Social Depolarization and Diversity of Opinions-Unified ABM Framework.

Most sociophysics opinion dynamics simulations assume that contacts between agents lead to greater similarity of opinions, and that there is a tendency for agents having similar opinions to group together. These mechanisms result, in many types of models, in significant polarization, understood as separation between groups of agents having conflicting opinions. The addition of inflexible agents (zealots) or mechanisms, which drive conflicting opinions even further apart, only exacerbates these polarizing processes. Using a universal mathematical framework, formulated in the language of utility functions, we present novel simulation results. They combine polarizing tendencies with mechanisms potentially favoring diverse, non-polarized environments. The simulations are aimed at answering the following question: How can non-polarized systems exist in stable configurations? The framework enables easy introduction, and study, of the effects of external "pro-diversity", and its contribution to the utility function. Specific examples presented in this paper include an extension of the classic square geometry Ising-like model, in which agents modify their opinions, and a dynamic scale-free network system with two different mechanisms promoting local diversity, where agents modify the structure of the connecting network while keeping their opinions stable. Despite the differences between these models, they show fundamental similarities in results in terms of the existence of low temperature, stable, locally and globally diverse states, i.e., states in which agents with differing opinions remain closely linked. While these results do not answer the socially relevant question of how to combat the growing polarization observed in many modern democratic societies, they open a path towards modeling polarization diminishing activities. These, in turn, could act as guidance for implementing actual depolarization social strategies.

Hammering with the telescope.

The rapid pace in which various Artificial Intelligence and Machine Learning tools are developed, both within the research community and outside of it, often discourages the involved researchers from taking time to consider potential consequences and applications of the technical advances, especially the unintended ones. While there are notable exceptions to this "gold rush" tendency, individuals and groups providing careful analyses and recommendations for future actions, their adoption remains, at best, limited. This essay presents an analysis of the ethical (and not only) challenges connected with the applications of AI/ML methods in the socio-legal domain.

On the role of chance in fencing tournaments: An agent-based approach.

It is a widespread belief that success is mainly due to innate qualities rather than external forces. This is particularly true in sports competitions, where individual talent is usually considered the main, if not the only, ingredient to reach success. In this study, we explore the limits of this belief by quantifying the relative weight of talent and chance in fencing, a combat sport involving a weapon, with the help of both real data and agent-based simulations. Fencing competitions are structured as direct elimination tournaments, where randomness is explicitly present in some rules. We focused on épée, which is one of three disciplines. We collected data on international competition results and annual rankings, in the range 2008-2020, for male and female fencers under 20 years old (Junior category). Then, we built the model calibrated on our dataset and parametrized by just one free variable a, describing the importance of talent-and, consequently, of chance-in competitions (a = 1 indicates the ideal scenario where only talent matters, a = 0 the complete random one). Our agent-based approach can reproduce the main stylized facts observed in data, at the level of both single tournaments and the entire careers of a given community of épée fencers. We find that simulations approximate very well the data for both Junior Men and Women when talent weights slightly less than chance, i.e. when a is around 0.45. We conclude that the role of chance in fencing is unusually high and it probably represents an extreme case for individual sports. Our findings shed light on the importance of external factors in both athletes' results in tournaments and throughout their career, making even more unfair the "winner-takes-all" disparities that often occur between the winner and the other classified competitors.

Agent Based Model of Anti-Vaccination Movements: Simulations and Comparison with Empirical Data.

Background: A realistic description of the social processes leading to the increasing reluctance to various forms of vaccination is a very challenging task. This is due to the complexity of the psychological and social mechanisms determining the positioning of individuals and groups against vaccination and associated activities. Understanding the role played by social media and the Internet in the current spread of the anti-vaccination (AV) movement is of crucial importance. Methods: We present novel, long-term Big Data analyses of Internet activity connected with the AV movement for such different societies as the US and Poland. The datasets we analyzed cover multiyear periods preceding the COVID-19 pandemic, documenting the behavior of vaccine related Internet activity with high temporal resolution. To understand the empirical observations, in particular the mechanism driving the peaks of AV activity, we propose an Agent Based Model (ABM) of the AV movement. The model includes the interplay between multiple driving factors: contacts with medical practitioners and public vaccination campaigns, interpersonal communication, and the influence of the infosphere (social networks, WEB pages, user comments, etc.). The model takes into account the difference between the rational approach of the pro-vaccination information providers and the largely emotional appeal of anti-vaccination propaganda. Results: The datasets studied show the presence of short-lived, high intensity activity peaks, much higher than the low activity background. The peaks are seemingly random in size and time separation. Such behavior strongly suggests a nonlinear nature for the social interactions driving the AV movement instead of the slow, gradual growth typical of linear processes. The ABM simulations reproduce the observed temporal behavior of the AV interest very closely. For a range of parameters, the simulations result in a relatively small fraction of people refusing vaccination, but a slight change in critical parameters (such as willingness to post anti-vaccination information) may lead to a catastrophic breakdown of vaccination support in the model society, due to nonlinear feedback effects. The model allows the effectiveness of strategies combating the anti-vaccination movement to be studied. An increase in intensity of standard pro-vaccination communications by government agencies and medical personnel is found to have little effect. On the other hand, focused campaigns using the Internet and social media and copying the highly emotional and narrative-focused format used by the anti-vaccination activists can diminish the AV influence. Similar effects result from censoring and taking down anti-vaccination communications by social media platforms. The benefit of such tactics might, however, be offset by their social cost, for example, the increased polarization and potential to exploit it for political goals, or increased 'persecution' and 'martyrdom' tropes.

Social Simulation Models at the Ethical Crossroads.

Computational models of group opinion dynamics are one of the most active fields of sociophysics. In recent years, advances in model complexity and, in particular, the possibility to connect these models with detailed data describing individual behaviors, preferences and activities, have opened the way for the simulations to describe quantitatively selected, real world social systems. The simulations could be then used to study 'what-if' scenarios for opinion change campaigns, political, ideological or commercial. The possibility of the practical application of the attitude change models necessitates that the research community working in the field should consider more seriously the moral aspects of their efforts, in particular the potential for their use for unintended goals. The paper discusses these issues, and offers a suggestion for a new research direction: using the attitude models to increase the awareness and detection of social manipulation cases. Such research would offer a scientific challenge and meet the ethical criteria.

Quantitative Agent Based Model of Opinion Dynamics: Polish Elections of 2015.

We present results of an abstract, agent based model of opinion dynamics simulations based on the emotion/information/opinion (E/I/O) approach, applied to a strongly polarized society, corresponding to the Polish political scene between 2005 and 2015. Under certain conditions the model leads to metastable coexistence of two subcommunities of comparable size (supporting the corresponding opinions)-which corresponds to the bipartisan split found in Poland. Spurred by the recent breakdown of this political duopoly, which occurred in 2015, we present a model extension that describes both the long term coexistence of the two opposing opinions and a rapid, transitory change due to the appearance of a third party alternative. We provide quantitative comparison of the model with the results of polls and elections in Poland, testing the assumptions related to the modeled processes and the parameters used in the simulations. It is shown, that when the propaganda messages of the two incumbent parties differ in emotional tone, the political status quo may be unstable. The asymmetry of the emotions within the support bases of the two parties allows one of them to be 'invaded' by a newcomer third party very quickly, while the second remains immune to such invasion.

Quantitative agent based model of user behavior in an Internet discussion forum.

The paper presents an agent based simulation of opinion evolution, based on a nonlinear emotion/information/opinion (E/I/O) individual dynamics, to an actual Internet discussion forum. The goal is to reproduce the results of two-year long observations and analyses of the user communication behavior and of the expressed opinions and emotions, via simulations using an agent based model. The model allowed to derive various characteristics of the forum, including the distribution of user activity and popularity (outdegree and indegree), the distribution of length of dialogs between the participants, their political sympathies and the emotional content and purpose of the comments. The parameters used in the model have intuitive meanings, and can be translated into psychological observables.

Discrete model of opinion changes using knowledge and emotions as control variables.

We present a new model of opinion changes dependent on the agents emotional state and their information about the issue in question. Our goal is to construct a simple, yet nontrivial and flexible representation of individual attitude dynamics for agent based simulations, that could be used in a variety of social environments. The model is a discrete version of the cusp catastrophe model of opinion dynamics in which information is treated as the normal factor while emotional arousal (agitation level determining agent receptiveness and rationality) is treated as the splitting factor. Both variables determine the resulting agent opinion, which itself can be in favor of the studied position, against it, or neutral. Thanks to the flexibility of implementing communication between the agents, the model is potentially applicable in a wide range of situations. As an example of the model application, we study the dynamics of a set of agents communicating among themselves via messages. In the example, we chose the simplest, fully connected communication topology, to focus on the effects of the individual opinion dynamics, and to look for stable final distributions of agents with different emotions, information and opinions. Even for such simplified system, the model shows complex behavior, including phase transitions due to symmetry breaking by external propaganda.