Untitled SOFTWARE AGENTS

Research with respect to software agents started in the late 1980s. Today, software agents are mainly understood as programs that are able to work independently (autonomous), are able to react to changes in their environment (reactive), are able to act proactively and can communicate with other software agents. In the early 1990s the term “software agents” was used quite broadly, covering different areas of research in software technology. To focus the broad use of the term Nwana (1996) suggested a typology for software agents that is still being used.  

It has been suggested to use software agents for many routine like works such as complex searches for information in libraries and network attached storages and processing of (digital) products in e-commerce (see for example [Jennings & Wooldridge (1996)]). Software agents on one hand need context awareness to understand e.g. the will of its user, on the other hand they need a certain type of “intelligence” to be able to make decisions. An important area of use for software agents today are simulations in science and computer games. In this context a number of highly specialised agents have been developed and are in use.

Software agents may play a major role in ambient intelligent environments to search autonomously for information, to evaluate them and to draw conclusions including adaptive decision making. 

Research in software agents is carried out in the private sector (mainly enterprises) and by public research institutions (mainly universities) world wide. In Europe coordination of stakeholders is supported by the EU in the context of the IST program (project AgentLink).

Affective Computing

In recent years there has been much diverse work which explores the use of computing in ways which involve human emotion. This area is commonly referred to as affective computing. This includes work on the use of emotions in human-computer interaction, Artificial Intelligence (AI) and agent architectures which are inspired by the mechanisms of emotion, the use of emotion in computer-mediated communication, the study of human emotion through computers and philosophical issues concerning, for example, the extent to which it is meaningful to talk about emotion in computational terms.

In psychophysiology there lies an assumption that all human behaviour, including perception, cognition, emotion, and action, has a physiological substrate. Thus, it may be possible to identify reliable physiological indicators of psychological states and personality. The idea that these can be ‘reliable’ comes from an understanding of the autonomic nervous system (ANS). Typically we are unaware of our ANS because it functions largely involuntarily, via sympathetic (giving us fight and flight responses) and parasympathetic (which for example allow resting and digesting) pathways. The ANS is thus ultimately responsible for involuntary physical effects associated with emotions such as anxiety, fear, anger, embarrassment, and joy, and the heightened mental focus associated with concentration and problem solving. If we are able to interpret the physical, and thus readily measurable, changes and states, then we should be able to deduce some of the underlying emotional states.

The physiological measures most often studied for relations to psychological state are electroencephalograms (EEG), skin conductance, heart rate, blood pressure, skin temperature, respiration, muscle tension, and eye movements (see [Lisetti & Nasoz, (2004)] for a review of research). Sensor technologies have been developed in all these cases, and indeed small wearable technologies, which incorporate them, are technically feasible. However, one of the most vigorously researched measures is currently that of EEG because it is considered a potentially richer source of information, and because changes are more immediate than in other physiological recordings, for example see [Berka et al. (2004)].