Contextual Deliberation of Cognitive Agents in Defeasible Logic

This article extends Defeasible Logic to deal with the contextual deliberation process of cognitive agents. First, we introduce meta-rules to reason with rules. Meta-rules are rules that have as a consequent rules for motivational components, such as obligations, intentions and desires. In other words, they include nested rules. Second, we introduce explicit preferences among rules. They deal with complex structures where nested rules can be involved. The basic deliberative process uses rules to derive goals (desires, intentions, obligations) based on existing beliefs, desires, intentions and obligations (beliefs concern the knowledge an agent has about the world: they are not in themselves motivations for action). Contextualising the deliberation requires to provide the agent with a mechanism for reasoning with rules, which are conditioned to some additional factors. In the simplest case, this can be done by adding such factors as new antecedents of the rules to be contextualised. But transformations may be problematic when complex reasoning patterns are considered. We therefore extend Defeasible Logic to deal with the contextual deliberation process of cognitive agents. First, we introduce meta-rules to reason with rules. Meta-rules are rules that have, as a consequent, rules to derive goals (obligations, intentions and desires): in other words, meta-rules include nested rules. Second, we introduce explicit preferences among rules to capture complex structures where nested rules can be involved in scenarios where rules are violated. The main challenge in the formal definition is to introduce a notion of compatibility between rules, for which we have given one possible solution in this paper, based on a subtle introduction of negated rules.