Adaptivity, Emotion, and Rationality in Human and Ambient Agent Models

A system which has the ability to take a flexible autonomous action in a dynamic, unpredictable and open environment is referred to as an agent. Agent technology has a great impact on all spheres of society. For instance, agent technology is being applied within the area of ambient intelligence, which has become one of the most dynamic, demanding and exciting domains of computational sciences nowadays. The ambient intelligence domain, aimed at having a ubiquitous computing environment with the focus of providing invisible support for humans, can play a vital role in future developments. The goal of this PhD thesis is to contribute to the domain of ambient intelligence by means of the development of ambient agents that are aware of the various characteristics of humans as well as their mental processes and states. Using this knowledge, ambient agents can provide highly personalized and dedicated support for humans, for instance to support them during decision making. In this PhD thesis three human aspects are considered, namely adaptivity, emotions, and rationality. To come to ambient agents that are able to understand these processes, models that represent the human processes have been developed and incorporated within ambient agents. These models are based on well recognized theories that are available in different disciplines, such as biology, neurology, psychology, and sociology. The goal of this PhD thesis is to determine a way in which such models of human processes can be developed, as well as to develop measures using which they can be evaluated. For this purpose, first, a variety of approaches is explored to design ambient agents equipped with the awareness of a human’s cognitive and affective states. The models designed and analyzed in this dissertation indeed show that ambient agents can be made aware of various characteristics of humans and their environment. In particular three levels have been explored: the cognitive level, the neurological level, and the level of interaction of the human with the environment. For the cognitive level, it has been shown how beliefs, desires, intentions, emotional states, and visual attention can be modeled. From the neurological perspective, models for mirroring, inverse mirroring and false attribution of self-generated (e.g., manual or verbal) actions to other agents have been developed. With respect to the interaction of the human with the environment, the ecological domain has been addressed, which covers the background for modeling an ambient agent that can reason about environmental dynamics. Secondly, to introduce measures through which the ‘human likeness’ of the proposed models can be assessed, a nontrivial question is addressed, namely whether the agent is behaving rationally in a changing environment or not. For this purpose, in this dissertation two rationality measures are introduced, which prove useful to evaluate the behavior of the proposed model.