BSEF10A040-PA-05(02).rar_The Web

Introduction In this chapter a brief description about the motivation for this research, research goals, research methodology, and a high level description of the component based design of a human based agent model is provided. Here the design of human based agent is proposed so that research work performed in this dissertation can be presented in collective framework. Finally, this chapter concludes with an overview of the parts and chapters presented in this dissertation. Motivation This section describes the main motivation and some of the background information behind the research work conducted for this dissertation. Section 2.1 advocates that computer simulation as a well accepted scientific method to study the systems which are physical inaccessible or too dangerous or unacceptable to engage. Hence it is considered one of the most suitable tools to study the dynamics of human trust phenomena. The argument presented in section 2.1 is extended for social simulations and the modelling of cognitive agents in section 2.2 and 2.3 respectively. It is argued that for the study of complex social phenomena which bears human as one of the main component the concept of cognitive modelling of the trust of a human is the primary requirement. In section 2.4 the role of human trust is emphasised to simulate realistic human behaviour in social simulation systems. 2.1 Simulation as scientific paradigm Science is an endeavouring process to explore the world around us and collect the knowledge that can be verified (Popper, 2002). In this process of scientific discovery scientists go through a rigorous cycle which is usually called the scientific method. The scientific method deals with observation of natural phenomena, generation of hypotheses explaining it, and the design of experiments in order to refute or accept these hypotheses. The result of the scientific experimentation provides the basis for the further inquiries and discoveries in science. The advent of thought experiments (Brown, 1993) and later in the last century the computer has given birth to the phenomenon of simulation as one of the promising scientific experimentation techniques. Simulation, or more precisely computer simulation, is very handy as a scientific experimentation tool when the studied system is physical inaccessible or too dangerous or unacceptable to engage (Sokolowski and Banks, 2009). Simulation can also be used prior to real experiments to branch and bound alternative possibilities of the experimentation process thus reducing the time and the cost. Besides the role of computer simulation as scientific experimentation technique over the years, computer simulation has been well accepted as a tool for education and training, healthcare, strategic and military planning, investment banking, and the manufacturing industry. As described in (Axelrod, 1997) the purpose of simulation is very diverse, including prediction, performance, training, entertainment, education, proof and discovery. The research conducted for this dissertation deals with a human�s internal state of mind, namely trust, which is not directly physically accessible, only indirectly via the behaviour it induces. Therefore computer simulation is the most appropriate scientific technique to be used in this process. 2.2 Social simulation Recently simulation techniques have been used to investigate the science behind complex social phenomena (Schelling, 1969; Epstein and Axtell, 1996). Social scientists have applied principles of the well known system dynamic theory (cf. Forrester, 1968; Forrester, 1969; Forrester, 1971) and the more recently developed agent-based techniques for simulation (e.g., Epstein and Axtell, 1996; Axtell, 1999; Axelrod, 1997). Both simulation techniques have pros and cons when applied in different contexts. Population-based simulation usually is based on system dynamics applied to global (non-individual) variables of the system. As population-based simulation lacks micro-level details of the system and hence is often assumed not to be a true representation of the system. In agent-based simulation internal processes and behavior of each individual agent is modeled (for example, using cognitive modelling techniques) and then the agents are allowed to interact with each other under some interaction protocol to generate both individual and collective social level effects. Though the agent based simulation is considered more realistic, for larger numbers of agents it becomes intractable. Population-based and agent-based simulations both, claim to represent social phenomena in there own respect. In this dissertation both techniques are applied and compared for these pros and cons, particularly in the context of trust. 2.3 Modeling of a cognitive agent An agent can be modeled at different levels, for example, at a physiological or psychological level. A cognitive agent is an intelligent agent based on human cognitive processes. Design of cognitive agents primarily obtains its inspiration from theories of cognitive science which claims to represent the human mental processes and behavior. Modeling and simulation of cognitive agents is equally important for the social sciences a well as for cognitive science for verification and validation of their respective theories (Sun, 2006). Another practical application of cognitive agents is in the domains of human-oriented ambient intelligence or human-aware computing where supporting software agents can reason (based on an internally represented cognitive model of the human) about a person�s state of mind and support whenever and wherever needed (Treur, J., 2008, Sharpanskykh and Treur, 2010). To realize intelligent personalized support to a human user different models of human cognition are being developed and embedded into devices that can reason about human mental processes and provide timely support, tailored to the person and his or her state. Trust can be one of the states of the person, which is estimated by the ambient system. Such trust states can relate to options or tools to be used by the person, but equally well to the support providing ambient agent. To support the human in a personalized way the trust human has in various options that can potentially aid in performing the task is an important factor. For instance, the manual for a task might contain a lot of flaws, whereas another human that has experience with the task can immediately aid in a correct fashion. Here, the trust value for the manual will be much lower, and an ambient agent showing this manual will most likely be ignored. In demanding circumstances, for example on board of a Naval vessel, where critical tasks are performed, such information is crucial for personal assistant ambient agents to be useful and help improve the overall effectiveness of a mission. Hence, the ambient agent must hold models of human trust dynamics to reason about and predict the current state of the trust of the user and adapt its own behavior accordingly. As the human would trust on those systems more which could understand and support human keeping his/her humanly nature under consideration hence the models presented in this dissertation are a step towards achieving a trustable ambient agent for the human support. 2.4 Trust as a cognitive phenomena Study of the dynamics of a human�s trust states has been performed across many academic disciplines which include neuroscience (King-Casas et al., 2005; Zak et al., 2004, Kosfeld et al., 2005), psychology (Deutsch, 1962; Worchel, 1979), sociology (Fukuyama, 1995; Gambetta, 1988), political science (Barber, 1983; Hardin, 2002; Knight, 2001), economics (Arrow, 1974; McCabe and Smith, 2000; Seabright, 2004; Zak and Knack, 2001), management science (Sitkin et al. 1998; Kramer and Tyler, 1996; Mayer, Davis and Schoorman 1995), computer science (Marsh, 1994) and artificial intelligence (Jonker and Treur, 1999; Shneiderman, 2000; Sabater and Sierra, 2005; Cast