A Contrast-Based Computational Model of Surprise and Its Applications

Area of Science:

• Artificial Intelligence
• Cognitive Science
• Psychology

Background:

• Surprise is a fundamental cognitive process influencing attention and decision-making.
• Existing models often lack a computational framework for quantifying surprise intensity.
• Psychological theories suggest surprise involves appraisal, attention shift, and belief revision.

Purpose of the Study:

• To present a contrast-based computational model of surprise.
• To explore the cognitive processes underlying surprise.
• To demonstrate the applications of computational surprise in artificial intelligence.

Main Methods:

• Developed a computational model where surprise intensity is a nonlinear function of probability contrast.
• Incorporated cognitive processes like appraisal, attention focusing, and belief revision.
• Applied the model to artificial agents in various domains.

Main Results:

• The model quantifies surprise based on the difference between subjective and alternative event probabilities.
• Surprise intensity is shown to be a nonlinear function of this probability contrast.
• Applications demonstrate surprise's role in exploration, creativity, and intelligent systems.

Conclusions:

• The contrast-based computational model provides a framework for understanding and implementing surprise.
• Surprise is crucial for decision-making, active learning, and selective attention in artificial agents.
• The model has broad implications for artificial intelligence, cognitive science, and psychology.