Bias, Brains, and the Bounds of Fairness in the Courtroom

Eleanor Chou

Edited by Samantha Tonini, Danush Gade, Mac Kang, and Sahith Mocharla

Impartiality is fundamental to any court of law. Accordingly, standards of objectivity—such as the reasonable person standard—are deeply ingrained in the judicial process. However, longstanding periods of prejudice, bias, and hate have made the pursuit of impartiality in the legal system an ongoing and iterative process. The persistent pursuit of objectivity has led many to look at using science, specifically neuroscience, to demonstrate biases in decision-making. Breakthroughs in understanding the neural functions underlying human decision-making have revealed areas of bias, making it difficult to assess the objectivity of juror rulings. Although legal safeguards like the Sixth Amendment’s guarantee of the right to an impartial jury exist, emerging research within the field of neuroscience suggests biases persist beyond conscious attempts at objective decision-making [1]. These findings bring to light the flaws in the notion of the impartial juror, highlighting a critical limitation in the current understanding of courtroom objectivity. 

Courts have implemented reforms to mitigate biases and ensure greater fairness in criminal trials. One of the primary steps is voir dire, or the jury selection process, which has been refined by the courts over time. During voir dire, potential jurors fill out questionnaires and are questioned by attorneys and the court about their biases and connections to the case. The Sixth Amendment of the United States Constitution guaranteed the right to an impartial jury, ensuring that defendants are held accountable by the people. 

Over the years, this process has evolved to include additional protections. In the ruling in Batson v. Kentucky (1986), the Supreme Court prohibited the use of peremptory challenges to exclude jurors based on race. Before Batson, it was common for lawyers to remove Black jurors from pools in cases with Black defendants as a matter of course [5]. The Court’s ruling established the Batson Test, a procedure that allows lawyers to object to juror exclusions that seem to be racially motivated. Subsequent cases have expanded protections in the process of voir dire. In J.E.B v. Alabama ex rel. T.B. (1994), the Supreme Court ruled that excluding jurors solely on the grounds of sex violates the Equal Protection Clause [6]. Both Batson and J.E.B have solidified fair, undiscriminatory selection procedures to safeguard the impartiality of jurors in criminal trials. 

Alongside selection procedures, modern courts also aim to promote impartiality through jury instructions. In criminal trials, the judge guides the jury through deliberations by explaining relevant laws, definitions, and standards of proof [7]. The specific details of the instruction given to the jury vary by state and case, but the majority emphasize the importance of objectivity in evaluation and thorough consideration of the evidence presented.

Though considerable progress has been made to create an objective system of decision-making within courts, the complexity of human thought makes it difficult for courts to account for every case. Emerging studies conducted on subconscious racial and socioeconomic biases reveal that many internal predispositions are undetectable by modern questionnaires and often even by the individual themselves [8]. Therefore, while the existing voir dire process ensures a thorough screening of an individual’s background, existing implicit prejudice makes it clear that more action must be taken to address the nuanced nature of human decision-making in the criminal justice system. 

People are not inherently free from bias. In a collaborative study conducted by researchers at MIT, New York University, and Yale University, White participants were presented with a series of Black and White male faces while their brain activity was recorded with fMRI technology [2]. After these tests, the researchers measured participants’ unconscious biases using the Implicit Association Test (IAT), their startle responses, and self-reported biases. Overall, the study found that there was greater amygdala activation—a response associated with emotional salience and fear—when the participants were presented with unfamiliar Black faces. These results correlated with both IAT results and startle responses but not with self-reported biases, suggesting that amygdala activity reflected only the participant's unconscious responses. As a whole, these findings grounded the existence of subconscious biases in scientific literature, challenging the assumption of the unbiased juror. 

While this study shows neural evidence of subconscious racial bias, factors other than race often influence people’s decision-making. One experiment examined 1,074 participants’ judgments of blame, guilt, and punishment for Black and White juvenile defendants across low to high socioeconomic backgrounds [3]. The result of this study showed that Black juveniles with a high socioeconomic status were judged less harshly than average. In addition, White juveniles with a high socioeconomic status were judged more harshly when compared to those with a low socioeconomic status. These findings demonstrate that understanding the interaction of various social factors is imperative in evaluating the reliability of juror decision-making. Another study, conducted at Duke University Medical Center, examined the influence of crime severity on juror decisions. Participants were presented with hypothetical scenarios that differed in crime severity, and were instructed to rate the strength of the evidence and recommend the level of punishment [4]. The researchers found that participants’ decisions in this study were significantly influenced by the severity of the crime. Even when there remained a similar amount of evidence across the scenarios, the jurors were more likely to assume guilt and recommend harsher punishments when the crimes were more severe. The results illustrate that biases are not limited to just racial or contextual factors and that they may also arise as a result of the type of crime committed.

Emerging reforms seek to address subconscious bias in the judicial process. One such example is the double-blind system. This system most commonly appears in eyewitness identification procedures, where criminal investigators and administrators of photo arrays conducting suspect lineups are unaware of which individual in the pool is the suspect. This process thus minimizes the risk of unintentional cues or subconscious signaling by law enforcement, improving the fairness and reliability of eyewitness evidence, which has been historically unreliable [9]. While the double-blind system is primarily implemented in criminal investigations, there has been increased discussion about its potential use in the trial process. An article submitted to the Indiana Journal of Law and Social Equality recommends an immediate implementation of the double-blind system. Under this model, the defendant is obscured from view—aside from when they testify by right—limiting the jury’s exposure to information like race, appearance, and socioeconomic status to ensure that jurors do not form opinions based on unconscious biases [10]. The defendant would be housed in a separate room—such as a jail conference room or an attorney-client meeting room—with an audio channel connecting to the defense counsel and a CCTV allowing live interaction with the witness. Additionally, their name would also be anonymized to initials, as courts do in cases involving minors. Williams argues that the defendant’s right to be present derives from Due Process and exists today mainly so the defendant can participate in cross-examination, not necessarily for the jury to view the defendant. Therefore, implementing double blinding can be done without violating the defendant's constitutional rights. Moreover, limiting the jury’s exposure to certain traits of the defendant may ensure that potential areas of bias, such as those based on variables like familiarity or race, do not cause skewed decisions. As discussed previously, subconscious biases persist through even the belief of impartiality, and there is potential for bias to exist in a multitude of different situations. Since bias is inevitable, restricting the juror’s view of the case to information pertinent to only the facts of the case may promote fairer, evidence-based decision-making rather than decisions based on factors irrelevant to the legal arguments surrounding a criminal case. However, the double-blind system is not a panacea for a broader problem of bias; it does not eliminate biases related to non-visual factors like crime location, language, or witness demographics. 

Another recent development is the increasing use of Artificial Intelligence (AI), where its growing usage has prompted discussion about its potential role in courts. Large Language Models (LLMs) like OpenAI’s ChatGPT and Anthropic’s Claude are capable of processing tremendous amounts of data and can therefore process legal arguments and facts in ways similar to and sometimes exceeding a human juror. This capacity has led some AI-forward proponents to consider whether AI could assist in jury deliberations. However, although AI is modernly advanced enough that it may be possible to supplement court proceedings, its use raises critical arguments about whether it truly supports objectivity or simply reinforces biases embedded in its ‘black-box’ algorithms. A study conducted at the University of North Carolina at Chapel Hill School of Law featured a simulated trial replicating a real-world case where LLMs like Grok, ChatGPT, and Claude served as the jury [11]. The case involved a teenager accused of aiding a robbery by standing behind the individual who was demanding money from the victim. While the original jury found the defendant guilty, the AI models unanimously returned a not guilty verdict. The original human jury concluded that the defendant's presence intimidated the victim and that his position behind the robber proved his intent to assist in the crime. In comparison, although presented with the same facts and argumentation, the AI models argued that the defendant took no real action and therefore there was insufficient evidence to prove intent. While the AI models proved their capacity for structured reasoning and decision-making, the difference between AI and human rule suggested that AIs process information in ways that may lack the real-world context and social intuition that humans bring to deliberations. The human decision concluded that the defendant’s presence intimidated the victim and that his position behind the robber proved his intent to assist in the crime, while the AI models argued that no actual action was taken by the defendant and therefore there was insufficient evidence to prove intent. Broadly speaking, this study underscores both the potential for AI implementation and the existing differences between it and human jurors. Furthermore, while the implementation of AI may eliminate common explicit or understood biases that currently influence the decisions of jurors, it may introduce different forms of implicit human bias. From the data used to train the models to the perspectives of the developers themselves, such biases have the potential to shape outcomes through the AI algorithm itself.

Ultimately, both the double-blind system and AI integration are potential solutions that address subconscious human biases. While the double-blind system offers a viable solution that may be implemented in courts today, further consideration of solutions pertaining to AI integration is required to ensure that its applications do not introduce new threats to fairness in juror decisions for criminal trials.

As seen in studies of racial, crime-type, and socioeconomic perceptions, research strongly suggests that subconscious biases are pervasive. While the criminal justice system has implemented safeguards against bias, like voir dire and the importance of a jury, these measures often fail to address the more deeply rooted subconscious biases revealed by scientific findings. As a result, potential developments like the double-blind system and AI courtroom integration should be considered, as they have the potential to reduce the influence of human biases. Ultimately, no single solution will fully ensure impartiality. Courts must continue to consider and refine potential tools that address bias from multiple angles while protecting the integrity of trials and the rights of defendants. As neuroscience research has continually revealed the depth of biases, the courts must adopt solutions that go beyond conscious deliberation alone, as fairness is not a fixed standard to be reached, but rather an iterative pursuit.

---

[1] Sixth Amendment, LII/LEGAL INFORMATION INSTITUTE,

https://www.law.cornell.edu/constitution/sixth_amendment (last visited Mar. 24, 2026).

[2] Elizabeth A. Phelps et al., Performance on Indirect Measures of Race Evaluation Predicts Amygdala Activity, 12 J. Cogn. Neurosci. 729 (2000).

[3] Elizabeth A. Gilbert, Alexander D. Guinn & N. Dickon Reppucci, Intersection of Race and Socio-Economic Status on Criminal Judgments: High Status Reduces Blame for Black Juveniles but Increases Blame for White Juveniles, 2 Front. Soc. Psychol. (2024).

[4] Jaime J. Castrellon et al., Social Cognitive Processes Explain Bias in Juror Decisions, 18 Soc. Cogn. Affect. Neurosci. 1 (2023). 

[5] Batson v. Kentucky, 476 U.S. 79 (1986).

[6] J.E.B. v. Alabama ex rel. T.B., 511 U.S. 127 (1994).

[7] Instructions to the Jury, AMERICAN BAR ASSOCIATION (Sep. 9, 2019), https://www.americanbar.org/groups/public_education/resources/law_related_education_network/how_courts_work/juryinstruct/.

[8] See [3].

[9] Eyewitness Misidentification, THE EXONERATION PROJECT, https://www.exonerationproject.org/issues/eyewitness-misidentification/#:~:text=Reducing%20False%20Identifications%20in%20Lineups,the%20importance%20of%20blind%20administration (last visited Mar. 24, 2026).

[10] Stanley P. Williams Jr., Double-Blind Justice: A Scientific Solution to Criminal Bias in the Courtroom, 6 Ind. J.L. & Soc. Equal. 50 (2018). 

[11] AI Jury Finds Teen Not Guilty in Mock Trial, UNC SCHOOL OF LAW (Nov. 4, 2025), https://law.unc.edu/news/2025/11/ai-jury-finds-teen-not-guilty-in-mock-trial/.