An Information Processing Pattern from Robotics Predicts Unknown Properties of the Human Visual System

摘要

We tested the hypothesis that an algorithmic information processing pattern from robotics, Active InterCONnect (AICON), could serve as a useful representation for exploring human vision. We created AICON-based computational models for two visual illusions: the shape-contingent color aftereffect and silencing by motion. The models reproduced the effects seen in humans and generated surprising and novel predictions that we validated through human psychophysical experiments. Inconsistencies between model predictions and experimental results were resolved through iterative model adjustments. For the shape-contingent color aftereffect, the model predicted and experiments confirmed weaker aftereffects for outline shape manipulations and individual differences in perceived aftereffects. For silencing by motion, the model predicted and experiments validated unexpected trends as well as individual differences. Our findings demonstrate AICON's ability to capture relevant aspects of human visual information processing including variability across individuals. It highlights the potential for novel collaborations between synthetic and biological disciplines.