Learning pragmatic frames in a computational architecture

Abstract

It still remains a mystery how social interactions can be learned só early in the human development process. Trying to understand better how social interactions work and how they emerge in infancy, Bruner (1983) presented the concept of a Pragmatic Frame (PF), a way to describe and study social interactions. This work addresses part of the mystery of learning a PF, proposing a taxonomy and a computational representation of Pfs, which are used in a solution for an artificial agent to learn a specific part of a PF through FAMA (Aineto, Jiménez Celorrio e Onaindia (2019), Aineto, Jiménez e Onaindia (2019)), an action model learning approach. This solution is implemented and tested with experiments that try to assess its viability.