Technologies in the twilight zone: early lie detectors, machine learning and reformist legal realism

This legal-historical analysis examines the introduction of polygraph technology in US courts circa 1920s-1940s and draws parallels to contemporary machine learning deployment in criminal justice. The paper argues that both technologies share fundamental similarities in attempting to predict individual behavior based on group patterns, and that early polygraph regulation failures offer critical lessons for AI governance today.

Historical legal analysis comparing early polygraph case law (particularly Frye v. United States 1923 and People v Kenny 1938) with contemporary machine learning litigation, examining judicial reasoning, legal realist influences, and regulatory frameworks across approximately 100 years of technological deployment in criminal justice contexts.

This study uses early polygraph litigation as a historical lens to understand contemporary artificial intelligence deployment in criminal justice. The paper examines how polygraph technology was introduced in American courts, analyzed through the Frye standard for scientific admissibility, and ultimately excluded from courtrooms while flourishing in investigative, employment, and interrogation contexts. The research reveals a bifurcated system where courts applied rigorous scientific standards while investigators and practitioners adopted the technology based on perceived utility and reformist legal realist principles. By drawing parallels between polygraph and machine learning, the study identifies how 'public good' aims can motivate technology deployment despite unresolved questions about scientific validity and compatibility with fundamental legal principles such as the Fifth Amendment privilege and jury decision-making authority.

This historical analysis provides critical guidance for regulators, developers, and practitioners considering machine learning deployment in criminal justice by demonstrating how reformist impulses can circumvent formal legal standards. The study suggests that without vigilant legal oversight, algorithmic decision-making risks replicating the dual-system phenomenon that enabled questionable technologies to operate outside courtroom constraints while influencing investigative and prosecutorial decisions.

Background & Context

This study represents a novel approach to understanding contemporary debates about artificial intelligence by examining the century-old introduction of polygraph technology in American courts. The paper argues that developers, practitioners, and regulators could usefully look back to the similar arguments made when polygraph machines were first introduced in the United States, noting that both polygraphs and machine learning attempt to predict something about a person based on how others have behaved. This historical lens provides critical insights into current struggles with algorithmic decision-making in criminal justice.

The paper examines the development of the polygraph within the justice system—both in courts and during criminal investigations—and considers the use of lie detectors in other justice contexts. The central thesis is that the promotion of lie detectors supported a reforming legal realist approach, something that continues today in debates over the deployment of machine learning where 'public good' aims are in play, raising fundamental questions about how principles of the rule of law can be upheld.

Research Design & Methodology

This is a historical legal analysis employing comparative methodology to draw parallels between early polygraph litigation and contemporary machine learning deployment. The research examines archival legal sources including the landmark Frye v. United States (1923) case, the 1938 People v Kenny trial where Raymond Kenny faced thirty to sixty years as a second offender, and appellate court decisions from the 1930s and 1940s.

The methodology involves detailed analysis of judicial reasoning, examination of contemporaneous legal commentary, and investigation of how lie detector technology was deployed outside courtroom contexts despite legal restrictions. Key analytical elements include:

• Historical case law review examining admissibility standards
• Analysis of the Frye "general acceptance" test and its application
• Investigation of reformist legal realist influences on technology adoption
• Parallel examination of modern machine learning deployment patterns
• Study of European Court of Human Rights decisions on polygraph use

The paper identifies how lie detectors were deemed not sufficiently established to gain general acceptance by experts, nor had their use moved out of the experimental 'twilight zone' to a 'demonstrable' stage—the phrase that gives the study its evocative title.

Results & Key Findings

The paper reveals a striking dual-system phenomenon: Two systems ran in parallel—one for criminal court proceedings requiring scientifically valid expert testimony, and one running parallel to the court system without such constraints, based on realist and utilitarian principles. This bifurcation enabled polygraph use to flourish in employment screening, criminal investigation, and suspect interrogation even as courts consistently excluded the evidence.

Key findings about early polygraph litigation include:

• US appellate courts in the 1930s and 1940s fairly consistently rejected lie detector evidence based on the Frye standard, though courts recognized the potential 'utility' of the polygraph particularly for investigation of crime, while rejecting its place in the courtroom
• The 1938 People v Kenny case involved conflicting eyewitness testimony, with lie detector results offered to demonstrate Kenny's innocence, and expert testimony provided by Father Walter Summers, a Jesuit and Head of Psychology at Fordham University
• Courts expressed nervousness not only about scientific validity, but about the test's potential impact on established legal norms such as the Fifth Amendment privilege against self-incrimination and the jury's role in determining credibility
• Use of the lie detector outside court forged ahead—for assessment of evidence, vetting potential employees, fraud investigations, and testing spousal fidelity

The machine learning parallel is equally striking: In the challenge by Eric Loomis to the use of the COMPAS risk assessment tool in his sentencing, the Wisconsin Supreme Court stated that the sentencing court 'would have imposed the exact same sentence without it', suggesting similar patterns of acceptance without rigorous scrutiny.

Discussion & Significance

The paper's significance lies in exposing how "public good" justifications enable deployment of contested technologies outside traditional evidentiary frameworks. Machine learning tools are consistently subjective—they can process only the data provided to them, historical data which will contain elements of subjectivity, and in the criminal justice context, tools making predictions based on historical offender data will be affected by past arrest history, force targeting decisions, social trends and prioritisation of certain offences.

Polygraphs, too, reconstruct individuals based on fragments of data—physiological responses—and place them along a line between truthful and deceptive. This fundamental similarity in predictive methodology underscores the relevance of historical lessons. The paper demonstrates that while polygraphs assume deception correlates with physiological responses over a threshold, machine learning assumes all relevant factors can be represented as data, measured, and analysed with significantly accurate predictive power—yet the deployment of both technologies in real contexts remains in the 'twilight' zone between experimental and demonstrable stages.

The legal realist framework is crucial: reformist impulses toward efficiency and scientific progress can override procedural safeguards when technologies operate outside formal court processes. This pattern suggests contemporary machine learning in policing, sentencing, and security screening may replicate the polygraph's trajectory—limited courtroom acceptance but expansive deployment in investigative and administrative contexts.

Limitations & Considerations

The author acknowledges that caution is required regarding generalisations based upon past legal treatment in one jurisdiction of one particular technology deployed within a social context very different from today's, but argues that sufficient parallels can be drawn between the two technologies and the aims of those who would promote their use to justify learning from the early polygraph experience.

This is a theoretical and historical analysis rather than empirical research, limiting its ability to directly test causal mechanisms or quantify impacts. The paper does not provide specific accuracy data or statistical comparisons, instead focusing on legal-historical patterns and regulatory frameworks. The single-jurisdiction focus (primarily United States case law) may not fully capture international variations in technological regulation or legal traditions regarding scientific evidence.

Practical Applications

For polygraph examiners and policymakers, this study provides crucial context about why polygraph evidence remains contested despite over a century of use. The paper proposes a number of regulatory solutions informed by the early lie detector experience, including preregistration of machine learning research design and precautionary legal frameworks around artificial intelligence, specifically the generalisation for AI of regulatory principles found in bodies of law relating to admission of expert evidence.

The analysis warns against accepting technologies based on utilitarian promises rather than evidentiary validity. For contemporary criminal justice practitioners, the lesson is clear: technologies deployed outside courtroom scrutiny can become entrenched despite lacking scientific consensus, creating parallel systems that undermine procedural fairness. The paper suggests that proper regulation requires preventing the bifurcation between formal legal standards and informal operational practices that characterized the polygraph's development and that now threatens to characterize machine learning's expansion in criminal justice.