Determining a Framework for the Generation and Evaluation of Ambient Intelligent Agent System Designs

The design and realization of Ambient Intelligence (AmI) systems using Artificially Intelligent (AI) agents is a rising field of research. However, the absence of clearly defined working criteria, supporting the generation and evaluation of AmI agent system designs, is a conspicuous obstacle to their advancement. The contribution of this paper is that we determine and test a framework for the generation and evaluation of AmI system designs, based on user experience and business criteria. Specifically, the process of designing a personal lighting AI agent, in collaboration with a leading lighting design company, is used as a case study to determine and test a framework for the generation and evaluation of AmI system designs based on feasibility and acceptability. First, we use storytelling videos to describe and communicate the user values and design scenarios to the stakeholders. Second, we generate design proposals for a lighting AmI agent based on five distinct systemic factors, namely: (a) the context of interaction; (b) the required system data; (c) the required sensing input; (d) the required user input; and (e) the desired system output. Finally third we determine an evaluation framework that is based on three distinct levels of in-built system intelligence, from lower to higher. The three levels reflect the feasibility and acceptability of the system. Feasibility is what a specific company is capable of producing, and in what timeframe. Acceptability is the potential of familiarity and trust that the users can feel while interacting with the AI agent.