The emergence of (artificial) emotions from cognitive and neurological processes
We address the BICA Challenge with a multi-agent system implementing Stanovich’s Tripartite Framework, which we have augmented with a diffuse control system modeled on biological neuromodulations. The Tripartite Framework shows how adaptive yet reflective behavior can emerge from the interaction of three sets of processes: processes responsible for fast context-sensitive behaviors (an autonomous mind), processes responsible for cognitive control (an algorithmic mind), and processes responsible for deliberative processing and rational behavior (a reflective mind). Working within this augmented Tripartite Framework, we were able to build a fully situated, goal-directed sensorimotor agent that can plan its behavior by reasoning on counterfactual situations. In this paper, we put neuromodulations to work towards giving the system well fleshed out emotions. Without them, the system’s emotions are purely semantic and cognitive. They are semantic in that the system’s conceptual map contains emotion words that are fully linked to other relevant words, and they are cognitive in that algorithmic-level control can focus goal-directed attention on the emotion words when the task demands it (as in an emotional Stroop task). With neuromodulations, we believe we can integrate the main physiological component of Lindquist’s situated conceptualization of emotions, core affect, understood physiologically as dynamical patterns of neuromodulations. Emotions in the resulting system are patterns of message passing activity between agents in which neuromodulations can increase sensitivity on salient emotional aspects of environments and focus attention on those aspects. We study the resulting emotions with the help of an emotional Stroop task in which the semantic and cognitive aspects of emotion are observed.