Distinguish learning from feedback artifacts and decoder fitting
Bounded experimental precedent / proposed causal test. Closed-loop neural-culture experiments motivate tests of adaptation under feedback. DishBrain reported results in an embodied game simulation; its title’s use of sentience is the authors’ terminology, not proof of subjective experience. Likewise, improved prediction by a digital readout does not alone demonstrate persistent biological learning.
Bounded experimental precedent / proposed causal test. This page expands the “Distinguish learning from feedback artifacts and decoder fitting” chapter in the neuroscience collection.
Closed-loop neural-culture experiments motivate tests of adaptation under feedback. DishBrain reported results in an embodied game simulation; its title’s use of sentience is the authors’ terminology, not proof of subjective experience. Likewise, improved prediction by a digital readout does not alone demonstrate persistent biological learning.
Necromancer proposes separating contingent feedback, matched noncontingent replay, baseline drift, and decoder-only learning. Prespecify the behavioral proxy, retention interval, stimulation exposure, artifact rejection, and analysis window. Separate biological preparations, not just correlated time windows, across evaluation splits. Record how much of any gain disappears when the trained decoder or the feedback contingency is removed.
Proposed gate: a reproducible feedback-specific effect that survives controls for exposure, preprocessing, decoder flexibility, and between-batch variation. Recovery of an experimentally induced response remains a circuit-level result. It does not establish autobiographical memory or personal identity. A useful near-term tool would detect reproducible functional change in an injury model before attempting broader interpretations.
Open this exact section in the Necromancer dossier. The original chapter status is “Bounded experimental precedent / proposed causal test.”
Take this topic further
Turn “Distinguish learning from feedback artifacts and decoder fitting” into a learning path
The Learning Studio will use this page’s overview and key ideas to shape a guided lesson, practice questions, and a small project around what you want to understand.
Related GiveGigs pages and features
Explore connected GiveGigs pages and tools for context and next steps.