Toward Designing Ethically Acceptable AI Security Systems Through Agent Modeling

Abstract

AI systems are often complex and they have ethical consequences. We propose careful modeling of the related agents and their ethical concerns for the design of acceptable AI systems. We have worked on an example case of shopping mall security, and we will exemplify our ideas with our case study. Our modeling is based on BDI (belief-desire-intention) theory, philosophical theories of social action, and cognitive research on individual minds and ethics. For our empirical part, we have met and interviewed personnel in a real shopping mall. We aim to give a theoretical basis for considering ethical aspects for the use and design of AI in security.