We May Be Biased to See Sexism When It Isn’t Obviously There

Matt Grawitch

This is a guest post by Matt Grawitch, describing research recently published in The Journal of Open Inquiry in the Behavioral Sciences (I am Editor). Matt is a Professor at Saint Louis University where he teaches and studies issues related to leadership, psychologically healthy workplaces, and evidence-based decision making

---

By Matt Grawitch

When we interact with or observe others in social situations (whether in the virtual or real world), we encounter situations where others’ language or behavior lacks clear intent. In such situations, it is human nature to make split-second judgments based on limited information. This occurs a lot on social media sites like Twitter, where a video snippet of a negative interaction between people will be shared and re-shared. People will line up to make all sorts of inferences and attributions about who was in the wrong. And they do this with a high degree of confidence.

But most of the time these snippets lack sufficient information to draw any reasonable conclusions. Even in more real-world interactions, such as a brief interaction with a salesperson, waiter, or some other stranger, we often lack sufficient information to make confident attributions about others’ motives, but that often doesn’t stop us from interpreting that interaction negatively (he’s clearly mansplaining, she’s just another Karen, they’re obviously racist).

One context in which this tends to happen involves attributions people make about sexism, even when there’s no clear evidence of intent. We all possess sex-based prototypes about who perpetrates and who experiences sexism. It’s possible that when situational factors sufficiently align with those prototypes, we’re biased to infer sexist intent, even if such intent isn’t obvious.

And that is what my colleagues and I tested in a series of studies we recently published in the Journal of Open Inquiry in Behavioral Sciences.

Across 3 studies, we asked people to read and respond to a scenario describing an interaction between a bank employee and a customer. Our experimental manipulation simply involved varying the sex of the banker and the sex of the customer:

[Julie/Tim] walks into the local bank and is greeted by [Anna/Bob], the assistant manager of the bank, whom [she/he] has dealt with before. They walk into [Anna/Bob]’s office to discuss what the bank can do to help [Julie/Tim] invest some money. As [Julie/Tim] walks in [Anna/Bob] says: “Hello [Julie/Tim]. That's a very nice suit you're wearing. You look great.”

[Julie/Tim] explains that [she/he] is interested in investing $30,000. As they are talking [Anna/Bob] notes that [Julie/Tim] has quite a bit of money that [she/he] wants to invest. [Anna/Bob] then goes on to say: “You're lucky to have so much money to invest.” [Julie/Tim] then begins to present [his/her] ideas concerning how [she/he] would like to invest [her/his] money. [She/He] explains that [she’s/he’s] worried about the stock market and that [she/he] is considering buying tax-free bonds. [Anna/Bob] responds by saying: “No, that's a bad idea. Tax-free bonds have a very low yield. You're better off investing in a mutual stock fund.”

At the end of the meeting. [Anna/Bob] gets up from behind her desk and puts [his/her] arm around [Julie/Tim]’s shoulders and says, “We will do all we can here to help you anytime you need us. Thanks for banking with us.”

Across all three studies, we found similar interaction effects (Figure 1 presents the aggregate effect across the 3 studies): when the banker was male and the customer was female, participants were much more likely to infer that the banker was sexist than in the other conditions. But we also found that the second-highest ratings occurred when the banker was female and the customer was male, suggesting that, though muted, the effect can occur when females are the perceived perpetrators. So, females can be perceived as sexist too, especially when interacting with males – the effects just tend to be weaker. These effects were consistent for both male and female participants.

Figure 1

In each of the three studies, we tested to see if there was some sort of mechanism that helped to explain what was happening:

In Study 1, we looked specifically at social justice attitudes (which concerns people’s positive attitudes toward social justice initiatives as a part of a scale developed by Torres-Harding et al., 2012) and concern for political correctness (which concerns emotional reactions to politically incorrect language using a scale developed by Strauts & Blanton, 2015). All we found was that those with higher social justice attitudes tended to rate the banker as more sexist when the customer was female, but there were no differing effects when the banker was male as opposed to female.

In Study 2, we looked specifically at attitudes toward men (using a scale that assessed hostility and benevolence toward men developed by Glick & Fiske, 1999), wondering if those who held more negative assumptions about men might be more inclined to demonstrate this effect. Hostility toward men did show a direct positive association with sexism ratings, but those who held more hostile attitudes simply tended to provide higher sexism ratings, regardless of the sex composition of the interactants.

In Study 3, we looked at neosexism, which is defined as the “manifestation of a conflict between egalitarian values and residual negative feelings toward women” (Tougas et al., 1995, p. 843), with the assumption that “those who are prejudiced couch their negatively charged beliefs about women in the language of equality rather than the language of inferiority” (p. 847).

For example, this is an item from the Neosexism scale:

In order not to appear sexist, many men are inclined to overcompensate women.”

Agreeing with this item is coded as a form of sexism.

• This is Lee commentiung here: The Tougas et al 1995 paper is a classic case in which the main conclusion – neosexism predicts sexist prejudice – is imported rather than demonstrated.  This is because the main outcome was support/opposition to affirmative action for women.  Higher neosexism scores predicted greater opposition to affirmative action – whether this is any type of sexism at all cannot possibly be demonstrated by the existence of the association because opposition to affirmative action is not inherently a form of sexism. Back to Matt’s report…

We found that when respondents reported higher levels of neosexism, the interaction effect disappeared. Those higher in neosexism reported no differences in sexism ratings across the conditions. In other words, they weren't inclined to make attributions of sexism when the situation was ambiguous. But for those who demonstrated moderate to low neosexism (i.e., they believed society was still highly tilted to favor males), the interaction effect was strong, i.e., they saw more sexism when a male banker interacted with a female customer than in the other conditions.

• Lee commenting here again, I just can’t resist: I find this result very ironically amusing. People “high in neosexism” {but see my earlier comment on the Tougas paper} showed less biased interpretations of the ambiguous interaction; people “low in neosexism” were biased in their attributions of sexism. Its almost as if social psychologists sometimes don’t know wtf their own findings mean, something I have addressed in tons of papers, such as this one. Ok, back to Matt…

We also found that the more sexist respondents believed the banker to be, the less likely they were to rate the interaction favorably. This effect was magnified for respondents holding low to moderate neosexist attitudes, though only in the male banker-female customer condition.

What Does All This Mean?

In a more general sense, the series of studies we conducted indicate that slight variations in the characteristics of a situation (the cues present in that situation) that have little to do with the intent of a message may play an important role in the attributions people make about others when that intent is ambiguous. The cues in the male banker-female customer condition were aligned with existing sexism prototypes – where the perpetrator is male and the victim is female – which biased participants in that condition toward stronger attributions that the behavior was sexist. But the fact that somewhat higher ratings were found when sex roles were reversed suggests that a female perpetrator-male victim prototype may also exist but is less accessible.

The presence of these prototypes could lead to a de facto bias toward making sexist attributions in opposite-sex interactions (based on the hierarchy of social prediction; Bach & Schenke, 2017). Because the cues present in the situation match the prototype and there’s no information present to suggest the prototype is inapplicable, people end up being more likely to conclude sexist intent (or at least being unsure whether to reject that conclusion).

What this all means is that we often make attributions about other people’s intent without evidence to validly draw that conclusion. Instead, when the situation matches our own preconceived assumptions about the way the world works, we tend to default to making attributions consistent with those assumptions unless there is evidence to contradict them. In other words, if we’re biased toward some conclusion and the situation does nothing to cause us to reject that conclusion, then we’re much more likely to stick with our bias. And that means the same seemingly benign situation may be judged differently, merely because of who (sex composition-wise) is interacting. And that judgment can lead to situations that amount to someone being guilty until proven innocent.

From a more practical standpoint, the implication here is that we should avoid, when possible, making attributions about others’ sexist intent without clear and convincing evidence of that intent. Otherwise, the attributions we make may say more about us than it does those whose behavior we judge.

References

Bach, P., & Schenke, K. C. (2017). Predictive social perception: Towards a unifying framework

from action observation to person knowledge. Social and Personality Psychology Compass, 11, Article e12312. https://doi.org/10.1111/spc3.12312

Glick, P. & Fiske, S. T. (1999). The Ambivalence toward Men Inventory: Differentiating hostile

and benevolent beliefs about men. Psychology of Women Quarterly, 23, 519-536. https://doi.org/10.1111/j.1471-6402.1999.tb00379.x

Strauts, E., & Blanton, H. (2015). That’s not funny: Instrument validation of the concern for

political correctness scale. Personality and Individual Differences, 80, 32-40. https://doi.org/10.1016/j.paid.2015.02.012

Torres-Harding, S. R., Siers, B., & Olson, B. D. (2012). Development and psychometric

evaluation of the Social Justice Scale (SJS). American Journal of Community Psychology, 50, 77-88. https://doi.org/10.1007/s10464-011-9478-2

Tougas, F., Brown, R., Beaton, A. M., & Joly, S. (1995). Neosexism: Plus ça change, plus c’est

pareil. Personality and Social Psychology Bulletin, 21, 842-849. https://doi.org/10.1177/0146167295218007