Intelligence begins
with what we
notice.

A probability tells us which explanation an agent favours. It may tell us less about which alternatives the agent can articulate, or which observation would distinguish them.

Current direction Testing whether alternative coverage and discriminator selection add predictive value for later revision.

When a model commits to one interpretation, what happens to the alternatives? I’m interested in connecting observable behaviour to representations in open-weight models.

Planned extension Studying representational correlates of interpretive capture, without assuming that an unspoken alternative is absent internally.

A group can produce a variety of statements while sharing the same mistaken premise. The question is whether its discussion keeps genuinely different explanations available.

Open question Connecting individual readiness to revise with collective patterns of convergence and dissent.