Sometimes, groups are genuinely found to differ a bit, on average. For instance, it may be found that men are a bit more dishonest than women or that kids from East Asian countries outperform American kids at math, on average. Situations like this often involve people jumping to inaccurate conclusions and spreading misinformation. In this article, we'll explore why, as well as how to think more clearly about these kinds of situations.
Here’s a (hypothetical) depiction of a small average difference in some trait between two groups, represented by two bell curves:
The x-axis represents a trait (e.g., honesty or math scores) and the y-axis represents the frequency of that trait. The blue bell represents the distribution of that trait in one group, and the yellow bell represents the distribution of that trait in the other group. As we can see in this depiction, one group is shifted to the left, indicating a somewhat lower amount of that trait, on average, than the other group.
Of course, this doesn’t mean that everyone from one group has more of the trait in question than everyone in the other group. Far from it. In fact, most people are in the large area of overlap, but we'll assume that there is nevertheless a real difference between the two groups.
Upon seeing results like this for real human traits, many people's thinking will go off the rails in one of two ways:
1️⃣ Some people greatly exaggerate the difference. Let’s call them oversimplifiers. They ignore that the gap is (i) small and (ii) only a difference in averages (not applying to all individuals). They go around saying it's proven that "X's are like this" (e.g., men are liars).
Frequently (but not always), Oversimplifiers adopt this narrative because the idea that there is a big chasm on this trait between these groups fits and reinforces their preconceived worldview.
2️⃣ Alternatively, some people (let’s call them Difference Deniers) pretend or convince themselves that the group difference doesn't exist, even though there truly is a non-negligible gap in the average of the two groups.
Frequently (but not always), Difference Deniers are motivated by their preference for a world without such group-level differences, their fear that such differences can be used to justify harmful things like stereotyping or worse, or their belief that thinking there are such group-level differences is indicative of being a bad person (even when such a difference is real).
The problem with turning small differences in averages into “A’s are like this, B’s are like that”, like Oversimplifiers do, is that it is an inaccurate oversimplification and often unfair to A’s or B’s or both.
The problem with denying the existence of average differences that, while small, really do exist is that you end up believing falsehoods, or you end up lying, or both. By doing so, you may end up unfairly misjudging people who are (without malice) reporting on real average differences, such as researchers who are genuinely trying to understand the topic. And, if you refuse to believe that genuine differences in averages exist, that may lead you to other false conclusions as a consequence, such as beliefs about which policies will be effective for improving outcomes.
How to deal with real differences
To avoid the weaknesses of both the Oversimplifiers and the Difference Deniers, we think the best way to handle these cases is to:
1️⃣ Avoid pre-judging people based on their membership in broad groups – learn about people as individuals before coming to judgments about them. A small difference in group averages is almost never a good justification for assuming those differences apply to individuals. Doing so is, in many cases, an example of the logical error known as the ecological fallacy.
2️⃣ Avoid language like “A’s are like this, B’s are like that” so that you aren’t an Oversimplifier. If what you're actually trying to say is that one group has an average that's a little higher, then state that in all its nuance - do your best to avoid language that could lead to misinterpretation.
3️⃣ Avoid denying that an average difference exists when it really does exist, so that way, you aren’t a Difference Denier. Denying the truth can cause a lot of harms: it can undermine your credibility, it can spread misinformation, and it can lead you to support ineffective policies.
4️⃣ When relevant, remind others that small average differences are not a good basis for judging individuals (epistemically and morally), and point out that the distributions between the two groups are heavily overlapping (when they are) to combat people using differences in the average as a justification for stereotyping.
5️⃣ Point to (when relevant, helpful, and accurate) policies that may help close the gap between the two groups (keeping in mind that some gaps in averages are fine if the trait in question is merely a difference and not something “good” or “bad”)
6️⃣ Point out (when the difference in question is not something people should be judged for) that this attribute should not be a basis for judgment, i.e., that having different values of that trait is completely okay.
Another approach that can be helpful in some cases, which was proposed to us by Guy Srinivasan, is as follows: “[Let's] agree to make decisions as if there were no average difference, since usually all such decisions would turn out the same, and usually when they wouldn’t it’s perpetuating systemic problems to make the decision differently." Note that, importantly, this doesn't mean denying what's true.
Now, we should say, as with any simple categorization, the distinction between Difference Deniers and Oversimplifiers is itself a simplification. Some people will only partly be Difference Deniers or Oversimplifiers, and most of us will not treat all small average differences the same way (e.g., some of us will be a Difference Denier on one topic, and an Oversimplifier on another). People are extremely complex, and this model is purposely simplified in order to help communicate an idea we think is helpful.
When Oversimplifiers and Difference Deniers are Right
Okay, but are there cases where either the Oversimplifiers or Difference Deniers are essentially just right?
Absolutely, there are some.
When a group difference is so huge that the distributions are (at least) nearly non-overlapping, then it’s reasonable to say, “A’s are like this, and B’s are like that.” For instance, it makes sense to say that “blue whales are big, mice are small.” In such cases, the Oversimplifiers aren’t really oversimplifying - they are just applying an accurate model to describe the fact that the groups are different. However, it's worth noting that when we’re talking about human groups, this kind of situation is very rare - almost always the distributions of the groups are heavily overlapped.
On the other hand, in situations when the difference in averages between groups is so small as to be essentially insignificant for all purposes, the Difference Deniers aren’t actually denying reality. For instance, if it turns out that right-handed people are 0.001% better at school work than left-handed people, that difference is so small as to not be meaningfully different from zero for essentially all purposes, and so saying there is “no difference” is an extremely reasonable thing to do. The phrase "no difference" doesn't literally have to mean a difference of zero - it can simply mean that the difference is too small to matter for anything important. There are, in fact, many attributes along which human groups differ so little that “no difference” is an accurate way to describe it (even though the difference is not literally zero to the 10th decimal point).
Sometimes you'll inevitably encounter group differences that genuinely exist. When you do, we hope you'll remember this article, and try to avoid both inaccurate extremes: don't be an Oversimplifier or a Difference Denier. Try to see the world as it actually is, while also treating all individuals fairly.
If you’re interested in taking this kind of thinking further, and practicing nuanced thinking in other domains, we suggest that you try out our Nuanced Thinking Techniques tool: