40 likes | 167 Vues
This paper explores the Affective Agent Architecture developed by Kevin Raison and Steven Lytinen, which integrates Tompkin’s theory of affect to create decision-making systems that respond to emotional states. The framework includes drives like hunger and bonding, as well as positive and negative affects, significantly influencing agent behavior in modified scenarios, such as a Wumpus world. The study highlights how these emotional inputs lead to appropriate actions, offering insights for advancements in artificial intelligence and robotics that require nuanced interaction with their environments.
E N D
An Affective Agent Architecture Kevin Raison Chatsubo Labs, Seattle Steven Lytinen DePaul University, Chicago Click Insert > Header & Footer to add Area/Division/Department name.
Tompkin’s (1962) theory of affect • Drives: Hunger, elimination, bonding, … • Positive affects: **Interest, Joy • Negative affects: **Fear, distress, anger, disgust, shame • Neutral: Surprise In short, drives + inputs -> affects -> actions Click Insert > Header & Footer to add Area/Division/Department name.
Modified wumpus world • Food replaces gold • No shooting • Agent never leaves • Senses dissipate • Drives: hunger, • elimination • Input: smells, breeze • Affects: interest, fear • Actions: move toward, • turn and run Click Insert > Header & Footer to add Area/Division/Department name.
Reactive Planning +F2 –S1 -B1 local maximum until hungry -B2 +F1 -S2 +F1 -S1 – B1 -B2 +F1 Click Insert > Header & Footer to add Area/Division/Department name.