Why are we so polarized? The role of spontaneous ideological inferences in political polarization

In recent years societies all over the world are becoming more and more polarized. In the United States, Poland, Turkey, Syria, Egypt, Kenya, India, Indonesia, and many other nations across the globe citizens are fundamentally divided across a wide range of different topics (Carothers & O’Donohue, 2019). In my PhD project I want to explore the role that spontaneous inferences of ideological labels like conservative, social justice warrior, racist or liberal play in processes of polarization. Previous research has demonstrated that when encountering descriptions of someone else’s behavior, we make sense of it by inferring stable dispositions about them – even when we do not have the intention to do so (e.g. Levordashka & Utz, 2017; Todorov & Uleman, 2002). I expect the same to be true for ideological labels.

There are two ways in which spontaneous ideological inferences might promote polarization. First, we might mentally represent ideological labels not just as mere personality traits but as social categories. In that case, spontaneously inferring that someone is for example a racist could imply identifying them as a member of our outgroup. This process of identifying someone as an outgroup member is known to lead to competitive and discriminatory behavior (Tajfel & Turner, 1979) and increased attitudinal polarization (Abrams et al., 1990). Second, there might be a trade-off between spontaneous inferences of ideological labels and spontaneous inferences of other potential causes of behavior, such as motives or beliefs. If so, then inferring an ideological label about someone would entail not inferring their motives and beliefs. In other words making spontaneous ideological inferences about someone might diminish our ability to take their perspective. Since social perspective taking increases our tendencies to approach and help outgroup members (Todd & Galinsky, 2014), this too might promote polarization.

Taken together, I examine whether we spontaneously infer ideological labels from someone else’s behavior, whether we use these labels to classify them as a member of our outgroup, and whether we simultaneously make less inferences about their motives and beliefs and are therefore less capable of taking their perspective. This research is theoretically informative because it will advance our knowledge on the interplay of different kinds of person inferences and bridge gaps between often separate but deeply related areas of research. Most importantly, however, it is crucial in developing a thorough understanding of the potentially polarizing effects these ideological inferences could have on society. If this research indeed reveals ideological inferences to play an important role in polarization, the ultimate aim of this project will be the development of strategies that might reduce the usage of ideological labels in public discourse and everyday interactions, so as to mitigate their polarizing effects.

Research Methodology

To investigate whether ideological labels are indeed spontaneously inferred from other people’s behavior, I will deploy two established experimental paradigms from research on spontaneous trait inferences: the savings-in-relearning paradigm (Carlston & Skowronski, 1994) and the probe recognition paradigm (Uleman, Hon, Roman, & Moskowitz, 1996). Together these two paradigms can test whether ideological inferences about the actor occur spontaneously during the encoding of behavior descriptions. In order to generate the material needed for these paradigms, I am currently conducting a series of pre-tests on prolific. To examine whether ideological inferences diminish our ability to take another’s perspective, I will adapt a procedure by Todd et al. (2012, Exp. 3) in which participants will be presented a catalogue of prewritten questions, from which they have to choose a certain amount to pose to a person they just made an ideological inference about. The amount of questions chosen that inquire information about the person’s motives and beliefs will be used as an indicator of attempted perspective taking. Lastly, to test whether ideological labels are mentally represented as social categories I will combine the savings-in-relearning paradigm with the who-said-what paradigm (Taylor et al., 1978), which is used to indirectly assess social categorization.

Sample Size

Even though effects for STIs are typically very robust (Bott et al., 2021), due to the novelty of the material and specific research questions, it is difficult to estimate the effect sizes for ideological inferences. For the first study I will therefore use a sample powered for a small effect of η² = .022 for the three-way interaction of interest. With a level of significance of 𝛼 = .05 and a power of 1 - 𝛽 = .80 the projected sample size needed (MorePower 6.0; Campbell & Thompson, 2012) is approximately N = 352 for each of the two paradigms. Given an expected exclusion rate of 5-10%, I will collect data for N = 387 participants. The sample sizes of following studies will be chosen depending on the observed effect sizes. As a rough estimate for study 2 I expect a necessary sample size of N = 112 and for study 3 of N = 387.

Study Costs

The savings-in-relearning paradigm will take around 45 minutes, so that the costs will amount to 387 * 0.75 * £7.80 = £2,264. The probe recognition paradigm will take around 30 minutes, so that the costs will amount to 387 * 0.5 * £7.80 = £1,509. Study 2 will take around 20 minutes, so that the costs will amount to 112 * 1/3 * £7.80 = £291. Study 3 will take around 60 minutes, so that the costs will amount to 387 * 1 * £7.80 = £3,019. In total the costs including the service charge will amount to £9,444.


Detailed pre-registrations for all studies will be progressively added to this project’s OSF page (OSF | Spontaneous Ideological Inferences) as soon as all methodological decisions are made and before data collection begins. The pre-registrations will be based on the “Pre-Registration in Social Psychology (van 't Veer & Giner-Sorolla, 2016): Pre-Registration” template.

Open Science

After the studies are complete, I plan to make my stimulus material, analysis code and the collected data publicly available on the project’s OSF page (see above). I intend to try and publish all of the findings from this project in open access journals.