This repository corresponds to the article's experiment "The Big-2/ROSe Model of Online Personality -- Towards a Lightweight Set of Markers for Characterizing the Behavior of Social Platform Denizens" (under review)
Here is the paper abstract: Background/Introduction: The Big-5/OCE-AN personality traits model, one of the central approaches to psychometrics, has been shown to have many applications over a variety of disciplines. In particular, correlations have been studied leading to effective characterization of people's behavior, and has become notorious for its role in the Cambridge Analytica/Facebook scandal surrounding the 2016 US presidential elections. Methods: In this paper, we develop Big-2 (or ROSe, for Relationship to Others and to Self), a model via which the personality of users of online platforms can be studied using a lightweight set of markers focused on on-line behavior, avoiding the major data privacy pitfalls afflicting approaches based on more powerful models that characterize personal aspects of the human psyche. Evaluation of Big-2's effectiveness is done in two parts: a quantitative evaluation on a specific prediction task and a qualitative evaluation based on an analysis of the different ways in which the Big-2 traits can be derived from online behavior, proposing a general template to guide such efforts. Results: Quantitative results show that our lightweight model can match or surpass the performance of Big-5 in a prediction task. Qualitative results show that it is feasible to implement the model based on the observation of basic online user behavior. Our main result is a general-purpose model that canbe used to characterize the personality traits of users of online platforms in an ethical manner. Conclusions: Our proposed model provides a valuable tool to carry out effective and explainable analyses of online personality, avoiding the collection of unnecessary user data that would open the possibility for ethical violations.