Logical Modeling of Emotions for Ambient Intelligence

Logical Modeling of Emotions for Ambient Intelligence

Carole Adam (RMIT University, Australia), Benoit Gaudou (UMI 209 UMMISCO, IRD, IFI, Vietnam), Dominique Login (University of Toulouse, CNRS, IRIT, France) and Emiliano Lorini (University of Toulouse, CNRS, IRIT, France)
DOI: 10.4018/978-1-61692-857-5.ch007


Ambient Intelligence (AmI) is the art of designing intelligent and user-focused environments. It is thus of great importance to take human factors into account. In this chapter we especially focus on emotions, that have been proved to be essential in human reasoning and interaction. To this end, we assume that we can take advantage of the results obtained in Artificial Intelligence about the formal modeling of emotions. This chapter specifically aims at showing the interest of logic as a tool to design agents endowed with emotional abilities useful for Ambient Intelligence applications. In particular, we show that modal logics allow the representation of the mental attitudes involved in emotions such as beliefs, goals or ideals. Moreover, we illustrate how modal logics can be used to represent complex emotions (also called self-conscious emotions) involving elaborated forms of reasoning, such as self-attribution of responsibility and counterfactual reasoning. Examples of complex emotions are regret and guilt. We illustrate our logical approach by formalizing some case studies concerning an intelligent house taking care of its inhabitants.
Chapter Preview


Ambient Intelligence (AmI) is the art of designing intelligent and user-focused environments, i.e. environments that can adapt their behavior to users and to their specific goals or needs at every moment, in order to insure their well-being in a non-intrusive and nearly invisible way. AmI systems should be embedded, context aware, personalized and adaptive (Jiao, Xu and Du, 2007). AmI is a highly multidisciplinary field of research, at the convergence between as various disciplines as for instance electronics, networks, ergonomics, robotics or computer science. Some examples of applications are ISTAG’s research scenarios (Ducatel et al., 2001), industrial ones such as Philips intelligent house HomeLab (Aarts and Eggen, 2002), or academic projects such as MIT’s Oxygen project and their MediaLab.

Since the focus of AmI is on the human users, it is of great importance in order to better understand them to have a model of human factors and to take these factors into account (Treuil, Drogoul and Zucker, 2008). In this chapter we focus on a particular human factor, namely emotions, that have been proven to be essential in the human decision making process (Damasio, 1994), but also in the interaction between users and machines (Picard, 1997). Therefore we think that AmI can take advantage of the results obtained in Affective Computing (Picard, 1997) or Artificial Intelligence (AI) about emotions.

In particular, the aim of this chapter is to show how logical tools and methods, traditionally used in the field of AI for the formal design of intelligent systems and agents, can be exploited in order to specify emotional agents for AmI applications. The term emotional agents refers to intelligent agents endowed with some emotional abilities, e.g. identifying and reacting to the users’ emotions. These agents may also be able to “feel” emotions1, to express them, and to behave accordingly. We think that there are two main reasons justifying the use of logical methods for the design of such emotional agents.

First, although physiological sensors can detect some simple users’ emotions such as happiness (Prendinger and Ishizuka, 2001), they are neither able to determine the object of these emotions (e.g. happiness about the sun shining or about being in holidays), nor able to differentiate between some close emotions such as sadness, regret and guilt (Oehme et al., 2007). Indeed, to do this, it is necessary to take into account the context and the users’ mental attitudes such as beliefs, desires, ideals which are the cognitive constituents of emotions. Now, so-called BDI (Belief, Desire, Intention) logics developed in the field of AI in the last fifteen years (see e.g. Cohen and Levesque, 1990; Rao and Georgeff, 1991; Lorini and Herzig, 2008) offer expressive frameworks to represent agents’ mental attitudes and to reconstruct on their basis the cognitive layer of emotions (Adam, 2007; Castelfranchi and Lorini, 2003; Steunebrink, Dastani and Meyer, 2007).

Second, in order to adapt to the users, an AmI system must be able to reason about their emotions and to analyze them. In particular, an AmI system must have a model of users’ emotions and of the ways in which they affect their behavior, like action tendencies (Frijda, 1986) (e.g. feeling fear entails escape) and coping strategies (Lazarus, 1991) (e.g. denying an information causing too much sadness). This model would allow the AmI system to reason about (e.g. anticipate) the effects of its own actions on the users’ emotions and actions. Now, logical approaches were specifically designed to endow agents with this kind of reasoning capabilities.

Key Terms in this Chapter

Counterfactual Reasoning: Reasoning based on what the current state of the world would be or would not be if I had acted differently, i.e. if I had performed or had not performed a certain action. For example, if I had slept this morning, I would be less tired; if I had not run this morning, I would feel sad; if I had taken the train instead of my car, I would not be caught in a traffic jam; if I had bought petrol earlier, I would not have run out of petrol on the freeway.

BDI Logic: Formal language used to represent agents’ reasoning in terms of their mental attitudes (beliefs, desires, intentions) as described by philosophers of mind (e.g. Bratman).

Cognitive Appraisal Theories of Emotions: In these theories, emotions are triggered by a process (called appraisal) determining the significance of the situation for the individual by assessing it w.r.t. various criterions called appraisal variables (e.g. desirability).

Actions Tendencies: Processes associated to emotions that allow individuals to make quick decisions, using their emotions as a kind of heuristic. For example, fear induces a tendency to escape while anger induces a tendency to attack.

Responsibility: An agent i is responsible for causing p if and only if i has performed a certain action and if i had not performed this action then p would not be true now. For example a driver is responsible for causing an accident by running a red light because if he had not, then the accident would not have occurred.

Coping Strategies: Strategies (or actions sets) used by an agent to restore its well-being in stressful situations. For example one can deny a piece of information that makes him too sad; one often starts by denying the death of a loved one.

Agent: Autonomous entity that is able to perceive its environment, reason about it, maintain personal goals, plan appropriate ways to achieve them, and act accordingly to modify in turn its environment. Some agents are also able to communicate with other artificial agents or with humans, or to express and understand emotions.

Complex Emotions: Emotions based on sophisticated forms of reasoning such as reasoning about responsibility and counterfactual reasoning. They are thus based on a larger set of appraisal variables than the set involved in basic emotions.

Emotion: Intentional affective mental state (in the sense of Searle, i.e. having an object) that arises automatically as a result of an individual’s interpretation of his environment and its relation to his attitudes (e.g. goals and interests). Emotions can then influence behavior in various ways, and were thus long considered irrational before research showed that they were actually essential in rational decision making.

Mood: Non-intentional affective mental state (in the sense of Searle, i.e. having no object). For example, one can be in a happy mood (about nothing in particular) which is different from feeling the emotion of happiness (about something, for example that the weather is sunny). Moods are also characterized by a longer duration than emotions.

Appraisal Variables: Criteria used to evaluate the significance of a situation w.r.t. an individual’s goals, ideals or desires, determining the triggered emotion. For example a pleasant situation (the pleasantness variable is measured by comparing the situation with the individual’s goals) leads to a positive emotion (e.g. joy).

Complete Chapter List

Search this Book: