Artificial Intelligence from the perspective of Information Science: where are we toward computational reasoning?

Abstract

Notable advances have been possible from the use of algorithms for automatic reasoning of Artificial Intelligence (AI). Such advances lead information scientists to question whether there is any sense in the traditional classification after the advent of search engines. The issue impacts not only CI, but Librarianship since the data of the 21st century are no longer on shelves. There is no physical restriction that requires the organization for which the classification schemes were originally designed. Some claim even if it is possible that a previous scheme predicts a user's need. This paper aims to discuss the limits of automatic reasoning and the extent to which it substitutes the human in classification tasks. The methodology adopted was: i) bibliographic review on essentials of human and computational reasoning; ii) the description of the automatic reasoning produced by two popular approaches: ontology and machine learning; iii) comparison between types of reasoning. Findings: we found that automatic approaches have clear limitations, but this is something well-known. The best contribution is to identify ontologies as closest to what people can do in the classification and, therefore, the closest to Information Science. Conclusions: AI assistance can be useful for humans, insofar as the analyzed approaches – classic automatic reasoning and probabilistic reasoning – are far from replacing human reasoning in classificatory tasks.