Bias: What Is It? And When Is It Bad?

Bias is a hot topic. Many people have expressed concern about bias in admissions, hiring, policing, sentencing, social media moderation, and many other aspects of society. The worry is that biases like these are wrong and detrimental.

Upon reflection, however, not all biases are bad. Further, some biases may be good. So bias is not always cause for concern. If that is surprising, then you may want to know what I mean by "bias." This post will briefly describe what bias means in cognitive science and various ways to evaluate whether a bias is bad.

1. What Do I Mean By Bias?

Many of us tend to buy from some products rather than their alternatives: MacOS vs. Windows, Starbucks vs. Dunkin, eBay vs. Amazon, etc. Cognitive scientists often refer to such measurable tendencies as biases (Evans, 1989).

We can measure biases in many ways. For example, we can measure the frequency with which I choose one brand over another. Or we can measure the time it takes me to determine whether I like one brand vs. its competitors. For example, if I tend to buy Apple products more than their competitors' products and be faster to decide that I like an Apple product than a similar product from another company, then I seem to exhibit a measurable bias in favor of Apple products (over their alternatives).

Of course, frequency and timing are not the only measurable forms of bias. There are many more (Hammersley & Gomm, 1997; Petty et al., 1997; Stanovich, Toplak, & West, 2008).

2. Is Bias Always Bad?

Discrimination on the basis of race, sex, and other social categories is widely considered bad. The point of this post is not to dispute the badness of such biases. However, many biases are probably not considered so problematic. For example, you may not care if we tend to pronounce or comprehend certain syllables faster than others. These phonetic biases seem neither malicious nor harmful.

Moreover, some biases may be good. For instance, you probably believe that logically valid arguments are better than fallacious arguments. You may also believe that we should discriminate between causal evidence and merely correlational evidence. These seem like obviously reasonable biases, but even seemingly irrational biases may have benefits (Bortolotti, 2020; Gigerenzer & Brighton, 2009; Johnson, Blumstein, Fowler, and Haselton, 2013).

There are a few implications of such unproblematic or even beneficial biases. One implication is the bias fallacy (Byrd, 2017): inferring that something is wrong or bad solely because of a bias that it exhibits.

3. What Makes A Bad Bias Bad?

If we cannot infer badness or wrongness from the mere existence of a bias, then how do we determine whether biases (or the decisions they cause) are bad? There may not be a universal rule to determine the badness of all biases. So we may need to evaluate each instance of bias on a case-by-case basis. And evaluation can be even more complicated: there is more than one way to evaluate biases.

Consequentialists evaluate biases according to their outcome—e.g., whether they produce unjust inequalities or increase the overall amount of harm. Deontologists evaluate biases by their intrinsic features—e.g., whether they involve harmful intent or constitute unacceptable behavior. Pluralists may embrace both forms of evaluation (as well as others).

So evaluating biases on our own is a challenge but evaluating biases together may be even more difficult. After all, if you and I have a reasonable disagreement about how to properly evaluate a bias, then we may arrive at a stalemate about whether a particular bias is bad (Elqayam & Evans, 2011; Thompson, 2011).

4. Why Does Bias Matter?

Many people get nervous or defensive as soon as I mention that I study bias (e.g., Byrd, 2019). They seem to assume that 'bias' refers to some kind of pernicious prejudice. A bit of reflection reveals that although this is sometimes correct, there seem to be exceptions—and the hard work of determining which cases are exceptional can be illuminating (Evans & Stanovich, 2013). So, when discussing the cognitive science of bias, we may find the need to highlight these exceptions and their implications. One way to do that is to refer to this post.