SOCIAL SCIENCES, EMOTIONS AND ALGORITHMS

This paper aims to provide a framework for questioning and rethinking the relationship between theories and techniques which, in social sciences, allow us to study emotional phenomena through computational text analysis. While commenting on the drawbacks of a culture that takes for granted both the theories of emotions and the techniques for data analysis, a sort of reflective approach will be proposed which refers to linguistic and hermeneutic methods. Section (1) poses some questions concerning a specific technique, namely the sentiment analysis, and argues that, in this case, the phenomena under examination are related with the metaphorical way of thinking and communicating through which we humans construct both our social relationships and our cultural models. Section (2) focuses on theories concerning emotional phenomena and proposes a model for interpreting their reciprocal relationships; as a result, emotions expressed through texts are defined as cultural phenomena inherently context dependent. Section (3) discusses the use of certain tools – namely word embedding and lexicons for emotional text analysis – within the framework of a theory of technique which refers to a sort of hermeneutical paradigm which strictly connects the interpretation work with the specific social contexts where the emotional phenomena originate.