Attention, learning, and the value of information

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Oculomotor Control

Background:

• Selective attention research has not fully elucidated its fundamental nature and neural underpinnings.
• Existing models often focus on specific aspects without a unifying framework.

Purpose of the Study:

• To propose an integrated view of attention drawing from animal behavior and machine learning.
• To explore how this integrated perspective can advance the understanding of visual attention and eye movement control.
• To connect attention mechanisms to higher-level cognitive functions such as curiosity and internal model inference.

Main Methods:

• Literature synthesis integrating findings from animal studies and machine learning.
• Conceptual framework development proposing attention as a mechanism for active learning and uncertainty reduction.

Main Results:

• Attention can be conceptualized as a process for reducing uncertainty and facilitating active learning.
• This unified view offers new insights into the control of eye movements (oculomotor control).

Conclusions:

• Integrating diverse perspectives on attention significantly enhances our understanding of oculomotor control.
• This framework provides a valuable lens for investigating complex cognitive functions like curiosity, exploration, and internal world modeling.