From Bench to Jury Box: The Impact of Bias on Judges and Jurors

Introduction

Legal systems aim to ensure justice. However, biased decisions cause it to be otherwise. Sometimes biases directly exist in the system. For instance, in tort law, the res ipsa loquitur doctrine, meaning the thing speaks for itself, contradicts the probability theory. It allows courts to attribute the cause of harm to the defendant if the defendant cannot come up with evidence suggesting otherwise. Thus, courts wrongfully attribute rare events to extraordinary causes without powerful evidence (Kaye, 1979). Observational studies and experiments yield data implying the effect of the unseen player in the justice system. (Liu & Li, 2019). The disparity between the conviction rates of different racial groups is one way we can examine biases, as African Americans are four times more likely to be imprisoned compared to whites (Wagner, 2012).

Legal Realism suggests that judges are under the influence of social, psychological, and political factors that hinder the mere application of legal principles. Conversely, Legal Formalism holds that judges only consider legal principles (Danziger et al., 2011). As a supporter of the latter, Kozinski suggests constraints of the process prohibit judges from using their ideologies in decisions (Kozinski, 1993). However, the disparity between sentencing times for different racial backgrounds on the same crimes or the hugely different acceptance rates given to asylum seekers are only two situations suggesting otherwise (Schoenholtz et al., 2007; Wistrich & Rachlinski, 2017). Besides, it is not argued that judges completely disregard legal principles. Biased decisions occur because they affect the way those principles are applied. For instance, when judges believe the defendant to be guilty, a tendency to look for incriminating evidence arises (Rassin et al., 2010).

Biases are categorized into two groups, namely social and cognitive biases (Stafford, 2017). The former is defined as our instinctive impressions of people based on their larger social group. These are compatible with implicit biases, which may result in “a deviation of an attitude” (Kang, 2021). Its effects on judicial decisions will be examined in detail with experimental results of the Implicit Association Test. The latter describes a systematic way of thinking that leads to systematic errors (Stafford, 2017). Two of the most common examples of cognitive biases -confirmation bias and anchoring effect- will be examined with specific examples in court settings.

Decision-making happens with two systems, System 1 and System 2 (Tversky & Kahneman, 1974). System 1 operates with little effort and tends to simplify complex problems with heuristics. These shortcuts can be useful when the cost of a potential mistake is less than the gain from time and effort. Hence, increase in detail and complexity of a problem with time constraints leads people to use heuristics, as the gain from effort and time increases more compared to the cost. System 2, on the other hand, is responsible for carrying out controlled operations, requiring more effort. But it is often lazy to make them. Consequently, decisions are significantly influenced by System 1, especially when cognitive constraint is high. These effects are prevailing in the courtroom with detailed cases under time constraints. However, in the judicial context, stakes are always too high, and the gain from time and effort will never offset the cost of a potential mistake, hindering one’s right to justice. Hence, the use of heuristics by judges and jurors is never justifiable.

Before my arguments, I want to explain how common law systems operate with judges and jurors. Jurors consist of ordinary citizens acting as legal decision-makers. They may operate independently from judges or form mixed decision-making bodies with them (Goldbach & Hans, 2014). In common law systems, jurors are more likely to act independently (Thaman, 2011). Therefore, the power dynamic between judges and jurors can be useful for policies, as jurors’ independence can create instances in which separate biases affect judicial decision-making bodies. Jurors also work with evidence as “fact-finders.”. The “story model” suggests that jurors can be anchored to their first-ever impression of evidence that influences their final decision (Guthrie et al., 2001). Once all evidence is presented during trial, jurors have closed deliberations to arrive at the verdict. Judges are responsible for managing the trial and standing neutral. After jurors arrive at a verdict, the judge makes the final decision accordingly.

The situation for jurors should be examined with more caution than that of judges. Since they lack the qualifications that could make judges less susceptible to biases in some contexts -like their sense of responsibility, education, and experience- jurors can be more likely to be influenced by biases in their decisions (Guthrie et al., 2001). Jurors can be anchored to initial offers for damage awards, can wrongfully evaluate possibilities, and can undermine the importance of statistical evidence (Broeder, 1959; Faigman & Baglioni, 1988; Guthrie & Korobkin, 1994). They can also fall for biases when interpreting confessions (Kassin, 2008). “Fundamental Attribution Error” and overwhelm can cause jurors to assume confessions to be true. Thus, situational causes are underestimated while outcomes are attributed to the personal character of the individual. Furthermore, group dynamics can affect their verdict. It may seem like the decision made by a group would be successful in balancing out biases; however, in-group effects, such as group polarization and majority effect, can also equally (if not more) affect jurors. (Bray & Noble, 1978).

This article stresses the importance of considering both judicial decision-making bodies together to construct the right policies. Only then could the systematic failure to ensure justice be mitigated. The first part discusses the effects of confirmation bias. The second part explains the role of anchors and how jurors could be less susceptible to some. The third part examines implicit biases and judges’ efforts in mitigation. The last part proposes possible solutions.

Discussion

Confirmation Bias

Kassin (2005) describes confirmation bias as a “warehouse of psychology research that suggests that once people form an impression, they unwittingly seek, interpret, and create behavioral data that verify it.” This applies to judicial decisions since judges’ initial decisions can lead them to look for evidence that proves them right (Rassin et al., 2010). The same tendency is also prevailing for jurors (P. G. Devine & Ostrom, 1985). Hence, initial decisions can affect the way the subsequent pieces of evidence are perceived. These factors -which should not be important for a case- affect the jury verdict and the decision of the judge. Carlson et al.'s (2001) mock juror experiment shows how confirmation bias distorts the evaluation of evidence. Jurors could struggle to keep an open mind until all the evidence is presented. The highly detailed cases with ample amounts of evidence are more likely to lead jurors to stick with their initial decisions as the cognitive constraint of the process increases.

Rassin et al.'s experiment (2010) illustrates this tendency. 79 law students were given a hypothetical case of physical abuse based on a previous study done by De Keijser and van Koppen (2007). After reading the case, participants indicated if they found the defendant to be guilty or innocent. After that, they were given a list of 20 additional investigations they could pursue. Half of the investigations were incriminating, and the other half were exonerating. Incriminating investigations were likely to result in evidence in line with their initial decisions, and exonerating investigations were likely to do the opposite. Participants who found the defendant to be guilty (33 out of 79) selected 4.6 investigations on average, and 57% of them were incriminating in nature. Those who thought the defendant was innocent (46 out of 79) averaged 5.4 additional investigations, and 45% of the investigations were incriminating. Additionally, the selection of incriminating evidence had a positive relation with evidence strength and crime severity. Hence, as the cognitive constraint increased, the use of heuristics increased. Despite definitive causal inferences, the study shows the extent to which the progression of real cases can be affected by confirmation bias.

Inadmissible evidence also triggers confirmation bias. Inadmissible evidence is defined as evidence that cannot be further used to prove a party’s claim (Cornell Legal Information Institute, n.d.). This type of evidence should be dismissed in decision-making. However, jurors cannot fully omit them. Its content, the source’s credibility, and its relevance to other evidence affect the degree of the effect of inadmissible evidence (D. J. Devine et al., 2001). Even though its effects vary, Steblay et al.‘s meta-analysis (2006) of 175 experiments shows its general pattern. The effects of inadmissible evidence and the relevant legal instructions given in all experiments were hypothesized in five categories, focusing on three main factors: exposure to inadmissible evidence, instructions given to the jurors, and the timing of instructions. The main findings of the analysis are coherent with the previously argued effects. Exposure to inadmissible evidence caused an increase in the conviction rates (in the experiments in which the evidence favored the prosecution), even with instructions, compared to the control group in which there was no mention of that evidence, 46% versus 37%, respectively. Additionally, end-of-trial instructions produced lower conviction rates compared to the group that was exposed to the evidence with no instruction. This highlights an aspect that future research should focus on. One more key finding of the analysis is the need for clear explanations on inadmissibility. If this was lacking, jurors’ efforts to disregard the evidence decreased. Hence, giving jurors clear instructions and explanations as to why the evidence was deemed inadmissible is crucial to successfully mitigate their effects.

The power dynamic between judges and jurors can reduce the effects of inadmissible evidence. Judges can decide not to inform jurors about the inadmissible evidence altogether to prevent them from forming biases. As it is evident in Steblay et al. (2006), the effect of inadmissible evidence is best mitigated when the evidence is not mentioned at all. However, this procedure only protects the jurors. There is no mechanism to protect judges from their effects. Nevertheless, it gives an insight into the way policies should be constructed. The policies that prioritize the relationship between judges and jurors could decrease the systematic failure to sustain justice.

Anchoring

Anchoring is the “excessive reliance on numeric reference points when making numeric judgments” (Tversky & Kahneman, 1974). Anchoring influences people when making estimations for an unknown value. The tendency to adjust the anchor is not enough to counter the effects, leading to biased final estimates. The same effects can also be seen in judicial contexts, as the process is also uncertain. Observational studies imply that anchoring can play a role in real cases. For instance, Mugerman et al.'s analysis (2020) shows how a procedure in bankruptcy cases anchors judges for a value that should be irrelevant. The official receivers, who are appointed by the judge, are responsible for gathering any related information and recommending a value. Even though this recommended value is not binding, it is highly influential in the final decision, and deviations from the value are extremely rare. The recommended value could be seen as a good estimate, considering the official receivers spend a large amount of time gathering the information. However, procedures like this hinder the independence of legal systems as they result in giving nearly full authority over judicial decisions to officials who are not judicial decision-making bodies. Thus, procedures like the one discussed can lead the legal systems to become vulnerable to outside influences. Further, the recommended value is also prone to the official receiver’s biases. Consequently, this procedure results in decisions driven by the biases of an official who is not responsible for making those decisions.

The influence of anchors can be hard to notice, as they still affect the final decision even when they are completely irrelevant. For instance, when the participants were asked to estimate the percentage of African nations in the UN, their decisions were shown to be anchored to an irrelevant anchor: the number on the wheel of fortune they had just spun. (Tversky & Kahneman, 1974). Englich et al. (2006) illustrate the same effect in a series of experiments by focusing on three irrelevant anchors: a journalist’s suggested sentencing time, a prosecutor’s sentencing time demand, and a random die the judges threw before making their final decisions. All anchors were shown to affect final decisions. Also, the expertise of the judges did not prevent this. Since expertise is not relevant to mitigate the effects, jurors and judges can likely be anchored to the same values during the trial. For instance, both judges and jurors can be under the influence of extreme settlement demands when deciding on damage awards. However, the knowledge of judges and jurors is different from each other. Judges can be more prone to some anchors, such as statutory damage caps, mandatory sentencing times, and prior cases, while jurors would simply lack the information about them. (Wistrich & Rachlinski, 2017).

The study by Rachlinski et al. (2015) on 115 Canadian trial judges and 65 newly elected trial judges in New York State shows the effect of statutory damage caps. Both Canadian and New York judges were separated into two groups. The difference in treatments was an additional sentence in which the defendant explicitly stated the damage cap. The results for both groups were similar. The explicit statement of the amount of the cap acted as an anchor, which resulted in higher damage awards. Sentencing times are also affected by anchors such as sentencing demands and mandatory sentencing times. When the initial demand for the sentencing time is longer, the final decision tends to be longer (Englich & Mussweiler, 2001). Mandatory sentencing times are expected to result in shorter sentencing times than the actual value, whereas damage caps are expected to result in higher damage awards than the actual value. Jurors, on the other hand, can mitigate the effects of statutory caps simply because they lack the information. Thus, the effect could be balanced between jurors and judges when there is an imposed statutory cap if the judges decide not to inform jurors about it (Guthrie et al., 2001).

The anchors discussed above, statutory damage caps and mandatory sentencing guidelines, can be argued to be reducing the effects of other, more irrelevant anchors. However, as every case is unique, these anchors also reduce the independence of legal systems, and this “solution” would only create more systematic errors.

Implicit Biases

Kang (2021) defines implicit bias as “deviation from some baseline of compassion which is presumably neutral or fair. Because we are thinking about human beings, bias here means some deviation in an attitude or stereotype about a social category”. The Implicit Association Test (IAT) is used to expose implicit biases (Bertrand et al., 2005). The test is typically done over a computer, and participants categorize words and faces that appear on the screen. Words are categorized as either good or bad, and pictures are categorized as African-American or white. Participants are instructed to press the side of the keyboard where the stimulus belongs. Every participant completes two versions of the task. First version pairs “African-American” with “bad” and “white” with “good.” The second version has the opposite pairing, with “African-American” with “good” and “white” with “bad.” Most participants respond more quickly in the first version. The difference in the response times signals an implicit bias against African Americans. The test is useful for understanding the underlying implicit biases because the time pressure leads people to use shortcuts that are driven by their biases (Tversky & Kahneman, 1974). This implies that the limited time judges have on cases can be a factor in causing the use of heuristics.

Judges and jurors who participated in the test responded faster in the first version (Wistrich & Rachlinski, 2017). The same study also examined the extent to which the judges’ and jurors’ decisions would be driven by their implicit biases. Results were hopeful for judges. The difference in treatments was an explicit statement of the defendant’s race. The percentage of judges who convicted the defendant did not differ between the two groups. However, similar results were not reached with jurors; they were more likely to convict the defendant when the race was explicitly stated. The results show that even though judges have implicit biases, as shown by the IAT, it did not affect their judgments when there was an explicit statement about a “potential and well-known bias,” like race. The same results could not be reached when judges were primed to the word African American. The judges who performed better in the first version of the IAT were likely to give more severe punishments to African American defendants. Similar results were also reached when the study was done on police officers (Graham & Lowery, 2004). Hence, there is enough evidence to argue that the difference between the conviction rates for different groups (Gross et al., 2017) can be reduced if judges are nudged to become aware of potential biases.

The results may look intricate, but they give important insights into the motives of judges when making decisions. Judges could mitigate biases when race was explicitly stated. However, when there was no explicit statement, judges acted upon their biases. Levinson et al. (2017) show that even though judges had biases against Jewish, Christian, and Asian litigants, there was no evidence that these biases affected judges when their race was explicitly stated. This can be attributed to judges’ experience, education, and motivation to be fair. Since jurors lack most of these qualifications, they are more likely to act upon their implicit biases. Hence, judges can mitigate the effect of biases when they are nudged to use their System 2.

Policy Suggestions

The first step for having effective policies is acknowledging their need. Like many professionals, judges tend to dismiss the evidence showing the influence of biases (Berthet, 2022; Dhami & Belton, 2017; Kukucka et al., 2017). When judges are also willing to work towards mitigation, more concrete steps can be taken.

Judges show effort, but even professionals are biased in their decisions (Chen et al., 2016). Thus, their efforts often fall short. Additionally, these efforts only come into play with well-known biases, like racial bias. For less obvious biases, training can be given to judges to increase their knowledge, which leads them to be cautious when they come across them (Wistrich & Rachlinski, 2017). The degree to which training would be effective and applicable to jurors is not clear, as they are less likely to act against their biases even when they are aware. For jurors, instructions that are clearly written and mentioned multiple times during trials for inadmissible evidence are effective (Lieberman & Sales, 1997; Nietzel et al., 1999; Steblay et al., 2006). Instructions should also include guidelines on evaluating evidence to mitigate effects of confirmation bias and reduce the cognitive constraint that leads jurors to use heuristics with evidence. Systems can be used to give instructions to be given in a standardized way multiple times during trials.

As inadmissible evidence is responsible for leading to the usage of biases for both judges and jurors, improvements should be made to the evaluation process. The effects could be mitigated by a system that decides on the admissibility of the evidence before judges to reduce their effects. As Steblay et al. (2006) show, the best way to mitigate the effect of inadmissible evidence is not to mention it at all. As discussed above, judges sometimes keep the information about inadmissible evidence from jurors. This is effective, but only for jurors. A system that evaluates the evidence before the judges would also be able to protect them. These systems examine the evidence first according to the legal principles and decide on its admissibility. Judges and jurors would not see the evidence deemed inadmissible. This way of evaluating evidence with a reliance on technology could result in concerns. For this reason, the party whose evidence was deemed inadmissible would have the option to appeal, which would be seen by a law clerk or a judge who is not on the case. To increase the efficiency of the system, departments could be created that would only be responsible for reviewing the appealed evidence.

A similarly structured system could be effective for anchoring. Extreme settlement offers and sentencing demands would be entered into the system. Instead of informing judges and jurors about these anchors, the system could provide guidelines according to legal principles on deciding damage awards and sentencing times. Further, for damage awards, mathematical computations can be carried out by the system specifically for these cases. Of course, judges would not be bound by the estimate of the system. However, the unbiased estimates created by the system would be better predictions than any anchor resulting from extreme offers. This system is different than damage caps and mandatory sentencing times because it would be able to account for the different aspects of each case. Hence, having reliable anchors for each case could be useful to mitigate the effects of extreme offers and artificial caps.

One other system to reduce the effects is an occasional data analysis system on decision patterns of judges. This could uncover the prevailing biases in judges. Additionally, systems would increase the sense of accountability. The motivation for making fair decisions would amplify, leading them to make conscious efforts for accurate decisions.

A more general way to mitigate the influence of biases would work to combat the factors leading to these shortcuts. As the time pressure increases, the gain from time and effort with the use of heuristics would also increase. Thus, the number of ongoing cases a judge can have could be limited to lead them to spend more time with each case.

Conclusion

All the biases examined above have one common aspect: leading to an unstable legal system, which can lead to diminishing levels of public trust. From biased jury selections to investigations that lead to only pursuing incriminating evidence, a case is prone to biases at many turns. For instance, the Central Park Five case is cited as one of the most notable examples of cognitive bias because the wrongful convictions of five teenage men were caused by the investigation that was guided by the initial beliefs of the police and the prosecution. (The Edelstein Firm, 2023). It is possible to come across cases in civil law as well, like Gregory v. Chohan in which the plaintiff claimed 71 million dollars in compensation (Gregory v. Chohan, 2023). This case reflected how anchoring can be used in unsubstantiated ways that affect final verdicts (Sampson et al., 2024). Improvements in the field are crucial for making legal systems stable and effective. Only the right policies can ensure justice is served regardless of race, gender, or the ungrounded demands of a party. If this cannot be achieved, then the trust between the public and the legal system can lead to a separation that further impairs justice.

Even though this paper only focused on three biases, others, like hindsight bias and status quo bias, can influence judicial decisions (Teichman & Zamir, 2021). Initial decisions of judges and jurors lead to confirmation bias, affecting the view of subsequent evidence. Views are distorted to prove them right. Inadmissible evidence, not being fully omitted can trigger confirmation bias. Judges have the option not to inform jurors about inadmissible evidence. However, judges themselves are not protected. Hence, systems that decide on the admissibility of evidence before the judges could minimize the effects on all decision-making bodies.

Anchors -extreme settlement offers, sentencing demands, damage caps, and mandatory sentencing times- affect both judges and jurors. Mandatory sentencing times and damage caps aim to create stability. However, these policies do not consider the differences of each case. They end up creating more systematic errors without addressing the problem. Thus, to truly counteract the effects, a system that dismisses extreme offers and performs mathematical computations tailored to each case could be implemented.

Finally, implicit bias is prevailing across both judicial decision-making bodies. Even though judges can mitigate them to some extent, judges are not always aware of them. Mandatory training can be given to inform them about lesser-known biases to increase mitigation efforts.