Theory 08/Consciousness & Attention/Attention
Attention Trades Precision for Bias
Attention is described as a spotlight or a filter. It may be a single parameter that trades two things.
Attention operates by trading sensory precision against prior bias, and the operating point is set by the ratio of expected reward to sensory noise, so the same cue can sharpen or bias perception depending on that ratio.
01The evidence
Attention both improves discrimination and shifts appearance, and the two effects vary across conditions in ways a single spotlight model does not capture.
02The tension
Gain models predict that attention improves precision without bias. Bias models predict the reverse. Both effects are observed, often together.
03The mechanism
Perception weights evidence against prior. Attention adjusts the relative weighting, which necessarily trades precision against bias because the two are linked by the same parameter.
That makes the trade-off the mechanism, and predicts that the ratio of expected reward to noise sets where on the trade-off a subject sits.
04What this is not
Those each predict one effect. This predicts both as one parameter, which explains their co-occurrence.
05Predictions
- P1Precision and bias should covary with cue value.
- P2The trade-off should shift with noise.
- P3A single parameter should fit both effects.
06Refutation
Precision and bias varying independently would remove the trade-off.
07What it changes
It would change how attentional deficits are measured.