Rationality and group behavior
Individuals in groups often behave in seemingly irrational ways that lead to highly undesirable individual and collective outcomes. Polarization of attitudes in a debate and bystander effects are typical examples of a detrimental group behavior which is mostly due to combined influence of individuals over each other. In the twentieth century, social psychologists have devoted much attention to such phenomena and provided explanatory clues for many of them. In a large number of cases, however, the question remains open as to whether these phenomena show that humans are truly irrational. The aim of this project is to understand some of these phenomena with the main tools available in epistemology: logic, probability theory, game theory, and argumentation theory. Using formal methods allows testing experimental hypotheses in a more controlled setting and makes the study immune to confirmation-biases and other disturbing factors that fuel the actual replicability crisis in social psychology. It also allows varying more parameters than is possible in a lab setting, such as group structure, size and time span. Formal analysis via models may therefore test and aid explanations of hypotheses that were provided by psychological experiments.
Final report
Purpose, development and implementation.
The social context in which individuals interact often causes suboptimal – sometimes catastrophic – collective behaviors to emerge. Lemming effects, extreme polarization of opinions, Kafkian social norms promoting inefficiency are just few examples of such behaviors. One relevant question for epistemology is whether individual irrationality is a necessary condition for these phenomena to arise and propagate. The purpose of this project is to answer this question with combined formal methods, provided by logic, game theory and abstract argumentation.
The project was conducted at the Department of Philosophy of the University of Lund. It was structured into three main blocks of execution, each one investigating a cluster of different types of dynamics generated by interaction, more precisely:
(a) Harmful social norms
(b) Group polarization and echo chambers
(c) Pluralistic ignorance and related phenomena
Blocks (a)-(c) were scheduled as fully independent, for a duration of 9-12 months each. Execution of block (a) was prioritized for the sake of a collaboration with members of the Department of Electrical and Information Technology of Lund University. This block consisted of a game-theoretical analysis and a batch of computer simulations on artificial societies. Afterwards, block (b) on group polarization was approached. The progress of this research gave rise to a number of further, more specific, research questions, which substantiated a spin-off project on the Epistemic and dynamic aspects of polarization. This project was awarded a Marie Curie Fellowship and started in September 2018, in agreement with the RJ board. The execution of the spin-off project shifted the major focus on the research issues related to block (b) (see results and new research questions). The project was terminated before its natural end and the execution of the final block (c), since the researcher entered a new academic position at the National Research Council of Italy.
The project’s three most important results and contributions to the international research front.
Block (a). This study consisted of a game-theoretic analysis and a simulative study of the effect of enacting fair social norms in a mixed society consisting of ”Low-doers” and ”High-doers”. The main aim was to assess whether pro-social in-group behaviour in such a mixed society has a negative impact on social efficiency and welfare. The initial conjecture was – based on straightforward game-theoretic considerations – that it does. To test the robustness of this conjecture, we devised and implemented a multi-agent computer simulation model, where “pro-social Low-Doers” interact with groups with different “moral profiles”. As a first result, the model confirms the game-theoretic insight, according to which the advantage of “fair” low-doing is quite robust in a society and leads, in the long run, to a strict-dominance (in terms of individual welfare) of Low-doers. Our computer model is also designed to test a number of possible countermeasures and their effectiveness. All the results of block (a) are published as an article in the Journal for Artificial Societies and Social Simulations (see publication list).
Block (b). Here the main focus was on the dynamics of polarization and bipolarization of opinions in groups of individuals, i.e. situations where the opinion of the members of a group either converge to one extreme or diverge towards opposite directions after debate. Abstract argumentation enables to explain these collective phenomena in a natural way. The goal was to assess whether polarization presupposes individual irrationality, understood as fully biased assimilation of pieces of information. The second otcome of this project was the elaboration of a rigorous framework to answer these questions, suitable for implementation on multiagent simulative models. A third result, generated by theoretical analysis in this framework, was to show, among other things, that both polarization and bipolarization are open to occur even in situations of transparent disclosure and non-biased assimilation of new information by individuals in a debate. Results are published in two papers in conference proceedings (see publication list).
New research questions generated through the project.
The elaboration of a formal framework of abstract argumentation in block (b) opened new insights and questions concerning the processes of exchange that cause polarization and bipolarization in a group – e.g. what kind of information disclosure and belief update by individuals are more likely to induce these effects. These issues were studied in depth in the spin-off project. The analysis shows that more subtle policies of information disclosure and belief update – hardly identifiable as irrational – have a significant polarizing effect. The formal framework elaborated in this research is ready for implementation in multi-agent simulations, and enables a refinement of existent models in social sciences, where strategic communication of information and biased assimilation are not included as a feature. The results of this line of investigation are open to validation with experimental tests on actual groups of debaters. This opens promising venues for future interdisciplinary research.
The project’s international dimensions, such as contacts and material.
In the initial phase of the project, the main vector of internationalization has been participation to major international conferences such as Logic Rationality and Interaction (LORI 2017), Sapporo, Japan; the European Conference on Multi-agent Systems (EUMAS 2017), Evry, France; and the International Conference of the Italian Association for Artificial Intelligence (AI*IA 2017), Bari, Italy. Exchange was undertaken with the group of Prof. Stephan Hartmann over the researcher’s academic visit (May 2018) at the Munich Center for Mathematical Philosophy (MCMP, LMU Munich), a leading center in Europe for research in social epistemology. Major collaboration was established with researchers at the Institute for Logic, Language and Computation of the University of Amsterdam, which was selected for the execution of the abovementioned Marie Curie project between 2018 and 2020.
Dissemination and collaborations.
The project’s results were mainly disseminated through contributed and invited talks at conferences and seminars, as well as journal articles and papers in the proceedings of major conferences in the field (see publications). Research material on abstract argumentation constituted the basis for an advanced level course (PhD/Master) on ”Topics in Argumentation Theory” – held at Lund University in 2018 – and for a course proposal, in collaboration with Prof. Davide Grossi (University of Groningen), for the European Summer School of Logic Language and Information 2020 (accepted, postponed 2021 due to the COVID crisis). This material will merge in a textbook on abstract argumentation, which is now in preparation. Furthermore, an international workshop on ”The wisdom and madness of crowds: argumentation, information exchange and social interaction” was held online in 2020.
The social context in which individuals interact often causes suboptimal – sometimes catastrophic – collective behaviors to emerge. Lemming effects, extreme polarization of opinions, Kafkian social norms promoting inefficiency are just few examples of such behaviors. One relevant question for epistemology is whether individual irrationality is a necessary condition for these phenomena to arise and propagate. The purpose of this project is to answer this question with combined formal methods, provided by logic, game theory and abstract argumentation.
The project was conducted at the Department of Philosophy of the University of Lund. It was structured into three main blocks of execution, each one investigating a cluster of different types of dynamics generated by interaction, more precisely:
(a) Harmful social norms
(b) Group polarization and echo chambers
(c) Pluralistic ignorance and related phenomena
Blocks (a)-(c) were scheduled as fully independent, for a duration of 9-12 months each. Execution of block (a) was prioritized for the sake of a collaboration with members of the Department of Electrical and Information Technology of Lund University. This block consisted of a game-theoretical analysis and a batch of computer simulations on artificial societies. Afterwards, block (b) on group polarization was approached. The progress of this research gave rise to a number of further, more specific, research questions, which substantiated a spin-off project on the Epistemic and dynamic aspects of polarization. This project was awarded a Marie Curie Fellowship and started in September 2018, in agreement with the RJ board. The execution of the spin-off project shifted the major focus on the research issues related to block (b) (see results and new research questions). The project was terminated before its natural end and the execution of the final block (c), since the researcher entered a new academic position at the National Research Council of Italy.
The project’s three most important results and contributions to the international research front.
Block (a). This study consisted of a game-theoretic analysis and a simulative study of the effect of enacting fair social norms in a mixed society consisting of ”Low-doers” and ”High-doers”. The main aim was to assess whether pro-social in-group behaviour in such a mixed society has a negative impact on social efficiency and welfare. The initial conjecture was – based on straightforward game-theoretic considerations – that it does. To test the robustness of this conjecture, we devised and implemented a multi-agent computer simulation model, where “pro-social Low-Doers” interact with groups with different “moral profiles”. As a first result, the model confirms the game-theoretic insight, according to which the advantage of “fair” low-doing is quite robust in a society and leads, in the long run, to a strict-dominance (in terms of individual welfare) of Low-doers. Our computer model is also designed to test a number of possible countermeasures and their effectiveness. All the results of block (a) are published as an article in the Journal for Artificial Societies and Social Simulations (see publication list).
Block (b). Here the main focus was on the dynamics of polarization and bipolarization of opinions in groups of individuals, i.e. situations where the opinion of the members of a group either converge to one extreme or diverge towards opposite directions after debate. Abstract argumentation enables to explain these collective phenomena in a natural way. The goal was to assess whether polarization presupposes individual irrationality, understood as fully biased assimilation of pieces of information. The second otcome of this project was the elaboration of a rigorous framework to answer these questions, suitable for implementation on multiagent simulative models. A third result, generated by theoretical analysis in this framework, was to show, among other things, that both polarization and bipolarization are open to occur even in situations of transparent disclosure and non-biased assimilation of new information by individuals in a debate. Results are published in two papers in conference proceedings (see publication list).
New research questions generated through the project.
The elaboration of a formal framework of abstract argumentation in block (b) opened new insights and questions concerning the processes of exchange that cause polarization and bipolarization in a group – e.g. what kind of information disclosure and belief update by individuals are more likely to induce these effects. These issues were studied in depth in the spin-off project. The analysis shows that more subtle policies of information disclosure and belief update – hardly identifiable as irrational – have a significant polarizing effect. The formal framework elaborated in this research is ready for implementation in multi-agent simulations, and enables a refinement of existent models in social sciences, where strategic communication of information and biased assimilation are not included as a feature. The results of this line of investigation are open to validation with experimental tests on actual groups of debaters. This opens promising venues for future interdisciplinary research.
The project’s international dimensions, such as contacts and material.
In the initial phase of the project, the main vector of internationalization has been participation to major international conferences such as Logic Rationality and Interaction (LORI 2017), Sapporo, Japan; the European Conference on Multi-agent Systems (EUMAS 2017), Evry, France; and the International Conference of the Italian Association for Artificial Intelligence (AI*IA 2017), Bari, Italy. Exchange was undertaken with the group of Prof. Stephan Hartmann over the researcher’s academic visit (May 2018) at the Munich Center for Mathematical Philosophy (MCMP, LMU Munich), a leading center in Europe for research in social epistemology. Major collaboration was established with researchers at the Institute for Logic, Language and Computation of the University of Amsterdam, which was selected for the execution of the abovementioned Marie Curie project between 2018 and 2020.
Dissemination and collaborations.
The project’s results were mainly disseminated through contributed and invited talks at conferences and seminars, as well as journal articles and papers in the proceedings of major conferences in the field (see publications). Research material on abstract argumentation constituted the basis for an advanced level course (PhD/Master) on ”Topics in Argumentation Theory” – held at Lund University in 2018 – and for a course proposal, in collaboration with Prof. Davide Grossi (University of Groningen), for the European Summer School of Logic Language and Information 2020 (accepted, postponed 2021 due to the COVID crisis). This material will merge in a textbook on abstract argumentation, which is now in preparation. Furthermore, an international workshop on ”The wisdom and madness of crowds: argumentation, information exchange and social interaction” was held online in 2020.