A computational model of consciousness for artificial emotional agents

A computational model of consciousness is explored by designing a software agent that simulates speech understanding and irony. The agent operates with scripts—productions or inferences—to process texts, extract semantic representations, and select reactions. It can simulate irony by replacing direct aggressive behavior with a positive sarcastic utterance, balancing between available scripts. This scheme may serve as a minimal architecture of consciousness, where the agent distinguishes its own representations from those of other agents through a two-stage processing: activating scripts and differentiating them by activation levels. The agent demonstrates features considered essential for conscious agents in philosophy and cognitive psychology, such as simulating imaginary situations and actions of itself and others.

Study at a glance

Abstract

Background. The structure of consciousness has long been a cornerstone problem in the cognitive sciences. Recently it took on applied significance in the design of computer agents and mobile robots. This problem can thus be examined from perspectives of philosophy, neuropsychology, and computer modeling. objective. In the present paper, we address the problem of the computational model of consciousness by designing computer agents aimed at simulating “speech understanding” and irony. Further, we look for a “minimal architecture” that is able to mimic the effects of consciousness in computing systems. method. For the base architecture, we used a software agent, which was programmed to operate with scripts (productions or inferences), to process incoming texts (or events) by extracting their semantic representations, and to select relevant reactions. Results. It is shown that the agent can simulate speech irony by replacing a direct aggressive behavior with a positive sarcastic utterance. This is achieved by balancing between several scripts available to the agent. We suggest that the extension of this scheme may serve as a minimal architecture of consciousness, wherein the agent distinguishes own representations and potential cognitive representations of other agents. Within this architecture, there are two stages of processing. First, the agent activates several scripts by placing their if-statements or actions (inferences) within a processing scope. Second, the agent differentiates the scripts depending on their activation by another script. This multilevel scheme allows the agent to simulate imaginary situations, one’s own imaginary actions, and imaginary actions of other agents, i.e. the agent demonstrates features considered essential for conscious agents in the philosophy of mind and cognitive psychology. conclusion. Our computer systems for understanding speech and simulation of irony can serve as a basis for further modeling of the effects of consciousness.