Evaluation of an Affective Model : COR-E Sabrina Campano, Etienne de Sevin, Vincent Corruble, Nicolas Sabouret Laboratoire d’Informatique de Paris 6 4, place Jussieu, 75005 Paris, France {name.surname}@lip6.fr

Abstract. In this paper, we present an evaluation of the affective model COR-E. This model intends to produce behaviors judged as emotional and believable ones. Emotions are seen as an emergent phenomenon, they are not encoded in the model. Our results show that COR-E effectively produce intended behaviors, thanks to its various characteristics. Keywords: affect, emotion, believability, behavior, virtual agent

1

Introduction

Most existing computational affective models rely on a number of numerical emotion variables that must be manually parametrized so as to outline believable affective responses and behaviors [5, 3, 2]. However, finding the correct value of these parameters and the influence of each one on the general model is a significant challenge. Other approaches, such as Pfeifer’s work [6] or the MicroPsi model [1], aim at obtaining emotional behaviors without using emotion variables. Emotions are considered as an emergent phenomenon. The model COR-E presented in this paper enters this category of model. COR-E (COR-Engine) is based on the psychological theory of COnservation of Resources (COR) [4]. The central tenet of the theory is that people strive to obtain, retain, and protect resources. The concept of resource refers to many types of subjective items : social ones such as self-esteem or caring for others, material ones such as a car, or physiological ones such as energy. COR-E intends to keep an architecture with a small number of parameters, while allowing the simulation of a wide variety of affective behaviors, including social ones. COR-E is based on the principle that an agent tries to protect and acquire resources that it values when they are respectively threatened or desired. The general architecture of the model is shown in figure 1.

2

Evaluation

In order to evaluate COR-E, we recorded videos clips of agents simulated by the model, in the scenario of a waiting line. Then we asked some human participants to answer an online questionnaire about these videos. The evaluation had two main objectives: (i) to determine whether agents’ behaviors simulated by

2

Evaluation of an Affective Model : COR-E Resource Sets

Environment

Threatened Resources

protect

Desired Resources

acquire

Possible Behaviours

Preferences

Behaviour Selection

Fig. 1. General Architecture of COR-E - Resources Sets determine possible behaviours ; behaviour selection depends on possible behaviours and agent’s preferences ; environment is updated with behaviour’s effects ; resource sets are updated according to the environment.

COR-E are considered as believable and emotional by human observers (general hypothesis H1); (ii) to validate the impact of the main characteristics of CORE’s architecture: acquisitive and protective behaviors (H2), preferences (H3), and the use of reputation psychological resources (H4). 113 participants contributed to this study. According to our results, the hypotheses about COR-E were all validated. COR-E allows the simulation of believable emotional behaviors (general hypothesis H1) thanks to its characteristics (H2, H3, H4). Acquisitive and protective behaviors, preferences, and the psychological resource of “Reputation” seem necessary to produce such behaviors. A large majority of participants recognized the behaviors produced by CORE as related to agents’ emotions (71.68 % to 92.04 %). These good results may be due in part to the use of the textual indication “protest”, used by an agent in order to react to an intrusion in the waiting line. Indeed, this term can be psychologically associated with the emotion of anger, thus facilitating the recognition of that emotion among participants. It would be interesting to know if the same results will be obtained without textual indications. Behaviors produced by COR-E model were rated as believable (mean from 5.08 to 6.01 on a scale from 1 to 7). It is possible that the score on believability was lowered because of a bias related to the interpretation of the term “believable”. As emotional behaviors tend to occur rarely, they might be judged as less believable. Another possible bias that could have lowered the score on believability may be related to agents’ moves. These elements indicate that the believability of agents’ behaviors could be further improved.

3

Conclusions and Future Work

We presented in this paper an evaluation aimed at assessing whether the behaviors produced by COR-E are judged as believable and emotional. In a further evaluation, we plan to assess whether human observers recognize appropriate emotional states in agents according to a given context, and also whether they recognize agents’ intentions. This will allow to check whether the internal state of an agent is well understood, according to the observation of its current behavior in the environment. This is an essential factor in agent believability [7].

Evaluation of an Affective Model : COR-E

3

References 1. D. D¨ orner, J. Gerdes, M. Mayer, and S. Misra. A simulation of cognitive and emotional effects of overcrowding. In Proceedings of the Seventh International Conference on Cognitive Modeling, pages 92–98, 2006. 2. C.D. Elliott. The Affective Reasoner: A process model of emotions in a multi-agent system. 1992. 3. P. Gebhard. ALMA: a layered model of affect. In Proceedings of the fourth international joint conference on Autonomous agents and multiagent systems, pages 29–36. ACM, 2005. 4. S.E. Hobfoll. Conservation of resources. American Psychologist, 44(3):513–524, 1989. 5. S.C. Marsella and J. Gratch. EMA: A process model of appraisal dynamics. Cognitive Systems Research, 10(1):70–90, 2009. 6. R. Pfeifer. The“Fungus Eater”Approach to Emotion: A View from Artificial Intelligence. Cognitive Studies, 1:42–57, 1994. 7. P. Sengers. Designing comprehensible agents. In International Joint Conference on Artificial Intelligence, volume 16, pages 1227–1232. LAWRENCE ERLBAUM ASSOCIATES LTD, 1999.