The Perfect AI lie Detector


Table of Contents

  1. Introduction
  2. Faults of the polygraph and other lie detector methods
  3. AI Lie detector
  4. Concluding Remarks

1. Introduction

Humans being able to detect those trying to deceive would be immensely important for many situations such as solving crimes or reading political intent. There have been numerous previous attempts to create methods that can detect those trying to deceive such as polygraphs or even semantic analysis methods.

Being able to discern between a lie and the truth traditionally has always been up to humans. However, we hear lies constantly. A study by the USC psychologist, Jerry Jellison, states that we hear up to 200 lies a day. Being able to detect lies without failure in the situations that matter most such as in crime or politics could be extremely game-changing.

  1. Faults of the lie detection methods

2.1. Polygraph

The most popular lie detection technology currently available in criminal prosecution is the polygraph. However, the accuracy of this technology is shoddy at best. Many law enforcement agencies choose not to use the technology at all citing its terrible record. Some organizations say that polygraphs have 70% accuracy. However, clearly guilty parties such as Gary Ridgway ( the Green River Killer) have fooled the polygraph systems. These failures cast doubt on the use of them in criminal cases. The National Research Council found that polygraphs are better than flipping a coin but are far from perfect.

2.2. Voice Stress Analysis

As the name of this method implies, this technique attempts to compare pitch and frequency of the voice over the course of an answer. This method is perhaps more versatile than the polygraph – it can be used over the phone or with a recording. Despite sounding promising, this method is barely better than detecting a lie based on flipping a coin. In a 2009 study by Damphousse and others, there was just a 48% accurate classification of those trying to deceive based on voice stress analysis.

2.3. fMRI

Perhaps one of the most sophisticated lie detection methods that are out there is the combination of fMRI and EEG. These are more rigorous scans of the brain which allow researchers to see changes in the brain over a period of time. A complete picture of which parts of the brain are activated – indirectly through observing which regions are using more oxygen. Certain regions of the brain are associated with truth-telling as well as telling a lie.

3. AI Lie detectors

3.1. EyeDetect

There are companies that are racing to develop the perfect lie detector based on AI technology. One of the companies, Converus, is working on creating a technology that can analyze video and predict truth based on eye movements and small changes in pupil size. The core technology, called EyeDetect, has already been used by FedEx to screen out drivers with criminal histories in Panama.

3.2. Discern Science International

Discern Science International is a startup that is developing a lie detection tool creatively named Avatar. The product is a virtual human that can interrogate travelers to screen them before they enter countries. Automating some of this work can make borders safer and more efficient.

The technology is being developed by former University of Arizona scientists and could be available as early as this year. While it is between 80 and 85% accurate, it is far more accurate than humans interrogating other humans which results in just a 54% accuracy rate.

Such systems are being used in border security in Europe where immigrants must answer questions such as their ethnicity and their reasons for moving.

3.3. University of Genova

Yet another study investigated 28 participants and analyzed the participants eye movements, time to respond, and eloquence. The researchers from the University of Genova found that pupil dilation, fixations, and blinks were the best features for lie detection. However, this work demonstrated just 71% overall accuracy.

  1. Concluding Remarks

Developing a lie detector that is more fault-free over the polygraph would be extremely important to identify those that are lying in criminal trials and other situations. However, very quickly the ethics question gets called into play as it was with the accuracy of the polygraph test.

Should we adopt a technology if it proves to be more accurate than the polygraph? Or eliminate the use of all such technology in our criminal prosecution practices until those methods are completely 100% accurate.

Looking beyond solving crimes, the technology could be applied more broadly. Do we want that? Would we want a Google Glass type of hardware that could discern if our friend is lying? It is easy to see both sides of the coin. In one sense, we could eliminate lying from humanity. On the other hand, isn’t that a uniquely human trait? In a world that is becoming increasingly dominated by computers, we may want to preserve those characteristics that make us less robotic.

These are open questions with no clear answers. We will need to wait and see how this potentially game-changing technology becomes integrated into our lives.

Show More

Nick Allyn

Hello, my name is Nick Allyn. I am extremely passionate about the field of artificial intelligence. I believe that artificial intelligence will save millions of lives in the coming years due in higher cancer survival rates, cleaner air, as well as autonomous cars.

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button