Artificial agents and their ontological status
Abstract
Can an artificial agent "really" think? can it be "really" intelligent ? can it "really" have beliefs, goals, intentionality? Such questions have been plaguing AI since its inception. The solution we propose is to assert that an "artificial agent" is a virtual agent - and that all the cognitive or mentalistic attributes we may be tempted to grant it are also virtual. Whereas the first qualifier is purely descriptive, the second one is about the ontological status of such things as appear to us as agents. Our solution relies on: 1°) eliciting a precise core meaning for the word "virtual", a meaning that differs radically from the one used by philosophers (Bergson, Deleuze,...) but that has always been implicit in specialized contexts ("virtual image", "virtual world") and that should now be given its full ontological generality; 2°) relating the virtual to a broadened notion of interoperability, which justifies our assertion on technical and psychological instead of philosophical grounds. We relate this interpretation of AI to strong AI, to weak AI and to Dennett's intentional stance. We stress its implications for the cognitive sciences project of "naturalizing" intentionality. Finally, we mention some consequences in the information systems domain, relative to the acceptance of agent concepts for the modelling of organizations and their business processes.
Fichier principal
Artificial_Agents_and_their_Ontological_Status.pdf (181.65 Ko)
Télécharger le fichier
Origin : Files produced by the author(s)
Loading...