A Modern Epistemological Reading of Agent Orientation

A Modern Epistemological Reading of Agent Orientation

Yves Wautelet (Université catholique de Louvain, Belgium), Christophe Schinckus (Facultés Universitaires St-Louis, Belgium) and Manuel Kolp (Université catholique de Louvain, Belgium)
DOI: 10.4018/978-1-60566-970-0.ch003
OnDemand PDF Download:


This article presents a modern epistemological validation of the process of agent oriented software development. Agent orientation has been widely presented in recent years as a novel modeling, design and programming paradigm for building systems using features such as openness, dynamics, sociality and intentionality. These will be put into perspective through a Lakatosian epistemological approach. The contribution of this article is to get researchers acquainted with the epistemological basis of the agent research domain and the context of the emergence of object and agent-orientation. This article advocates the use of the Lakatosian research programme concept as an epistemological basis for object and agent orientation. This is done on the basis of how these frameworks operationalize relevant theoretical concepts of the Kuhnian and Lakatosian theories.
Chapter Preview


Data and information becomes very rich if not overwhelming in many modern applications, such as real-time driving assistance systems that integrate information from the sensors installed at different locations of a vehicle, or military C3I systems that make critical decisions based on efficient command and intelligence processing in complex battlefield. The 911 tragedy asks for essential security applications such as event surveillance and detection, intelligence monitoring and analysis, intrusion and deception detection, as well as their implementations in terrorism tracking, analysis, and response. Within various research fields, knowledge discovery and integration algorithms can lead to efficient solutions to assist model integration and decision making in complex and distributed environment. With so much information in hand in such applications, a suitable strategy that must apply intelligent and active information fusion is the key enabling technique.

Information fusion/integration aims at combining multiple evidences, which normally contain overlapping (correlating) information relating to the underlying hypothesis due to the generating mechanism of these evidences, in order to gain better accuracy or higher confidence on information of interest (Hall & McMullen, 2004; Klein, 2004). Thus any algorithm that takes multiple inputs to evaluate a target output can be candidates for information fusion. The main challenges in information fusion include uncertainty and incompleteness in evidence, requirement on processing speed and cost, etc.

In this article, we set the knowledge discovery and integration task used in discussion and experimentation to be human subject modeling. Such a model integrates video and acoustic information of human subjects as well as other related contextual information, and infer about the internal status about the subject’s mental and emotional aspects. This internal status can provide information of significant safety and security interests. Moreover the model could be about not only individuals but also a crowd of people. Through certain active strategy, it generates in a timely manner important indicators such as anxiety and stability of an individual or a crowd, which are useful measures of danger level of a special event involving human participants. Based on the given information, further analysis could be done with the help of other information such as the interacting machinery like a car or a fighter aircraft, or identification of the subject and report about possible terrorist plot, leading to proactive procedures of warning and countermeasures. Therefore such a human subject model can in fact function as a core component of intelligent assistance in battlefields, or scene surveillance and analysis in homeland security settings, providing real-time support for efficient decision making in emergency response and control.

Machine learning, artificial intelligence and psychological models have been extensively used in the information fusion in user modeling or human mental state detection (Cohen et al, 2000; Duric et al, 2002; Horvitz et al, 1998; Hudlicka & McHeese, 2002; Jameson, 1996; Picard, 1997; Salvucci et al, 2001). Of them, Bayesian network (BN) models have been applied in many user modeling and information fusion applications because of the capability to handle uncertainty and the analogy to human reasoning process, exemplified in a set of studies of intelligent office software assistants in the Lumiere Project other studies by Microsoft (Horvitz et al, 1998; Horvitz & Paek, 1999, 2000) and studies in affective state detection (Ball & Breese, 2000; Conti, 2002). In addition there are a range of information fusion applications for Bayesian networks, such as security modeling for service and health enterprise (Li et al, 2009; Li & Chandra, 2008).

Complete Chapter List

Search this Book: