Designing a mood-mediated multi-level reasoner

Gomes, Samuel; Rocha, José Bernardo; Dias, João; Martinho, Carlos

http://hdl.handle.net/10400.1/26157

Use this identifier to reference this record.

Name:	Description:	Size:	Format:
Designing_a_Mood-Mediated_Multi-Level_Reasoner.pdf		3.54 MB	Adobe PDF	Download

Send Feedback

Authors

Abstract(s)

Psychology-oriented research extensively studied how affect influences human decision making. Particularly, the cognitive tuning assumption suggests that mood can serve to regulate between shallow and deliberative decisions - a more negative mood means higher deliberation. This work proposes a model that mimics the cognitive tuning assumption. To that end, the Collective Risk Dilemma (CRD) game For the Planet was designed and created, a process allowing for distinct levels of reasoning was defined and tested in that game, and mood was integrated to control the level of reasoning of such a process. Several distinct Artificial Intelligence (AI) profiles were created to verify that the developed model was able to dynamically change the level of reasoning and adjust the resulting AI behavior in our CRD game. Results revealed that the distinct AI profiles were influenced by their affective states and experienced circumstances, and that the emergent behaviors were consistent with the cognitive tuning assumption, thus demonstrating that we managed to construct an innovative and flexible model that can use mood to dynamically adjust the level of reasoning of an AI agent.

Keywords

Mood Games Cognition Appraisal Decision making Tuning Computational modeling Affective computing Cognitive models Appraisal processes Mood or core affect