Across 20 Atari games, removing peripheral visual input drops human-action prediction accuracy by median 35–44%, far more than removing gaze maps (~2%) or past states (1.5–15%).
The role of central and peripheral vision in expert decision making
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Estimating Central, Peripheral, and Temporal Visual Contributions to Human Decision Making in Atari Games
Across 20 Atari games, removing peripheral visual input drops human-action prediction accuracy by median 35–44%, far more than removing gaze maps (~2%) or past states (1.5–15%).