Computing with speculations

Abstract

A novel, alternative approach to the fundamental logic of commonsense reasoning has been recently advanced by Enric Trillas, based on the idea that in order to allow for creativity in automated reasoning, the standard operations of inducing and abducing must be complemented by guessing, or speculating. In this paper, after a recall of Trillas’ skeleton formal model, a number of interactive examples that visualize some of the concepts involved in Computing with Speculations are presented, followed by a discussion and some proposals on how to overcome some technical hurdles in implementing speculations in a real computational context. Such examples want to be a contribution towards an experimental inquiry, whose concrete implementation ought to be based on the Computational Learning approaches described. The general aim is to introduce, albeit in a prototypical form, a theoretical framework for Natural Reasoning, and the tools to evolve such a framework, in combination with other similar endeavours, into a full system of Natural Computing.