Practical Reasoning with Attention Mechanisms

One of the most interesting questions in the field of artificial intelligent agents is whether it is feasible to design agent architectures that merge the classical emphasis on practical reasoning, with advanced psychological capabilities that are crucial for action guidance. This paper proposes an agent architecture, named CATALINA (Cognitive AgenT prActicaL reasonINg Architecture), that augments Bratman’s classic BDI paradigm with attentional mechanisms, executive functions (Buehler 2022), and a global workspace (Baars 2005). The architecture is applied to a scenario illustrating how the collaboration of the Executive Functions and the Global Workspace enables the action-guiding behaviour of the agent.