Caring about the user’s view : the joys and sorrows of experiments with people

Abstract

Deploying AI, and specifically P&S, technology into the real world entails many stimulating problems for researchers and system designers. Various smart solutions have been produced both for specialized problems and, in fewer cases, for more general purpose domains. An important aspect which is fundamental for any successful application, and more specifically for those which address a broad audience, concerns the users' perception and acceptance of technology. This issue is often either neglected or underestimated. We have been working for two years on the issue of importing experimental techniques from HCI and experimental psychology into smart system development. This approach has both pros - interesting features from the user perspective are discovered and can be used to bias design and research activities - and cons - experimenting with humans adds additional difficulty to the project, and applying a correct methodology is very time consuming. This paper describes a fielded experimental investigation of a fully implemented AI system named ROBO CARE . The system uses constraint-based scheduling technology to actively monitor a pattern of activities executed by an assisted person and uses detected temporal constraint violations to trigger meaningful and contextualized proactive interaction. Dialogue with the users is managed by a robotic mediator who acts as the main communication channel between the users and the intelligent domestic environment. The paper presents an evaluation of elderly people's perception of the intelligent system, focusing on aspects related to the robot's aspect, interaction modalities, content and timing of interaction, providing suggestions and hints for system designers.