|
|
|
| Rethinking on Attention: A Reinforcement-learning
Model of Attention |
| CHENG Shaozhe, SHI Bohao, ZHAO Yang, XU Haokui, TANG Ning, GAO
Tao, ZHOU Jifan *, SHEN Mowei? |
| Department of Psychology and Behavioral Science, Hangzhou, 310028 |
|
|
|
|
Abstract Though summarizing and analyzing the current attention theories, the present
article proposes a new hypothesis that attention is a phenomenon of information selection,
rather than a mental architecture or cognitive resource. Inspired by the reinforcementlearning
algorithm in artificial intelligence field, we suggest a reinforcement-learning model
of human behavior that is able to show the phenomenal attention. This model describes the
interaction between agent and environment: the agent takes action to interact with the
environment and get feedbacks, by which the mental state updates to produce a new policy
for taking the next-step action, in order to maximize the cumulative reward. In this learning
procedure, attention emerges as a phenomenon that high-value information gradually get
processing priority. This framework of modeling provides a new approach to rethink the
nature of attention.
|
|
Published: 09 January 2017
|
|
|
|
|
|
|
