Just a big misunderstanding? Bias and Bayesian affective polarization

Abstract

I present a model of affective polarization---growth in hostility over time between two parties---via quasi-Bayesian inference. In the model, two agents repeatedly choose actions. Each choice is based on a balance of concerns for private interests and the social good. More weight is put on private interests when an agent's character is intrinsically more self-serving and when the other agent is believed to be more self-serving. Each agent Bayesian updates about the other's character, and dislikes the other more when she is perceived as more self-serving. I characterize the effects on growth in dislike of three biases: a prior bias against the other agent's character, the false consensus bias, and limited strategic thinking. Prior bias against the other's character remains constant or declines over time, and actions do not diverge. The other two biases cause actions to become more extreme over time and repeatedly be ``worse'' than expected, causing mutual growth in dislike, i.e., affective polarization. The magnitude of dislike can become arbitrarily large---even when both players are arbitrarily ``good'' (unselfish). The results imply that seemingly irrelevant cognitive biases can be an important cause of the devolution of relationships, in politics and beyond, and that subtlety and unawareness of bias can be key factors driving the degree of polarization.

“Unmotivated bias” and partisan hostility: Empirical evidence

“…Congleton (2001), for related work on rational ignorance. The present paper is similar in that I also study biases in political beliefs-about the Birther issue in particular, and implicitly about character traits and actions more broadly that might drive favorability ratings (Stone, 2016). Just as voters may face low private costs of holding biased beliefs on issues such as the benefits of free trade, the private costs of holding biased beliefs underlying favorability ratings of the parties are likely also low.…”

“…The primary focus of this paper is direct effects of overprecision on partyism (effects holding fixed social distance and group identity, i.e., holding fixed one's own ideology, partisanship, and/or perception of the out-party's ideology). This focus is motivated partly by a companion paper, Stone (2016). In that paper, a model is presented showing how two (unmotivated) cognitive biases can cause misinterpretation of political actions leading to increased dislike of the out-party, more so as the actions grow more extreme.…”

“…Overprecision could lead to greater hostility indirectly by enhancing one's own political extremism and/or perceived social distance between the parties, in turn directly causing greater partyism for other reasons, such as motivated reasoning. Overprecision could also lead to greater out-party hostility directly by increasing confidence in observed information that superficially reflects negatively on the out-party, namely, extreme political actions by the out-party, as shown theoretically by Stone (2016), and or media content slanted against the out-party. Since these mechanisms (extremism of political actions and slanted media) have become more prevalent over time, they are also potentially relevant to the observed growth over time in partyism.…”

Optimal Echo Chambers

“…Golman, Hagmann, and Loewenstein (2017) review the literature on avoiding free information (by contrast, in our model information has an opportunity cost). Stone (2019) develops a model of affective polarization, in which players' distrust of each others' selflessness grows, whereas in our model there is no misalignment of incentives. De-Marzo, Vayanos, and Zwiebel (2003) suppose that people suffer from "persuasion bias"-a failure to account for the repetition of information.…”