Towards robots with common sense

Winfield, Alan

Abstract

This study explores the question: could a robot be equipped with an artificial “common sense” and if so, how? The study outlines a method for equipping social robots with an artificial theory of mind. The method is based on a simulation-based internal model that allows a robot to “imagine” and hence predict the likely consequences of both its own actions and others in its environment. An approach that we have experimentally tested with real robots, demonstrating both enhanced safety and simple ethical behaviours. The chapter describes these experiments – alongside current work to extend the method so that robots are able to tell each other stories – and argues that our approach offers a practical starting point for social robots with artificial common sense.

• Online Publication Date: Jun 28, 2024
• Publication Date: Jun 28, 2024
• Deposit Date: Jun 18, 2024
• Pages: 66-81
• Book Title: AI and Common Sense: Ambitions and Frictions
• Chapter Number: 5
• Public URL: https://uwe-repository.worktribe.com/output/12066791
• Publisher URL: https://www.taylorfrancis.com/chapters/edit/10.4324/9781032626192-9/towards-robots-common-sense-alan-winfield