OPIDPOLA:
Offensiveness, inverse planning and the «concrete/abstract» dissonance. New proxies for measuring affective polarization.

Hyphotesis 1.

The first hypothesis of this project is that both a) the distance between the level of perceived offensiveness when the same sentence is uttered by somebody from our own group versus when it is uttered by someone from the opposite group, and b) the role that abstract-concrete dissonance plays in our perceptions of offensiveness, are associated with polarized contexts and serve as proxies, as reliable proxy indicators, for measuring affective polarization.

Justification of hypothesis 1.

A rough example for a) is the following. If at time t1 the statement p is perceived as a 3 -on a scale from 1 to 7, where 1 is not at all offensive and 7 totally offensive- when made by our group and as a 4 when made by the other group, and at a later time t2 the same statement p is perceived as a 2 when made by our group and as a 4.5 when made by the other group, then at t2 the level of affective polarization is higher than at t1.

In the case of b), the abstract-concrete dissonance has to do with the fact that the abstract principles claimed to be held regarding offensiveness are not, in fact, those that actually modulate our perceptions of offensiveness in cases where the general principle would clearly be involved. In previous studies (Almagro, Hannikainen & Villanueva, 2021; in press), it has been observed that speaker identity is the most important factor guiding concrete judgments of offensiveness, followed by intention and harm. However, when asked in abstract terms about the relevance of these factors, participants indicated intention as the most important one, and identity as the least relevant for considering the same statement as offensive or not.

A rough example for b) is the following. If at t1 the distance between the effect of a certain principle on our concrete judgements and our perception of the importance of that principle in our judgements is, for example, two points, and at t2 the distance is 4 points, then at t2 the level of affective polarization is higher than at t1.

Polarization has to do with having a high confidence in the core ideas of our preferred political group, but we usually do not fully understand what those ideas entail (Funkhouser, 2020; Ganapini, 2021; Williams, 2021), or we do not have enough information to be ascribed such beliefs (Hannon, 2021). In this sense, our increased confidence in the truth of an abstract idea, central to our group (affective polarization), does not materialize in our concrete judgements in specific situations; with our statements we simply express unconditional support for our group, we show that we perceive whatever the other group does as more offensive, but we do not possess sufficient stability to be ascribed the beliefs expressed by our statements.

Hyphotesis 2.

The second hypothesis of the project is that attribution of intent mediates our perceptions of offensiveness, especially when they depend on the speaker’s identity, and this can be modeled by using the model of inverse planning. 

Justification of hypothesis 2.

To formalize this hypothesis, we draw on important advances in cognitive science, which link the understanding of other people’s behavior to a mental ‘inversion’ of the process of planning one’s own actions (or ‘inverse planning’). This computational framework, inspired by the Bayesian reasoning model, stipulates that we infer intention or other underlying mental states as the goal of an agent by virtue of its observed actions and the environment or context in which it performs them (Baker, Saxe & Tenenbaum, 2009; Riordan et al., 2011). Specifically, it assesses whether the observed action is approximately rational (or likely) in that specific context under different assumptions about the agent’s mental state. The idea is that explaining an agent’s actions in terms of mental states requires reversing the model by which similar actions are planned: proceeding backwards to infer the desires, beliefs and intentions that causally explain the agent’s behaviour.

We transfer this model, whose application to motor behavior interpretation tasks has already been successful, to our domain of interest: language interpretation. Except in exceptional cases where the intention is unambiguous (e.g. insults…), in most natural contexts, intentions are uncertain. Applying the inverse planning model to these tasks, interpreters or listeners can be expected to rely on contextual elements (such as the speaker’s membership of the target group, previous attitudes towards the target group, etc.), beyond the action performed (i.e. the content of the utterance), to infer the speaker’s intention in situations of uncertainty. Coupled with the notion that mental states modulate various normative judgements (e.g. of moral blame or condemnation), this thesis implies that certain contextual information (which helps to infer the speaker’s intention) may determine their degree of offensiveness and censure. 

As discussed above, the greater our perceived offensiveness of a statement when it is made by someone from the other group, the greater our inclination to attribute negative intentions. This, in turn, reinforces abstract-concrete dissonance: the greater our inclination to think that it is the others who are offensive or easily offended, when in fact it is a symmetrical effect (Almagro, Hannikainen & Villinueva, in press). The explanation is as follows. Since we perceive the statements of the other group as more offensive than those of our group because we attribute offensive intentions more often to the others than to our own, it is reasonable to conclude that it is the others who are easily offended or who are really offensive. This leads us to defend abstract principles that do not actually guide our behaviour (e.g., «intention is more relevant than identity», «it is others who are offended»). In this sense, the mediation of the attribution of intentions in our perceptions of offensiveness can be explained in terms of inverse planning: although it is the speaker’s identity that triggers our perception of offensiveness, we think that it is the speaker’s intention, as the attribution of negative intentions to others mediates our perception of offensiveness.

The mediation of intentions in our attributions of offensiveness is in turn a control for hypothesis 1: the perceived offensiveness distance between a statement of our group and the same statement of the other group should correlate with the predisposition to attribute negative intentions to the other group in the face of a potentially offensive statement.