Skip to main content

The Epistemic and Dynamic Aspects of Polarization.

Periodic Reporting for period 1 - EDAPOL (The Epistemic and Dynamic Aspects of Polarization.)

Reporting period: 2018-09-01 to 2020-08-31

Exchanging arguments with others plays a crucial role when forming opinions on many issues, e.g. the trustworthiness of a political candidate or the usefulness of an electronic device. Providing persuasive arguments and assessing new ones influences our decisions, very often, more crucially than isolated inquiry. However, this process has a number of shortcomings, many of which have been identified and studied by social scientists. One most relevant is group-induced attitude polarization. Group polarization is a collective phenomenon that typically occurs when an initial tendency of individual group members toward a given direction is enhanced following group discussion. For example, a group of members of the same party will get more convinced of their opinion after an internal debate. Detrimental echo chambers and radical divergence between different communities are easily generated by this dynamic. Although it is widely recognized, by experimental and simulative studies, that the exchange of persuasive pro and contra arguments influences polarization, the exact mechanisms by which this process unfolds are still unclear. In particular, it is crucial to assess the impact played by the strategic disclosure of arguments and their more or less biased assimilation by individuals. The main aim of the EDAPOL project is to develop a formal understanding of these factors, so to unveil the step-by-step structure of the process of information flow and opinion change that generates polarization in a group. To this end, the project combines formal methods from argumentation theory and dynamic epistemic logic. Understanding polarization on a rigorous basis meets an important societal challenge in the era of social media. Indeed, virtual online discussion witnesses a more pronounced tendency for groups to polarize towards opposite directions than face-to-face discussions, with a detrimental impact on the associated life of our society.
The first preliminary objective of the EDAPOL project is to model the argumentative knowledge base of individuals and the effect on their opinion of different policies of argument disclosure (more or less transparent) and of belief update (more or less credulous). The second objective is to assess how different combinations of such policies, if held by individuals in a small group of debate, determine polarization of opinions. The subsequent and final step of this research in EDAPOL is to set the ground for implementation of this framework in multi-agent simulations and experimental work in social sciences, by harmonizing it with existent models in the area.
The project EDAPOL was organized in three sequential phases. The initial phase consisted of a literature review and preliminary work on the modelling tools of this research, i.e. abstract argumentation and dynamic epistemic logics. This made possible to provide a formal modelling of the fundamental types of epistemic actions that individuals perform in a debate, such as strategies for information disclosure by speakers and policies of belief update by the hearers. The outcomes of this first stream of work are published in two papers in the proceedings of the International Conference on Logic, Rationality and Interaction (LORI 2019) and of the 2nd International Workshop on Dynamic Logic, new trends and applications (DaLì 2019), held in the context of the Third World Congress on Formal Methods. This material will also be taught at the 2020 edition of the European Summer School of Logic, Language and Information (ESSLLI 2020, postponed to 2021) and will converge in a textbook on logic and abstract argumentation which is now in preparation.
In the second phase, the main effort was directed towards systematizing the initial findings and elaborating a general framework of analysis. This took the form of a general dynamic logic for abstract argumentation, now submitted as a journal paper. An international workshop on the topic of the EDAPOL project was organized in early 2020 and a popular science paper was published. The results were also disseminated through participation at conferences - the 18th International conference of the Italian Association for Artificial Intelligence (AI*IA) and the European conference on Artificial intelligence (ECAI) - and invited talks at seminars - the Logic and Interactive Rationality seminar (LIRa) of the ILLC and the seminar of the Norms and Networks Cluster (NNC) of the University of Groningen.
The third phase aimed at connecting our theoretical framework with simulative and experimental research in social sciences, so to implement our findings in simulative models of opinion change.
The EDAPOL project elaborates the first formal framework to analyse strategic communication of arguments and various policies of belief update in a debate. This has a wide spectrum of applications for the theoretical understanding of opinion change in smaller and larger groups of discussion. In particular, this work opens the way to the refinement of existing simulative models of argument exchange between individuals. Such models are devised to investigate how communication of arguments determines opinion shifts in groups, e.g. polarization. At the present state of the art, models devised by social scientists mostly run under the simplifying assumption that agents have no strategic behaviour. As the logical modelling in this project shows, strategic behaviour has a major impact on dynamics of opinion change. Thus, the implementation of different policies of epistemic actions is expected to accomplish a more realistic modelling of the behavioural features of agents. The collective phenomena arising in a simulative framework where such strategic agents interact are to be tested against experimental results and empirical data to provide new clues for understanding the dynamics of opinion change in groups.