Taking Cognizance of Cognitive Bias

Patrick Shin, Liability for Unconscious Discrimination? A Thought Experiment in the Theory of Employment Discrimination Law, Suffolk University Law School Research Paper No. 10-21, Hastings L. J. (forthcoming), available at SSRN.

It’s not that the question hasn’t been adverted to in much of the cognitive bias scholarship in the employment discrimination area, but most of the articles to date have focused on the empirical questions of the degree to which “implicit bias” exists and the extent to which it might influence real world employment decisions. As for whether truly unconscious discrimination is illegal, the analysis is often extremely truncated. The most common answer is essentially a textualist argument (often by scholars who would eschew that approach in other contexts). That is, those who would make such bias actionable, look to Title VII’s prohibition of discrimination “because of” race or sex and conclude that, where it can be proven to have resulted in an adverse employment action, unconscious bias is prohibited. Another common reaction is to look to the Court’s various condemnations of stereotyping and equate them with cognitive bias, although it is not so clear that, say, the partners in Price Waterhouse weren’t aware that they were unhappy that Ann Hopkins wasn’t conforming to views about appropriate behavior for women.

The absence of deep analyses of what Patrick Shin in Liability for Unconscious Discrimination calls the “naïve question” he addresses — “should implicit bias be a basis for disparate treatment liability” — is understandable. Before we make definitive judgments about whether certain conduct should be illegal, we probably should understand it better than we do at this point in our history. Plus, in the garden-variety employment discrimination case, the hard question is rarely reached because the jury is inferring bias from conduct, and whether the bias is conscious but covert or merely implicit doesn’t matter. Of course, one can imagine issues which force the legal question to the surface – whether to admit expert testimony on cognitive biases, for example, or whether the jury should be instructed that, if they believe the defendant’s disavowal of bias, they must find for it. But it’s not an accident that, fifteen years after Professor Krieger wrote Content of Our Categories, we still don’t have a judicial answer to the question of the legal significance of the phenomenon.

Professor Shin’s article, posted on SSRN, Liability for Unconscious Discrimination: A Thought Experiment in the Theory of Employment Discrimination Law, addresses the normative question directly. Needless to say, I can’t capture the subtleties of his article here. Indeed, the complexity of the enterprise is underscored by the lack of a consensus on what to label what Shin’s title calls “unconscious discrimination” since “cognitive bias,” “implicit bias,” and “subtle discrimination,” among others, are competing for pride of place. But his basic method is to avoid the debates that have so far consumed the literature by making two basic assumptions that take him directly to the normative question he prefers to address: (1) that unconscious discrimination is real, and (2) that it is provable. He then poses a hypothetical he calls “Work Experience I,” which involves an employer who chooses a white with greater work experience over an African American who has interviewed more effectively. Consistent with his assumptions, Shin asks whether, if we could know that experience was trumped presentation in this situation because of the decisionmaker’s unconscious bias, the law should make that illegal, even though the employer honestly believed he was making a decision on the merits of the two candidates.

As is obvious, this is a very difficult question, and Shin approaches it from a variety of perspectives.  He recognizes that Title VII already bars discrimination absent conscious intent – after all, that is the core of the disparate impact theory. He also recognizes reasons why the law might want to bar unconscious discrimination, including achieving the equality goals of the antidiscrimination project and simple fairness to victims who, after all, will be equally out of a job whether the discrimination is conscious or unconscious.  As for contrary arguments, he sets to one side those that question his two assumptions or require empirical judgments (for example, how effective any legal regime making unconscious discrimination might be).

Instead, he believes “liability for unconscious discrimination would remain controversial even if there were convincing evidence that it would in fact reduce workplace inequality overall.”  That is because we are accustomed to assessing decisions in terms of the “justificatory rationales” rather than a “causal conception.” In short, a “causal conception of actionable discrimination allows for the imposition of liability even when the employer acted on considerations that provide legitimate, adequate reasons” for the decision in question.  This “departs radically from our current conception of discrimination because it requires cleaving the causation inquiry entirely from any evaluation of the adequacy of the agent’s subjective rationale for the action in dispute.”

So far, so good. Shin has demonstrated that holding the employer liable in the “Work Experience I” hypo requires a shift in the concept of discrimination; but, as he recognizes, it might be more accurate to say that it’s a shift in the concept of disparate treatment discrimination. But other than being new, why is such a shift problematic? Shin has several worries. The first is that such a shift undercuts the moral objection to discrimination. “Arguably, [discrimination] ceases to have any significance at all, at least as a criticism of the actor.” Secondly, he argues that a pure causation approach renders antidiscrimination law incoherent. In what he labels “Workplace Experience II,” Shin posits an unbiased decisionmaker who prefers a white candidate with more experience to a black candidate with less. Unlike the original hypothetical, the decisionmaker is not influenced by the candidates’ races, but the black candidate’s lack of experience can be traced to his race, and a pure causation approach would seem to require the two decisions to be treated equally.

One common theme of scholarly things I like a lot is that they make me think, and Professor Shin’s piece made me think about the normative question. On the one hand, I think he overstates the significance of “justification” in the discrimination context. Indeed, the word “justificatory” suggests the presence of a legitimate reason — which in fact was at play in his “Work Experience” hypotheticals. But antidiscrimination law has pretty definitively rejected any requirement that the employer’s reason be sensible. So long as the employer’s reason is non-racial, current doctrine does not make it actionable, regardless of whether it is rational, silly, or even illegal under some other statute. “I don’t hire Capricorns” is a fine justification for not hiring a black plaintiff, so long as it was really zodiacal considerations and not race that caused the decision. To the extent current concepts privilege arbitrary actions, why is it so bad that we change our approach?

Shin’s second point, that causation sweeps too broadly, is more difficult for me. Even disparate impact discrimination is narrower than pure causation. It’s narrower, first, because it’s not enough for race to be a factor in a particular decision regarding an individual if there’s no impact on the race to which he belongs, and, second, because the business necessity defense permits some practices with a disparate impact. I guess one response is that, unlike “Work Experience II,” the employer in “Work Experience I” does have some control over his conduct, at least in the sense that he can take steps to debias himself (perhaps by deciding how important experience is before he meets any candidates). But I remain uneasy, which I guess is ultimately what I liked about the piece.