The roots of polarization in the individual reward system

Abstract

Polarization raises concerns for democracy and society, which have expanded in the internet era where (mis)information has become ubiquitous, its transmission faster than ever, and the freedom and means of opinion expressions are expanding. The origin of polarization however remains unclear, with multiple social and emotional factors and individual reasoning biases likely to explain its current forms. In the present work, we adopt a principled approach and show that polarization tendencies can take root in biased reward processing of new information in favour of choice confirmatory evidence. Through agent-based simulations, we show that confirmation bias in individual learning is an independent mechanism and could be sufficient for creating polarization at group level independently of any additional assumptions about the opinions themselves, a priori beliefs about them, information transmission mechanisms or the structure of social relationship between individuals. This generative process can interact with polarization mechanisms described elsewhere, but constitutes an entrenched biological tendency that helps explain the extraordinary resilience of polarization against mitigating efforts such as dramatic informational change in the environment.

Keywords: affective polarization; confirmation bias; motivated reasoning; polarization; reinforcement learning.

Figure 1.

Attitude polarization in the simple environment. (a) Virtual environment parameters. The images introduce simulations parameters of the virtual environment in the basic version. (b) Attitude distributions. The histograms represent the distributions of the attitudes (i.e. the difference in individual Q-values) of agents at the end of the simulation, for three different confirmation bias levels. Colours represent the extremeness of the preferences.

Figure 2.

Attitude polarization in heterogeneous populations. (a) Agents' heterogeneity. Two types of heterogeneity are considered here. Heterogeneity in terms of initial attitudes (top panel) and heterogeneity in terms of confirmation bias strength. (b) Attitude distributions. Histograms represent the distributions of the attitudes (i.e. the variability in individual Q-values) of agents at the end of the simulation, with heterogeneous initial preferences (top panel) and heterogeneous confirmation bias levels (bottom panel). Colours vividness represents the extremeness of the attitude. Heatmaps represent the distributions of the attitudes of agents grouped according to their initial attitudes (top panel) or confirmation bias levels (bottom panel).

Figure 3.

Attitude polarization in online networks. (a) Information transmission disruptions. In echo chambers mechanism, agents in a social network are more strongly and frequently connected to homogeneous others in terms of attitudes. Consequently, they received preferentially confirmatory information from others. In filter bubbles, algorithmic strategies exploit the agents' style of information consumption and feed them confirmatory information to retain them in the network. (b) Attitude distributions. The histograms represent the distributions of the attitudes (i.e. the difference in individual Q-values) of agents at the end of the simulation, when echo chambers effects (top panel) and filter bubbles effects (bottom panel) are implemented. Faded-colour histograms represent the distributions of the attitudes without any disruption in the information transmission (as seen figure 1b, middle panel). Colours represent the extremeness of the attitudes.

Figure 4.

Attitude polarization resilience. (a) Breakthrough information. The images introduce the implementation of breakthrough news in the simulation. 180 days are now simulated and a breakthrough information is delivered to the world at time t = 90. At that time, the outcome values of the now objectively incontestable object become [0, 1] instead of [−1, 1], increasing its expected value. (b) Attitude distributions. The histograms represent the distributions of the attitudes (i.e. the difference in individual Q-values) of agents at the end of the simulation, when a breakthrough information is delivered at time t = 90 in three different scenarios: confirmation bias only, confirmation bias and echo chambers, confirmation bias and filter bubbles. (c) Polarization evolution. Line plots represent the evolution of the population's attitude towards one object (0 indicating no preference) and the evolution of attitudes standard error, used as a measure of polarization. Colours represent the three different scenarios simulated. Vertical dotted lines represent days threshold for echo chambers and filter bubbles (x = 8) and the onset of the breakthrough information (x = 90).