Computational models have shown how polarisation can rise among deliberating agents as they approximate epistemic rationality. This paper provides further support for the thesis that polarisation can rise under condition of epistemic rationality, but it does not depend on limitations that extant models rely on, such as memory restrictions or biased evaluation of other agents’ testimony. Instead, deliberation is modelled through agents’ purposeful introduction of arguments and their rational reactions to introductions of others. This process induces polarisation dynamics on its own. A second result is that the effect size of polarisation dynamics correlates with particular types of argumentative behaviour. Polarisation effects can be soothed when agents take into account the opinions of others as premises, and they are amplified as agents fortify their own beliefs. These results underpin the relevance of argumentation as a factor in social-epistemic processes and indicate that rising issue polarisation is not a reliable indicator of epistemic shortcomings.

计算模型已经表明，当深思熟虑的代理人接近认知理性时，两极分化是如何加剧的。本文进一步支持了这样的论点：在认知理性的条件下，两极分化可能会上升，但它并不依赖于现有模型所依赖的限制，例如记忆限制或对其他代理人证词的偏见评估。相反，深思熟虑是通过代理人有目的地引入论点以及他们对他人引入的理性反应来建模的。这个过程本身会引起极化动力学。第二个结果是，极化动态的效应大小与特定类型的争论行为相关。当代理人将他人的意见作为前提时，极化效应可以得到缓解，而当代理人强化自己的信念时，极化效应就会被放大。这些结果支撑了论证作为社会认知过程中一个因素的相关性，并表明日益严重的问题两极分化并不是认知缺陷的可靠指标。