Opinion Types on Social Media: A Review of Approaches to What Opinions Are in Social vs. Computational Science

Today’s public opinion research has rapidly transformed to include opinions expressed within new forms of mediatized communication, including those on social networks and messengers. Computational communication studies have provided for new approaches in opinion mining and detection, and many new approaches to how ‘opinions’ are seen in research have emerged. The growth of the academic field that (re)conceptualizes opinion as a term, including for automated and AI-based opinion detection, has gradually turned from a spurring advantage to an inhibitory obstacle for further public opinion studies. The diversity of this field, being a problem itself, also produces scientific sub-problems that we try to address. The three of them are: (1) the widening differences between the ‘traditional’ (pre-computational) opinion studies, already highly diverse, and computational opinion mining; (2) the shared methodological basis of the computational communication studies that mostly reduce ‘opinion’ to the lexical-semantic levels of speech, largely ignoring social-group, political, cultural, temporal, and narrative aspects of opinion formation; (3) our unawareness of whether the gap between human and AI-created human-like representation of opinions can ever be overcome, and, if not, how the machine sees opinions and why they cannot reach the logic of human judgment on opinionated content. In this paper, we review the three periods of formation of opinion studies, which have brought on the ‘traditional’ (very diverse), computational, and human-like computational views on how opinions look like in academic representation. Our main focus is on how the pre-computational and computational methods of opinion detection could enrich each other and whether full similarity between opinion formulation and/or representation by artificial intelligence and human beings could ever be reached. In the conclusion, we hint on the future directions of conceptual research on the nature and essence of human and human-like opinion.