The Evolution and Current Status of Ocular Deception Detection – September 22, 2026

Deception Detection Tools Advance, but Outcomes Still Depend on Human Judgment and Critical Thinking

Law enforcement, investigative, and intelligence analysis organizations are testing artificial intelligence–driven deception detection tools as the need to verify identity, intent, and credibility expands across physical and digital environments.

A recent BizTechReports executive vidcast interview with former Federal Bureau of Investigation professionals Michael Howell and Melody Bruns indicates that while these technologies are advancing rapidly, their effectiveness remains closely tied to the judgment, experience, and analytical discipline of the people using them.

The discussion focused in part on research conducted through Texas A&M University, where Howell, now an associate professor at the Bush School, and Bruns, founder of PathLink Consulting, evaluated ocular deception detection systems. These tools attempt to infer deception by measuring physiological responses in the eyes, including blink rate, pupil dilation, and fixation patterns.

Many of these systems are built on the assumption that deception increases cognitive load, which in turn produces measurable biological signals. Advances in artificial intelligence and computer vision have made it possible to capture and analyze these signals at scale.

Despite these advances, the results produced by these systems remain highly dependent on how they are applied in practice.

Human Judgment as the Governing Factor

The effectiveness of these technologies depends less on the sophistication of the tools themselves and more on how they are used.

Both Howell and Bruns emphasized that emerging detection systems are best understood as force multipliers for trained practitioners rather than standalone solutions. Their value is realized only when applied by investigators and analysts with strong critical thinking skills, domain expertise, and the ability to validate findings through corroboration.

Howell described rapport-based interviewing techniques developed over decades of investigative work, noting that the most reliable way to assess whether someone is being truthful is to engage them in detailed conversation and look for information that can be independently verified.

This human-centered approach reflects a broader shift away from earlier interrogation models that relied heavily on eliciting confessions. Modern practices prioritize building trust, gathering verifiable information, and testing narratives against independent evidence.

The research conducted by Howell and Bruns underscores how this dynamic plays out when these tools are tested in practice.

Study Findings Highlight Variability and Risk

Their study analyzed the results of 420 subjects. Separately, their review of other ocular deception studies found accuracy rates ranging from 68 to 95 percent depending on methodology and conditions. That variability introduces uncertainty in environments where decisions may affect employment, legal outcomes, or public safety.

Bruns noted that the foundational theory behind many of these tools remains unresolved. Her research was unable to identify any study confirming that lying consistently increases cognitive load, and she raised the question of why these devices appear to produce results in some cases if the underlying theory has not been validated.

These findings informed a more cautious assessment of the technology's current readiness.

Howell emphasized that claims made for these technologies must be supported by evidence, and that much of the existing science does not support the reliability of the behavioral cues these tools are designed to measure.

He noted that the assumption that lying consistently increases cognitive load does not hold in all cases. Recalling a detailed memory from a decade ago, for example, may require more cognitive effort than fabricating a response, underscoring the difficulty of using physiological signals alone to distinguish truth from deception.

That variability reflects a broader challenge for organizations evaluating AI-driven decision tools.

The gap between vendor claims and validated evidence reflects a broader issue in the adoption of AI-driven solutions. Claims of accuracy and efficiency often outpace independent validation. In the case of deception detection, that gap introduces operational risk, particularly when tools are used without sufficient human oversight.

Cognitive bias further complicates the equation. Howell pointed to confirmation bias and availability bias as persistent influences on how investigators interpret both evidence and technology outputs. He stressed that experience and training are essential for recognizing those biases and being willing to reassess conclusions when the evidence warrants it.

Bruns described this as a structural gap in how many organizations approach decision-making. While intelligence and law enforcement agencies often rely on formal analytical frameworks, many enterprises lack comparable capabilities. She noted that structured methods are necessary for genuinely challenging assumptions and evaluating sources, and that critical thinking is not something most people develop automatically without those frameworks.

This gap becomes more consequential as AI tools are introduced into areas such as hiring, compliance, and internal investigations. While certain deception detection technologies face regulatory constraints in the United States, similar capabilities are being explored in other jurisdictions and contexts.

Evaluation and Investment Considerations

Organizations evaluating these tools must weigh financial investment against demonstrated effectiveness, while also accounting for the risks of false positives and false negatives. Bruns stressed the importance of reviewing multiple independent studies rather than relying on a single source, noting that a single study can produce an incomplete or misleading picture of how a technology actually performs.

Collaborative evaluation models may help mitigate these risks. Howell pointed to task force structures used in law enforcement, where multiple agencies contribute expertise and share insights, enabling broader testing and more balanced assessments of emerging technologies.

Ongoing innovation is expected to expand beyond ocular signals to include full-body behavioral analysis and multimodal detection systems. Advances in artificial intelligence are likely to accelerate development, particularly as demand grows for tools that can operate in digital and remote environments.

Adoption will depend on whether these technologies can demonstrate consistent and reproducible results under real-world conditions and whether organizations can integrate them effectively into human-led processes.

Howell acknowledged that AI represents a significant and promising technology, but cautioned that it is not infallible. He emphasized that when decisions affect people's lives, the standard for accuracy must be high.

The challenge for decision-makers is not simply adopting new tools, but ensuring they are placed in the hands of practitioners capable of interpreting results, challenging assumptions, and grounding outcomes in verifiable evidence. In deception detection, that balance will determine whether these technologies enhance judgment or introduce new uncertainty.

###