Table of Contents Introduction I. The Courts' Unprincipled Treatment of Profile Evidence A. Profile Evidence B. The Courts' Analysis of Profile Evidence: Problems Beyond Unpredictability 1. The influence of bias based on a defendant's personal characteristics 2. Incorrect findings of harmless error 3. Illogical rules under the common law II. Predictive Character Evidence A. Accuracy and the Bias-Variance Tradeoff B. The Meaning and Properties of Ordinary Character Evidence 1. The rule against character evidence 2. Clarifying the meaning of character evidence 3. Character evidence as a bias-variance tradeoff C. The Meaning and Properties of Predictive Character Evidence III. Implications A. Descriptive Implications: Explaining Unpredictability and Confusion in Admissibility Decisions Surrounding Profile Evidence B. Prescriptive Implications: Applying the Rule Against Character Evidence to Assess the Admissibility of Profile Evidence C. Legal Outcomes Based on Race and Other Personal Characteristics D. Implications for the Use of Artificial Intelligence and Other Technologies that Rely on Extrapolations from Population-Level Data Conclusion
Introduction
Criminal profiling aims to connect a specific defendant to a crime or act by describing the probable features or conduct of a type of criminal offender or actor. (1) Its use is common (although controversial) in police investigations--for example, to detect securities fraud or investigate a murder. (2) However, when a court admits profiling as evidence, it takes on a particularly pernicious character.
Consider, for example, the case of People v. Masters, in which the police relied on the guidance of a profiling expert (a forensic psychologist) to identify probable features of the perpetrator of a murder and thereby arrest a suspect. (3) At trial, the same expert testified regarding the probable features of the perpetrator--for example, the likely existence of a triggering event in the perpetrator's life, prior fantasies about committing the murder, preference for a surprise attack, and a link to the perpetrator's past. (4) He then connected these features to the defendant. (5) The trial court admitted this evidence, and the prosecutor relied heavily on it to prove the government's case and obtain a conviction. (6) The court of appeals and the state supreme court later upheld the admission. (7) Approximately nine years after the defendant's conviction and imprisonment, exculpating DNA evidence came to light, leading to the defendant's exoneration. (8) The case was built on profile evidence, and it resulted in a false conviction.
Profile evidence is not limited to murder cases. It is common in cases involving drug trafficking, domestic abuse, fraud, and many other offenses. (9) For example, a drug-trafficking case may involve testimony that the defendant fits the profile of a drug courier because he arrived in the United States from Colombia and paid for his ticket in cash. (10) A domestic violence case may involve evidence of a defendant's behaviors to suggest that, based on data from a broader population, the defendant was an abuser who was more likely to become a murderer. (11)
Indeed, the United States Supreme Court recently highlighted the significance of profile evidence in Diaz v. United States, a drug-trafficking case in which the prosecutor sought to prove the defendant knowingly possessed narcotics by introducing profile evidence indicating that it is rare for drug couriers to be unknowing participants in their trafficking activities. (12) Notwithstanding the defendant's objections, the Ninth Circuit upheld the district court's admission of this evidence and affirmed the defendant's conviction. (13) The Supreme Court granted certiorari but ultimately affirmed the judgment of the Ninth Circuit. (14)
Profile evidence is a type of "base rate evidence"--a broad category of evidence that has become increasingly common in civil and criminal cases. (15) In statistics, a base rate indicates the prevalence of a feature in a population. (16) For example, assume that a thirty-year-old patient who has not recently traveled abroad enters a New York hospital with a fever, headache, chills, and fatigue. A doctor is likely to be influenced by the base rates of diseases involving these symptoms in the relevant "reference population"--a comparison population, such as New York's population of thirty-year-olds. These base rates are likely to lead the doctor to conclude that the patient has a cold, the flu, or COVID-19 rather than malaria, which has a very low base rate, or prevalence, in the reference population. (17) Therefore, base rates may lead the doctor to test for certain diseases and not for others.
Base rates are relevant to legal cases for the same reason they are fundamental in statistics and medicine: They suggest the probability of an occurrence, feature, or act based on its prevalence in a broader population. They appear in various forms and can be crucial to achieving an accurate verdict. For example, base rate evidence may appear in the form of DNA evidence offered to connect a defendant to a crime based on the prevalence of a DNA profile in a population, or it may appear as evidence that 80% of a certain population committed a particular act (such as trespass) to suggest that a specific defendant who is member of the population committed the same act. (18) But while base rate evidence has the potential to be highly impactful on the outcome of a case, (19) admissibility decisions surrounding this evidence are at best unprincipled and unpredictable. (20) Arguably, many of these decisions are altogether incorrect.
The courts' treatment of profile evidence is particularly problematic. There is little predictability in how a court will decide the admissibility of this evidence--or even which rule it will apply in the first instance. (21) While forensic evidence and many other forms of base rate evidence are commonplace and often beneficial, profile evidence is different, and although courts seem to recognize its problematic nature, they seem confused as to how to analyze it or why it should be excluded--as though courts are missing an important factor in their analysis.
In this Article, I argue that courts are in fact missing a central component of the analysis. I trace the unpredictability and incoherence surrounding the admissibility of profile evidence to a misunderstanding about the nature of this evidence and what separates helpful from harmful forms of base rate evidence. Specifically, I show that although some forms of base rate evidence are desirable and even critical to achieving an accurate case outcome, profile evidence often constitutes unrecognized character evidence, a type of evidence that is highly prejudicial and prohibited by federal and state evidentiary rules.
In criminal and civil trials, Federal Rule of Evidence (FRE) 404 and analogous state rules prohibit character evidence. (22) Character evidence is evidence of a person's propensities or other acts (acts not at issue in a case) offered to suggest that the person acted in conformity with their propensities or prior acts on the occasion in question. (23) For example, a prosecutor generally may not introduce evidence that a defendant committed two prior robberies to prove that the defendant acted in accordance with a propensity to commit robbery on the occasion in question. (24) Although this evidence can be quite probative (and, as such, is sometimes admitted under an expanding set of exceptions (25), the general ban on character evidence is based on the profound prejudicial impact that such evidence has on the outcome of a case. (26) As one court has stated, "an obvious truth is that once prior convictions are introduced the trial is, for all practical purposes, completed and the guilty outcome follows as a mere formality." (27)
To demonstrate that a significant subset of base rate evidence--and profile evidence in particular--constitutes prohibited character evidence, and to describe the relationship between base rate evidence and character evidence, I begin by clarifying the meaning, properties, and effects of character evidence. I then draw on an area of statistics called Bayesian inference to define a concept that I call predictive character evidence. Predictive character evidence is base rate evidence that fulfills four criteria summarized by the acronym PBIC: predictive, behavioral, individualized, and characterizing. It describes evidence regarding a behavioral propensity of a population to suggest that an individual member of the population acted in accordance with this propensity. Predictive character evidence is a form of base rate evidence that involves behavioral stereotyping, and it is surprisingly common in civil and criminal trials. As I show below, however, it relies on impermissible character reasoning. This is because inferring a defendant's behavior from a population's propensity requires an inference that the defendant acted in accordance with a character or propensity on the occasion in question. (28) As such, predictive character evidence is the unrecognized link between base rate evidence and ordinary impermissible character evidence: Predictive character evidence involves character reasoning and is therefore character evidence; however, it is a special type of character evidence that involves base rate evidence that satisfies the PBIC criteria. That is, when base rate evidence fulfills the PBIC criteria, it constitutes predictive character evidence and is impermissible under FRE 404 and analogous state rules. (29)
Finally, after carefully defining predictive character evidence and applying the PBIC criteria to delineate precisely when base rate evidence constitutes impermissible character evidence, I show that this relationship resolves significant problems in evidence law. Specifically, my analysis answers two important questions--one...
