As I’ve written before, it is never safe to trust a potential juror’s own opinion about whether they are biased or not. That is because there has never been much support in the social science for that ability to self-diagnose. That, of course, has not stopped self-diagnosis from being the courts’ most common assumption in voir dire. Somewhat recently, the U.S. Supreme Court, in the case of Skilling v. U.S., has reinforced that the proper and reliable way for trial courts to uncover bias is the simple way: ask them. As an earlier decision held, “It is fair to assume that the method we have relied on since the beginning…usually identifies bias.”

Only that is not fair to assume. The social science has run in the other direction: Not only is it unreliable to use self-diagnosis, it is not even necessarily better than nothing. That is the conclusion of a recent comprehensive study (Yokum, Robertson & Palmer, 2019). Researchers looked at the effectiveness of this most-central tool of American voir dire, and in a nutshell, they found it does not work. The team experimentally tested jury selection in a medical malpractice civil trial using a variety of different research participant populations. They looked at potential bias created by pretrial publicity (because that is an easy bias to create in the lab), but their findings reasonably apply to all biases. Looking at whether the court-approved questions worked to identify those who were biased, and whether those who admitted to bias were any different than those who didn’t, the researchers concluded that the process simply did not work: “Jurors were unable to diagnose their own biases.” In this post, I will take a look at the study, and also consider the question, “What would be different about our jury selection system and practice if we took the failure of self-diagnosis seriously?”

The Research

The study selectively exposed research participants to either pretrial publicity biased against a medical negligence defendant or to control material. Then they walked them through the kinds of questioning they would undergo in court to see who would acknowledge or deny that their views had been affected by the material. Finally, they exposed all, including those who admitted to bias, to a condensed video-recorded medical negligence trial with opening, testimony, closings and instructions, before recording their verdict.

Ultimately, the questioning failed to identify those whose views were impacted by the publicity. Those who were exposed to prejudicial information and who denied bias were still three times more likely than the control group to find for the plaintiff, and the median damages they awarded were five times the median damages of the control group. This is despite the fact that these jurors gave themselves a clean bill of health when it comes to bias.

There are good reasons why the “just ask them” approach wouldn’t work. Self-diagnosis and affirmation of the ability to not be influenced requires awareness of bias in the first place, knowledge of the direction and strength of the bias, a motivation to acknowledge or correct that bias, as well as an ability to adjust one’s response. One might think that the current approach of asking potential jurors for a self-assessment of that ability, while imperfect, is better than nothing in at least weeding out the extremes. However, that is not what the study shows. Ultimately, the team could find no evidence that self-diagnosis procedures helped at all. In fact, while the differences were not statistically significant, they were in the wrong direction: Post-screening, the panels were worse on bias rather than better. Specifically, without the screening questions, 52 percent of those exposed to pretrial publicity found for the plaintiff, and with the screening, 56 percent did.

My takeaway from this study is that it is worse than we think. It is not just that self-diagnosis is incomplete or flawed, it is that it is wholly useless. Ultimately, in the study some admitted to bias and some didn’t. And researchers were unable to identify what made some subjects admit to bias, but it wasn’t the degree of bias. Among those exposed to the biased publicity, there was equal bias among the groups that admitted bias and denied bias. “To be sure,” the authors concluded, “our data does not support the hypothesis that asking jurors whether they can be impartial is at all diagnostic of partiality versus impartiality.” Those are words that litigators and consultants should really take to heart: “Not at all diagnostic.”

The Thought Experiment: What Would It Look Like If We Took It Seriously that Jurors are Unable to Self-Diagnose Bias?

So, trying to take that to heart, it occurred to me to ask that question: What would be different if the system and its practitioners no longer trusted self-diagnosis? I think three things would be different.

There Would Be Expanded Use of the Best Tools

If we asked, “What leads to greater candor from jurors?” the answer from the research would be more reliance on written questionnaires, ideally completed at home. It would be less reliance on leading questions from the judge. And it would be expanded attorney-conducted voir dire, including questioning on relevant attitudes. In all cases, the best practice would be to use questions which don’t play to social desirability by signaling the “correct” response, and the key would be to ask “What do you think?” and not “Can you be fair?”

There Would Be Better Arguments Over Cause Challenges

If we took the failure of self-diagnosis seriously, then the standard for cause challenges would also not center on the ‘magic question’ alone. Instead of simply accepting an affirmative answer to the question, “Will you be able to decide this case solely on the evidence?” judges would more properly apply the standard that applies to judges themselves. The self-perception of immunity to bias is not enough, and if a reasonable person would perceive that a potential juror is likely to be biased based on the circumstances or views expressed, then that appearance of prima facia bias should be enough to justify a challenge for cause, even if that juror does not supply the magic words of a self-diagnosis.

There Would Be Increased Access to Peremptory Strikes

It is currently a common response to the problem of racial bias in jury selection to call for the reduction or elimination of peremptory strikes. While that racial bias (usually by prosecutors in criminal cases) needs to be addressed, throwing away the peremptory strike simply does not make sense as long as the cause-challenge system is so broken. In the real world, bias is going to be subjective, and even as counsel needs to make informed judgments, they will still be judgments. As long as parties are likely to have a good-faith belief in bias that cannot rise to the court’s standard of proof sufficient for a cause-challenge, and as long as cause challenges themselves are tied to the “not at all diagnostic” magic questions of self-diagnosis, then litigants will need to rely on peremptory strikes. For that reason, parties should have more rather than fewer such strikes.

In any setting where we cannot trust, “I can be fair” statements, we still want the parties to be able to make the trial more fair than it would otherwise be. That means divorcing the process from self-diagnosis, and instead equipping litigators with the tools and the training to question effectively and to remove likely bias when they see it.