Mind as a Service: Building Socially Intelligent Agents
The ability to exhibit social behaviour is paramount for agents to be able to engage in meaningful interaction with people. In fact, agents are social beings at the core. That is, agent behaviour is the result of more than just rational, goal-oriented del
- PDF / 360,109 Bytes
- 15 Pages / 439.37 x 666.142 pts Page_size
- 60 Downloads / 182 Views
Abstract. The ability to exhibit social behaviour is paramount for agents to be able to engage in meaningful interaction with people. In fact, agents are social beings at the core. That is, agent behaviour is the result of more than just rational, goal-oriented deliberation. This requires novel agent architectures that start from and integrate different socio-cognitive elements such as emotions, social norms and personality. Current agent architectures however, do not support the construction of social agents in a structured, modular and computational- and designefficient manner. Inspired by service-orientation concepts, in this paper we propose MaaS (Mind as a Service) as a modular architecture for agent systems that enables the composition of different socio-cognitive capabilities into a running system. Depending on the characteristics of the domain, agent’s deliberation will require different social capabilities. We propose to model these capabilities as services, and define a ‘Deliberation Bus’ that enables to design deliberation as a composition of services. This approach allows to define deliberation architectures that are situational and dependent on the available components in order to cope with the complexity of social and physical environments in parallel. We furthermore propose a Service Interface Descriptor language to encapsulate service functionalities in a uniform way.
1
Introduction
The potential of artificial intelligent systems to interact and collaborate not only with each other but also with human users is no longer science fiction. Healthcare robots, intelligent vehicles, virtual coaches and serious games are currently being developed that exhibit social behaviour - to facilitate interactions, to enhance decision making, to improve learning and skill training, to facilitate negotiations and to generate insights about a domain. In all these cases, the ability to exhibit social behaviour is paramount for successful functioning of the system. We informally define social intelligent agents as systems whose behaviour can be interpreted by others as that of perceiving, thinking, moral, intentional, and behaving individuals; i.e. as individuals that can consider the intentional or rational meaning of expressions of others, and that can form expectations about the acts and actions of others [27]. In this light, functionalities required from social intelligent agents include the ability to reason about norms, beliefs and culture-specific c Springer International Publishing Switzerland 2016 V. Dignum et al. (Eds.): COIN 2015, LNAI 9628, pp. 119–133, 2016. DOI: 10.1007/978-3-319-42691-4 7
120
V. Dignum
contexts, to display and understand emotions, to balance between goal-directed and reactive behaviour, maintain a sense of identity, to form expectations about the other’s acts and actions, etc. An important aspect of social behaviour is the capability to integrate and to choose between different types of behaviour, such as e.g. utility-based, mimicry or altruistic behaviours based on the physical and social c
Data Loading...
