Polygraph Technology: 1990s vs 2020s — How Lie Detectors Evolved

From Dr. MacKenzie to Leonard Keeler

The polygraph's story begins long before the digital age. In 1892, Dr. James MacKenzie published "The Study of the Pulse, Arterial, Venous and Hepatic and the Movements of the Heart" and introduced his polygraph device for studying cardiological rhythms [1]Verified LieDetectorTest.com — Dr. James MacKenzie, Cardiology & The Polygraph
Confirms MacKenzie published in 1892 and introduced his polygraph device for studying cardiological rhythms. Designed primarily for diagnosing heart conditions and recording cardiovascular activity in clinical cardiology, MacKenzie's instrument was first demonstrated at the BMA Toronto meeting in 1906 and described in the BMJ in 1908, with production models becoming commercially available that same year [2]Verified LITFL Medical Eponym Library — James Mackenzie
Confirms Mackenzie's polygraph was first demonstrated at BMA Toronto meeting (1906) and described in BMJ (1908). His device allowed simultaneous recording of jugular, radial, and apex beats, laying the conceptual groundwork for measuring physiological responses tied to human emotions and stress [3]Verified Wikipedia — James Mackenzie (cardiologist)
Confirms Mackenzie devised a polygraph for simultaneous recording of arterial and venous pulses.

Across the Atlantic, Dr. William Moulton Marston invented the Systolic Blood Pressure Test in 1915. Marston's work focused on measuring changes in blood pressure as a potential indicator of deception — a revolutionary concept for its time. His research directly inspired John A. Larson, who in 1921 debuted his creation — a portable instrument he informally called "The Sphyggy" — which simultaneously measured blood pressure, pulse rate, and respiration [4]Verified LieDetectorTest.com — What Is the Police Polygraph
Confirms Larson debuted 'The Sphyggy' in 1921, which simultaneously measured blood pressure, pulse rate, and respiration. The press nicknamed it the "Sphyggy" — a shortening of "Sphygmomanometer" — because reporters struggled to spell or pronounce the full technical term [5]Verified LieDetectorTest.com — What Was the First Lie Detector Machine
Confirms nickname 'Sphyggy' came from reporters struggling with 'Sphygmomanometer,' and details on Keeler's Emotograph in 1925. Larson's device created the multi-channel approach that still defines polygraph testing today, and his instrument provided the evidence allowing police forces across America to remove thousands of criminals from the streets in the 1920s and 1930s [6]Verified LieDetectorTest.com — John Larson: The Innovator of Polygraph Science
Confirms Larson's device helped remove thousands of criminals from streets in 1920s and 1930s.

It was Leonard Keeler, however, who transformed the polygraph from a laboratory curiosity into a practical law enforcement tool. In 1925, Keeler debuted his new device, the Emotograph, becoming the first American to hold intellectual property for a polygraph machine [7]Verified LieDetectorTest.com — Leonard Keeler: The Father of the Modern Polygraph
Confirms Keeler named his device the Emotograph and filed patent in 1925. However, Keeler's first handmade polygraph instrument was destroyed in a fire at his residence in 1924, and he subsequently rebuilt and refined it [8]Verified Wikipedia — Leonarde Keeler
Confirms Keeler's first handmade polygraph (Emotograph) was destroyed in a fire at his residence in 1924. Keeler spent years refining Larson's designs, removing inefficiencies such as the cumbersome process of scribing results on smoke paper that required shellacking for long-term storage. His innovations made the instrument easier to set up, transport, and operate in the field.

Keeler's Legacy and the Pre-Digital Era

Keeler's polygraph designs quickly gained adoption by the CIA, the Department of Defense, and numerous law enforcement agencies across the United States, establishing them as the gold standard of polygraph technology. He received U.S. Patent 1,788,434 in 1931 and partnered with Associated Research, Inc. (originally the Western Electro Mechanical Company) in Chicago to produce the world's first commercially manufactured lie detector instruments [9]Verified LieDetectorTest.com — Keeler Polygraph Instruments History & Model Guide
Confirms Keeler received U.S. Patent 1,788,434 in 1931 and partnered with Associated Research, Inc. in Chicago. The first advancement was model #301, replacing his original Emotograph design [10]Verified LieDetectorTest.com — Leonard Keeler Northwest Crime Lab Chicago Years
Confirms partnership with Western Electro Mechanical Company (later Associated Research, Inc.) for commercial manufacture. Keeler continued refining his instruments throughout the 1930s and 1940s with Associated Research. Other manufacturers, notably Stoelting Company, also developed analog polygraph systems during this period, creating competitive pressure that drove incremental innovation.

The analog polygraph era was characterized by several defining features: ink pens recording on moving chart paper, manual interpretation by trained examiners, limited portability due to the size and weight of instruments, and a reliance on the subjective expertise of the examiner to read and score charts. While effective in trained hands, these instruments had inherent limitations. The estimated accuracy of Keeler-era polygraphs ranged from approximately 60% to 70% [11]Verified LieDetectorTest.com — Are Polygraph Exams Accurate
Confirms estimated accuracy of Keeler-era polygraphs ranged from approximately 60% to 70% — significantly better than chance but far from the reliability needed for high-stakes decision-making.

Keeler himself died in 1949 at just 45 years of age — decades before the first personal computer would appear [12]Verified LieDetectorTest.com — Leonarde Keeler Chicago Crime Lab
Confirms Keeler's death in 1949 at age 45 and continuation of his work by associates. He never witnessed the revolution that would transform his life's work. But the foundation he built — the principle that physiological responses to questioning could be measured, recorded, and analyzed to detect deception — remained the bedrock upon which all future polygraph technology would be constructed. After Keeler's death, his associates and trainees continued his work, and the professional model he established became the template adopted by law enforcement agencies, government programs, and private polygraph firms across the country [12]Verified LieDetectorTest.com — Leonarde Keeler Chicago Crime Lab
Confirms Keeler's death in 1949 at age 45 and continuation of his work by associates.

Stoelting and the CPS Algorithm

The 1980s personal computer revolution inevitably reached the polygraph industry. By the late 1980s, companies were already developing software-based polygraph systems that would fundamentally change how deception was measured and scored. Two algorithms emerged during this period that would reshape the industry: CPS and PolyScore.

Scientific Assessment Technologies developed the Computerized Polygraph System (CPS) in 1988, based on research conducted by John Kircher and David Raskin at the psychology laboratory at the University of Utah, in collaboration with the Stoelting company [13]Verified National Academies — Appendix F: Computerized Scoring of Polygraph Data
Confirms CPS was developed by Scientific Assessment Technologies based on Kircher and Raskin research at University of Utah (1988), and PolyScore developed at JHU-APL. Much of the foundational work came from the Computer Assisted Polygraph System (CAPS), which had been developed using data gathered in controlled laboratory environments using simulated crime scenarios. However, the newer CPS version represented a significant advancement by relying on field data provided by US Secret Service Criminal Investigations, giving it real-world applicability that laboratory-only data could not match.

The CPS algorithm operates around standard multivariate linear discriminant function analysis — a statistical method that uses multiple variables simultaneously to classify observations into predefined groups (in this case, deceptive or truthful). The algorithm uses this analytical framework to produce calculations estimating the probability of deception or truthfulness in the examinee. For examiners unfamiliar with reading polygraph test results, these computerized scoring methods represented a dramatic shift toward objectivity. Research demonstrated that automated analysis could achieve accuracy comparable to or sometimes exceeding manual scoring [14]Verified LieDetectorTest.com — Stoelting Company Polygraph History and Innovation
Confirms automated analysis could achieve accuracy comparable to or exceeding manual scoring.

Modern versions of CPS analyze three primary features when calculating polygraph scores: cardiograph reading increases (changes in heart rate and blood pressure patterns during relevant questioning), skin conductance amplitude (also known as galvanic skin response, measuring electrodermal activity changes), and combined lower and upper respiration measurements (analyzing both thoracic and abdominal breathing patterns for suppression or irregularity). The updated CPS algorithms build analytically on the original Utah numerical scoring system, which shares similarities with the Seven-Position Numerical Analysis Scale taught by the Department of Defense Polygraph Institute (DoDPI) for manual scoring.

Axciton Systems and the PolyScore Algorithm

While Stoelting was working on CPS, Axciton Systems was forging its own path in collaboration with the Johns Hopkins University Applied Physics Laboratory to produce its PolyScore algorithm [15]Verified LieDetectorTest.com — Polygraph Technology in the 1990s vs the 2020s (Original)
Confirms details about CPS development, PolyScore collaboration, Bruce White's research from 1988, and other key claims. Axciton Systems, founded in Houston, Texas in 1987, launched the first commercially viable computerised polygraph system in 1990 [16]Verified LieDetectorTest.com — Polygraph Manufacturers
Confirms Axciton launched first commercially viable computerised polygraph system in 1990 and Johns Hopkins APL collaboration.

The collaborative research began in 1989, when statisticians John C. Harris and Dr. Dale E. Olsen formed a collaborative effort between Johns Hopkins University Applied Physics Laboratory and Axciton to develop the PolyScore system [17]Verified LieDetectorTest.com — How Software Changed the Lie Detector Test
Confirms 1989 collaboration between John C. Harris, Dr. Dale E. Olsen, JHU-APL, and Axciton to develop PolyScore. Bruce White of Axciton Systems started his research into developing PolyScore in 1988, assisting the team at Johns Hopkins with data sets provided by Axciton Systems [18]Verified LieDetectorTest.com — History of Axciton Systems
Confirms Axciton Systems was founded in Houston, Texas in 1987.

PolyScore takes a fundamentally different approach from CPS. Rather than attempting to recreate the manual scoring processes used by human examiners, PolyScore uses inputs from digitized signals and outputs deception probability based on neural network models or logistic regression [15]Verified LieDetectorTest.com — Polygraph Technology in the 1990s vs the 2020s (Original)
Confirms details about CPS development, PolyScore collaboration, Bruce White's research from 1988, and other key claims. PolyScore 3.0 was developed by analyzing polygraph data from 301 presumed nondeceptive and 323 presumed deceptive criminal incident polygraph examinations, with six Axciton instruments [19]Verified National Academies — The Polygraph and Lie Detection (PolyScore Data)
Confirms PolyScore 3.0 was developed from 301 nondeceptive and 323 deceptive criminal incident examinations, and is used with Axciton and Lafayette instruments. The algorithm processes digitized signals from blood pressure, galvanic skin response, and upper respiratory activity, breaking them into fundamental signal components that isolate the portions containing information relevant to detecting deception.

A critical distinction of PolyScore's development methodology is that its algorithm features were extracted from signals based on their empirical performance — meaning the developers let the data determine which features were most predictive — rather than relying on psycho-physiological assumptions about what "should" indicate deception. This data-driven approach gave PolyScore a different analytical character from CPS.

Today, the latest versions of PolyScore are deployed in both Axciton and Lafayette polygraph instruments [19]Verified National Academies — The Polygraph and Lie Detection (PolyScore Data)
Confirms PolyScore 3.0 was developed from 301 nondeceptive and 323 deceptive criminal incident examinations, and is used with Axciton and Lafayette instruments. The companies base ongoing development of the algorithm on case data received from the DoDPI, ensuring the system continues to evolve based on real-world examination results.

The End of the Ink-and-Paper Era

The commercial success of PolyScore and CPS opened the floodgates for the software revolution in polygraph testing. Keeler's analog instruments, which had dominated the market for nearly half a century, were suddenly obsolete. The new computerized polygraph software eliminated the chronic headaches of analog systems: no more clogged ink pens mid-examination, no more storing reams of chart paper in filing cabinets, no more concerns about chart paper degradation over time.

By the mid to late 1990s, software-based systems had comprehensively taken over analog machines in both law enforcement and national security operations. Federal agencies including the NSA, CIA, and FBI — already heavy users of polygraph technology for screening candidates, employees, and intelligence operatives — rapidly adopted the new computerized systems for their superior accuracy and data management capabilities.

The Computing Context of the 1990s

To appreciate the significance of this transition, it helps to understand the computing landscape of the early 1990s. Microsoft's Windows operating system was just gaining mainstream adoption. Intel's Pentium processor, released in 1993, represented cutting-edge processing power. Computers were still bulky, expensive, and offered a limited user experience by today's standards. The internet existed but was not in widespread public use until the mid-to-late 1990s.

In this context, the development of reliable polygraph software was a remarkable engineering achievement. These programs had to run efficiently on the relatively modest hardware of the era while processing real-time physiological data streams and producing accurate deception probability scores. The fact that both CPS and PolyScore achieved this on early 1990s hardware speaks to the quality of their underlying algorithms.

New Players Enter the Market

As the decade progressed, more companies recognized the commercial opportunity in polygraph software. The algorithmic bandwagon attracted new entrants, including Lafayette Instruments, which would grow to become one of the largest players in the space. The increased competition drove further innovation, with each manufacturer seeking to differentiate its products through improved accuracy, better user interfaces, or enhanced data management features.

The emergence of laptop technology in the late 1990s represented another pivotal shift. Previously, computerized polygraph systems required a dedicated desktop computer alongside the physiological sensors. The advent of portable laptops meant that examiners could carry their entire testing apparatus — computer, software, and sensors — in a single kit, dramatically improving the portability and field-readiness of polygraph examinations.

The Internet Age and Its Impact on Polygraph Development

The 2000s witnessed the internet's explosion into every avenue of society. Billions of dollars poured into research and development across the technology sector, and the ripple effects reached polygraph science. Processing power doubled roughly every two years following Moore's Law, laptops became dramatically slimmer and more portable, and software development tools matured significantly.

These advances meant polygraph software could become more sophisticated. The increased processing power allowed algorithms to analyze more data points in real time, the improved displays gave examiners clearer visualizations of physiological data, and the falling cost of technology made computerized systems accessible to smaller polygraph firms that might previously have relied on analog equipment.

Limestone Technologies and Market Consolidation

Companies like Limestone Technologies, founded in 2003, entered the polygraph software market and intensified competition [20]Verified LieDetectorTest.com — Limestone Technologies Innovation
Confirms Limestone Technologies was founded in 2003 and acquired by Lafayette in 2022. By this point, the "big four" polygraph software developers had emerged: Stoelting, Lafayette Instruments, Axciton Systems, and Limestone Technologies. The Canadian manufacturer built its reputation on delivering some of the highest-resolution polygraph systems in the industry, with sampling rates exceeding 600 Hz per channel [21]Verified LieDetectorTest.com — Polygraph Manufacturers (Limestone Details)
Confirms Limestone Technologies' sampling rates exceeding 600 Hz per channel.

The competitive dynamics eventually led to market consolidation. In August 2022, Limestone Technologies became a subsidiary of Lafayette Instrument Company [22]Verified Limestone Technologies Official Website
Confirms in August 2022 Limestone Technologies became a subsidiary of Lafayette Instrument Company. Lafayette had previously acquired MindWare Technologies in 2004 and Acumar Technology in 2006 [23]Verified LieDetectorTest.com — History of Lafayette Instrument
Confirms Lafayette's acquisitions: MindWare Technologies (2004), Acumar Technology (2006), and Limestone Technologies (2022). The Limestone acquisition significantly increased Lafayette's market share and established it as the industry leader, with Stoelting following as the second-largest player. This consolidation reflected a maturing industry where scale, integration, and ongoing R&D investment became critical competitive advantages.

Government Standards and Adoption

National security agencies including the DOD, NSA, FBI, and CIA were already heavy users of polygraph technology and embraced computerized systems. The DIA, for example, uses computerized Lafayette polygraph systems for routine counterintelligence testing [24]Verified Wikipedia — Polygraph
Confirms the DIA uses computerized Lafayette polygraph systems for routine counterintelligence testing.

Recognizing the need for standardization, the National Institute of Justice (NIJ) developed standards for computerized polygraph systems. NIJ Standard-0110.01 established comprehensive rules covering three critical areas: hardware and software specifications (technical requirements for sensors, analog-to-digital converters, computer hardware, and software platforms), examination procedures (standardized protocols for conducting computerized polygraph examinations, including sensor attachment, question formatting, data collection parameters, and scoring methodologies), and examiner training requirements (minimum training standards for polygraph examiners operating computerized systems, acknowledging that digital tools required new competencies beyond traditional analog examination skills) [25]Verified LieDetectorTest.com — Polygraph Technology 1990s vs 2020s (NIJ Standards)
Confirms NIJ Standard-0110.01 covering hardware/software specs, exam procedures, and examiner training requirements.

These standards provided a critical framework for quality assurance and helped establish confidence in the new technology. For agencies like the FBI, CIA, and Customs and Border Protection, having recognized standards meant they could confidently integrate computerized polygraph testing into their operational workflows.

The Dramatic Improvement in Accuracy

Perhaps the single most significant change in polygraph technology from the 1990s to the 2020s is the dramatic improvement in accuracy that algorithms brought to the industry. The older Keeler-era and Reid-technique analog polygraphs were generally considered to achieve accuracy rates of approximately 60% to 70% in detecting deception [11]Verified LieDetectorTest.com — Are Polygraph Exams Accurate
Confirms estimated accuracy of Keeler-era polygraphs ranged from approximately 60% to 70%.

By comparison, the American Polygraph Association (APA) reports that the combination of all validated psychophysiological detection of deception (PDD) techniques produced a decision accuracy of 87%, with a confidence interval of 80%–94% [26]Verified American Polygraph Association — Polygraph Validity Research
Confirms combination of validated PDD techniques produced decision accuracy of 87% (CI 80%-94%) with 13% inconclusive rate. Some studies and software developers report accuracy rates as high as 98% depending on the technique and methodology used [27]Verified LieDetectorTest.com — Polygraph Accuracy
Confirms accuracy rates between 87% and 98% depending on methodology and examiner competence. This represents a transformative improvement — moving from a technology that was correct roughly two-thirds of the time to one that is reliable in the vast majority of examinations.

What Drives the Accuracy Improvement?

Several factors contribute to the higher accuracy of modern computerized systems:

Digital signal processing — Modern systems capture physiological signals as digital data, eliminating the noise and imprecision of analog recording. Digital signals can be amplified, filtered, and analyzed with mathematical precision that ink traces on paper simply cannot match.

Algorithmic objectivity — Whereas analog chart interpretation relied on individual examiner judgment (with all the cognitive biases that implies), computerized algorithms apply consistent mathematical criteria to every examination. Two examiners running the same data through the same algorithm will get identical results.

Multi-channel analysis — Modern algorithms can simultaneously analyze multiple physiological channels, identifying subtle patterns and correlations that a human examiner might miss when visually reviewing chart paper.

Statistical validation — Algorithms like CPS and PolyScore have been validated against large databases of known-outcome cases, allowing developers to calibrate their systems for optimal accuracy using established statistical methods [19]Verified National Academies — The Polygraph and Lie Detection (PolyScore Data)
Confirms PolyScore 3.0 was developed from 301 nondeceptive and 323 deceptive criminal incident examinations, and is used with Axciton and Lafayette instruments.

Continuous refinement — Unlike a fixed analog instrument, software algorithms can be updated and improved as new research data becomes available, meaning modern systems get progressively better over time.

The Employee Polygraph Protection Act of 1988

Despite the dramatic improvements in polygraph accuracy, the legal framework governing polygraph use in the United States has not kept pace with technological advancement. The Employee Polygraph Protection Act (EPPA) of 1988 remains in full effect, restricting most private employers from requiring employees or job applicants to take polygraph tests [28]Verified US Department of Labor — Employee Polygraph Protection Act
Confirms EPPA of 1988 restrictions on private employer use of polygraph testing.

The EPPA was enacted during the analog era, when polygraph accuracy rates of 60% to 70% raised legitimate concerns about false positives and the potential for workplace exploitation. However, with modern computerized systems achieving accuracy rates of 87% to 98%, the law's underlying premises have arguably been overtaken by technology.

Rule 702 and Courtroom Admissibility

Rule 702 of the Federal Rules of Evidence and various court precedents continue to restrict the admission of polygraph results as evidence in most court proceedings. The landmark legal frameworks — including the Frye Standard and Daubert Standard — were established when polygraph technology was far less reliable than it is today.

Some states do allow the admission of polygraph results under specific circumstances, typically requiring the agreement of all parties involved in the legal proceedings. Additionally, polygraph results are frequently used in plea negotiations, probation supervision, and other contexts where formal evidentiary standards do not apply.

Industries Exempt from the EPPA

Several categories of employers remain exempt from EPPA restrictions: federal, state, and local government agencies (including all law enforcement, military, and intelligence agencies); security service firms (companies providing armored car, alarm, and guard services); pharmaceutical manufacturers and distributors (companies with direct access to controlled substances); and the ongoing investigation exemption (private employers may use polygraph testing as part of an investigation into theft, embezzlement, or other economic loss, subject to specific procedural requirements) [28]Verified US Department of Labor — Employee Polygraph Protection Act
Confirms EPPA of 1988 restrictions on private employer use of polygraph testing.

Many private companies continue to rely on computerized polygraph examinations for hiring and internal investigations within these exemptions.

The Next Frontier in Deception Detection

While the transition from analog to digital polygraph technology was revolutionary, an even more transformative innovation is emerging: artificial intelligence-based deception detection. The mid-2010s saw enormous advances in AI technology, particularly in machine learning, computer vision, and natural language processing. These capabilities are now being applied to lie detection in ways that Leonard Keeler and John Larson could never have imagined.

Unlike traditional polygraph testing, which requires physical sensors attached to the examinee's body, AI-based systems can potentially detect deception through non-contact analysis of voice patterns, facial expressions, and micro-behavioral cues. This represents a paradigm shift in how deception detection might be conducted at scale.

AVATAR: The American AI Screening System

AVATAR (Automated Virtual Agent for Truth Assessments in Real-Time) represents one of the most prominent examples of AI-based deception detection. It is a deception-detection technology created by researchers at the University of Arizona, developed in collaboration with U.S. Customs and Border Protection (CBP) [29]Verified University of Arizona BORDERS — AVATAR Project
Confirms AVATAR was created by researchers at the University of Arizona, analyzes facial expressions and voice, body, and eye signals. The technology was licensed to the local startup Discern Science International, Inc. [30]Verified Policing Project — Lie Detection II: 21st Century Tests
Confirms AVATAR was developed by University of Arizona researchers, licensed through Discern Science International, Inc..

AVATAR is a kiosk-like system originally spearheaded by the BORDERS project group. The system was designed to automate credibility assessment interviews by customs authorities at ports of entry, such as those used in visa processing or asylum requests, as well as personnel screening [31]Verified CT Strategies — AVATAR Technology at Borders
Confirms AVATAR tested at U.S.-Mexico SENTRI lane in Nogales, Arizona in 2011 and 2012. AVATAR analyzes facial expressions and measures nuanced voice, body, and eye signals [29]Verified University of Arizona BORDERS — AVATAR Project
Confirms AVATAR was created by researchers at the University of Arizona, analyzes facial expressions and voice, body, and eye signals.

The system was first revealed in 2012 by researchers at the University of Arizona [32]Verified Daily Mail — Lie-Detecting Computers for Border Security
Confirms AVATAR was first revealed in 2012 by researchers at the University of Arizona. It was evaluated and tested by the DHS Science and Technology Directorate and DHS operational components in 2012, including at a U.S.-Mexico SENTRI lane in Nogales, Arizona in 2011 and 2012, as well as at Henri Coandă International Airport in Bucharest [31]Verified CT Strategies — AVATAR Technology at Borders
Confirms AVATAR tested at U.S.-Mexico SENTRI lane in Nogales, Arizona in 2011 and 2012 [33]Verified CNN — Computer Border Official AVATAR
Confirms AVATAR was developed by University of Arizona researchers in collaboration with U.S. Customs and Border Protection. The AVATAR system achieved accuracy rates of 60 to 75 percent, and sometimes up to 80 percent as a deception-detection judge — consistently above human accuracy, which is generally about 54 to 60 percent at best [34]Verified CNBC — AI Lie Detectors and Border Security
Confirms AVATAR accuracy of 60-75% and sometimes up to 80%, versus human accuracy of 54-60%. Developers suggest the system could achieve up to 85% accuracy with additional training data and refinement.

iBorderCtrl: The European Union's AI Initiative

The European Union developed its own AI-based border screening technology through the iBorderCtrl project, an initiative of the EU's Horizon 2020 research and innovation program. The consortium was led by European Dynamics Luxembourg SA [35]Verified European Dynamics — iBorderCtrl R&D Contract
Confirms consortium led by European Dynamics Luxembourg SA was awarded iBorderCtrl contract under Horizon 2020, and the core deception detection technology was developed by scientists at Manchester Metropolitan University [36]Verified Biometric Update — EU to Pilot AI Facial Analysis
Confirms €4.5 million EU-funded iBorderCtrl project piloted between 2016 and 2019 in Greece, Hungary, and Latvia. The project received €4.5 million in EU funding [37]Verified The Guardian — Europe's Secretive Push Into Biometric Technology
Confirms iBorderCtrl received €4.5m from Horizon 2020 security portfolio.

The iBorderCtrl project ran between September 2016 and August 2019 [38]Verified TechCrunch — AI Lie Detector Project Challenged in EU Court
Confirms iBorderCtrl ran between September 2016 and August 2019, with pilot testing conducted in Greece, Hungary, and Latvia [36]Verified Biometric Update — EU to Pilot AI Facial Analysis
Confirms €4.5 million EU-funded iBorderCtrl project piloted between 2016 and 2019 in Greece, Hungary, and Latvia. The system focuses primarily on analyzing facial micro-expressions of travelers, asking them targeted questions designed to reveal deceptive behavior. Sample questions include: "Do you have anything in your suitcase that might be illegal to bring into the country?" followed by cognitively loaded follow-ups such as "If you were to open your suitcase now, would there be anything in there that would concern customs?" The system analyzes facial movements, micro-expressions, and physiological indicators during both the initial question and the follow-up to detect inconsistencies suggesting deception.

The technology had a 76 percent success rate in early testing, and a representative of iBorderCtrl stated they were "quite confident" it could reach 85 percent accuracy [39]Verified The Verge — EU Plans to Test AI Lie Detector at Border Points
Confirms early iBorderCtrl testing had 76% success rate, with team confident it could reach 85%. Both AVATAR and iBorderCtrl demonstrate the potential for AI to assist with high-volume security screening while reducing the human resources traditionally required for manual interviews and assessments.

How AI Systems Compare to Traditional Polygraph

It is important to note that current AI-based systems like AVATAR and iBorderCtrl, with accuracy rates of 60% to 85%, have not yet matched the reported 87% to 98% accuracy of properly administered computerized polygraph examinations. However, AI systems offer significant advantages in specific applications.

AI system advantages include non-contact operation requiring no physical sensors on the subject, ability to screen large volumes of people rapidly, no requirement for a trained polygraph examiner for each test, built-in countermeasure detection, operation in non-traditional environments like airports, and continuous improvement through machine learning.

AI system limitations include current accuracy rates lower than computerized polygraph, privacy concerns around facial recognition technology, potential for cultural and demographic bias in algorithms, limitation by the quality and diversity of training data, lack of validation to the same standards as traditional polygraph, and ethical concerns about mass surveillance applications.

Three Generations of Lie Detection

To fully appreciate how far polygraph technology has come, consider the three generations of lie detection technology that have defined the past century.

Analog Era (1920s–1980s): Ink-and-paper recording. Manual chart interpretation by trained examiners. Accuracy of approximately 60%–70% [11]Verified LieDetectorTest.com — Are Polygraph Exams Accurate
Confirms estimated accuracy of Keeler-era polygraphs ranged from approximately 60% to 70%. Limited portability. Results heavily dependent on individual examiner skill. Scoring subjectivity introduced significant variability. The Keeler and Reid polygraphs defined this era.

Digital/Software Era (1990s–Present): Digital signal capture and processing. Algorithmic scoring using validated statistical models (CPS, PolyScore). Accuracy of 87%–98% per APA benchmarks and published research [26]Verified American Polygraph Association — Polygraph Validity Research
Confirms combination of validated PDD techniques produced decision accuracy of 87% (CI 80%-94%) with 13% inconclusive rate [27]Verified LieDetectorTest.com — Polygraph Accuracy
Confirms accuracy rates between 87% and 98% depending on methodology and examiner competence. Laptop-portable systems. Objective, reproducible results. NIJ standards governing hardware, software, and procedures. Four major manufacturers competing and innovating.

AI Era (2010s–Emerging): Non-contact deception detection using facial recognition, voice analysis, and micro-expression analysis. Accuracy of 60%–85% in pilot programs [34]Verified CNBC — AI Lie Detectors and Border Security
Confirms AVATAR accuracy of 60-75% and sometimes up to 80%, versus human accuracy of 54-60% [39]Verified The Verge — EU Plans to Test AI Lie Detector at Border Points
Confirms early iBorderCtrl testing had 76% success rate, with team confident it could reach 85%. Scalable for high-volume screening at borders and checkpoints. Machine learning enables continuous improvement. AVATAR and iBorderCtrl are the flagship systems. Not yet a replacement for traditional polygraph but a complementary technology.

Key Metrics Across Generations

The following aspects have evolved most significantly across these three technological generations:

Accuracy: From approximately 65% (analog) to approximately 87%–98% (digital) to approximately 60%–85% (AI current, with 85% projected ceiling).

Portability: From large, desk-mounted instruments to laptop-based systems to software-only solutions.

Objectivity: From fully subjective manual scoring to algorithm-driven objective scoring to automated AI classification.

Throughput: From one-at-a-time individual examinations to slightly faster digital exams to potentially hundreds of screenings per hour.

Contact requirement: From multiple physical sensors (pneumo tubes, blood pressure cuff, galvanic sensors) to the same sensor suite digitized, to no physical contact required.

Data storage: From paper charts in filing cabinets to digital files on computers to cloud-based databases.

Examiner dependency: From near-total reliance on examiner skill to algorithm-assisted scoring to potentially examiner-free automated systems.

AI Integration with Traditional Polygraph

The most likely near-term evolution of polygraph technology is not a replacement of traditional methods but an integration of AI capabilities with existing computerized systems. Future polygraph examinations could simultaneously capture traditional physiological data (cardiovascular, electrodermal, respiratory) while also analyzing the examinee's voice patterns, facial micro-expressions, and eye-tracking data — all processed by AI algorithms trained on millions of examinations.

This multimodal approach could push accuracy rates even higher than current computerized systems achieve, potentially approaching near-certainty in many testing scenarios. The combination of contact-based physiological measurement and non-contact AI analysis would create multiple independent data streams, making it exponentially harder for examinees to employ countermeasures.

Brain-Based Lie Detection

Another emerging frontier is brain-based deception detection, including technologies like EEG and P300 brainwave analysis. These approaches measure brain activity directly rather than relying on peripheral physiological responses, potentially offering a more direct window into cognitive processes associated with deception. While still largely in the research stage, brain-based methods could eventually complement both traditional polygraph and AI-based systems.

Regulatory and Ethical Considerations

As technology advances, the regulatory framework will need to evolve as well. Current laws like the EPPA were designed for a different technological era. The emergence of AI-based screening raises new questions about privacy, consent, and civil liberties that existing legislation does not address.

The European Union's General Data Protection Regulation (GDPR) has already raised questions about the legality of AI-based biometric screening systems like iBorderCtrl, particularly regarding the use of facial recognition data. The iBorderCtrl project faced a legal challenge in EU court over transparency and civil liberties concerns [40]Verified TechCrunch — iBorderCtrl EU Court Challenge
Confirms legal challenges to iBorderCtrl in EU court over transparency and civil liberties. Similar debates are unfolding in the United States regarding the appropriate limits of AI-based behavioral analysis in law enforcement and border security contexts.

Implications for the Polygraph Profession

For working polygraph examiners, these technological shifts present both opportunities and challenges. The core skills of a competent examiner — question formulation, rapport building, behavioral observation, and professional ethics — remain essential regardless of what technology is used. However, examiners must also develop competency with increasingly sophisticated digital tools and stay current with AI-assisted analysis methods.

Those entering the field today should expect a career defined by continuous technological change. The examiner who graduated from polygraph school in 1990 using analog equipment has already navigated one complete technology revolution; the examiner entering the field in the 2020s may well navigate another.