Temporal Agents

The finite bounds on the rate and duration of physical processes is a natural resource limit for a rational agent, but processes are a not prominent in logical presentations of rationality. Here we consider some of the logical issues in portraying processes in a way which may help explicate linguistic and computational aspects of human agents, assist the design of artificial agents, and perhaps too, explain some mistakes of reason. 1. Appreciating Psychologism By embracing psychologism in their “New Logic”, Gabbay and Woods admit inference which is fallacious in traditional logic. This is justified by arguments such as the usefulness of short cuts for achieving effective response from an agent with limited computational time and space. The cited paper stands as an informal prelude to a logic of wider ambit, including abduction and discovery, an unconventional, ground-breaking logic whose consequences are to be induced from inference steps which include the seemingly fallacious. Although inspired by the new logic, this paper is more concerned with processes and their logical presentation; the computational processes of a bounded agent in an unbounded environment, with processes to be composed from primitive steps, and where the way processes are composed ultimately constitutes the mental architecture of the agent. If these processes are to operationalise the inferences of a more conventional logic, its consequences will perhaps be perceived as epistemic states of the agent. We take the view that in modelling an agent the processes are no more ephemoral than the states they may produce, that the observational behaviour of an agent depends on rich, dynamic, and communicative ingredients that stative concepts alone do not capture, and that the reasoning processes of agent, through being temporally bounded, can be distinguished from the atemporal stative inference which may be proper for an institutional agent. From this perspective, psychologism may be the misplaced adoption of temporal reasoning processes when not appropriate. Despite the encapsulated rationality of traditional mathematics and logic, and the hunger of computational theorists for guidance, these disciplines have not been overly successful in characterising real computation. One can argue that a faithful description of computational engines, and of operational agents, depends on a temporal sense of causality and consequence which can be hidden by atemporal interpretation when rationality is expressed in traditional logics. The discipline of symbolic logic may come as inherited background in the drive to represent computational processes, and it has been elegantly exploited by causal rules in logic programming styles, but all too often the underlying logic of a computer program is obscured by cumbersome notation with explicit temporal order, whether Turing machine, 1Dov Gabbay and John Woods, The New Logic, Logic Journal of the IGPL, vol.9 No. 2, pp157-190, 2001