CollabPEAS: A Collaborative VR System for Learning AI Agent Design Through the PEAS Framework

Abstract

We present CollabPEAS, a collaborative system for artificial intelligence (AI) literacy in which pairs of learners come to understand how an AI agent’s behavior is produced by human specification. Working in shared virtual reality (VR), learners jointly specify a virtual cleaning robot using the PEAS framework (Performance measure, Environment, Actuators, Sensors), developing competencies for reasoning about how AI agents sense, decide, and act. The system integrates three coordinated design elements: (1) a five-step PEAS-guided design flow, implemented in VR, that makes agent specification choices spatially grounded and observable, (2) a block-based coding interface that preserves the perceive, decide, act reasoning cycle, and (3) collaboration supports designed to structure negotiation and shared ownership between partners. We situate CollabPEAS within AI literacy and collaborative learning research and discuss directions for future evaluation.