Definition
Plain language
Feelings and moods — things like fear, anger, or distress.
As stated in the literature
Umbrella term for emotion-like states; in interpretability work, 'affect-like' internal directions are candidate representations of valence and arousal rather than verified experiences.
Also called: affect-like, affective
Why it matters: Affect is the thing welfare research is ultimately trying to detect, so being clear that 'affect-like' internal patterns are candidates rather than confirmed feelings keeps claims honest.
For example, when a model writes about being scared of a barking dog, researchers can look for an internal pattern that also shows up across other fear-related passages.
Heard on the show
“We don't know how alignment errors, drift, and shot noise affect performance at the scales where attention actually matters.”Episode 002 — An AI Ran a Real Optics Lab for 21 Hours and Found a Transformer-Shaped Pattern in Light