those positions is settled by how authorship feels. Introspection can report a decision arriving unbidden and still leave the useful question standing: what capacities make a system meaningfully agentic?
The operational list is shorter than the metaphysical one. Goals. Reasons. Prediction error. Self-modification. Responsiveness to consequences. Those features do not prove libertarian freedom or subjective experience. They do give a way to compare people, animals, institutions, and artificial systems without pretending “it outputted a plan” is the same as “it chose.”
Blame, goals, and machines
Moral responsibility is the pressure point. If behaviour is caused, communities still have to decide how accountability, rehabilitation, incentives, and care should work. That is a civic problem, not a proof that the metaphysics came out one way.
For AI, the useful cut is between inheriting an objective and setting one. A system can optimize a target it was given. That is not yet evidence that it authors goals, feels authorship, or deserves blame. Agency talk gets cheaper every time a product is said to “decide.”
Session history
Deep Dive #2 was the capacity night (May 22, 2025, Vancouver): first sold-out MAC session, waitlist of three. The room stress-tested Sam Harris’s hard-determinist frame against Lee Cronin, Dennett’s compatibilism, and the Assembly Theory / time thread. No consensus on free will. A deeper question emerged: maybe free will is the wrong frame, and agency, goals, and prediction error are the ones that travel.
The public record keeps speakers and sources as labeled history. It does not publish participant messages or reconstructed dialogue.
Nearby questions in the series
If agency needs a history of selection, read whether consciousness evolved for survival, not selfhood. If choice needs stakes, see whether consciousness requires a body. If “deciding” is being confused with grasping, the companion page is what it means for AI to understand.
Further reading
The curated shelf is the agency current of the MAC Library, including the Making Sense of Free Will entry. Search receipts on the Community Source Index. The MAC group home remains a sibling, not a canonical winner, for the next night and how to join.