It's still subjective and based upon interpretation of the Lie Detector Consultant.
Torture has been used throughout history too. It is based upon forced testimony. Sometimes the testimonies have been proven correct, sometime as lies and sometimes inconclusive. Such is human nature....sin nature that usually places the interrogator on the " moral high ground."
For instance a TV series from the early 2000s called 24 with Keiffer Sutherland. I hated the show because it was a promotion for the wars and often justified torture. I warned back then that the acceptance if such by the Bush republicans would come back and bite them. They went along with it providing the rope for their own hanging now. The Sutherland characters already designate the conservatives as such and more of a threat than Alc👁️ada.
When the brain implants and hand implants become a requirement as the state IDs are today, there may be such torture incentives or the Chinese model. All media like these forums show a lot to the 55 + agencies and many more corporations. Then the interface asks questions about various things that either reveal dopamine or cortisol hormones to determine truth and lies. Those sensors probably exist in the implantables. I am simply speculating, but trying to see your point.
Lie detection is both art and science.
There is no "silver bullet". But AI can be trained in spotting cues both micro expressions, voice strain, touching your nose, a shrug, self comforting, etc. Torture is not an effective technique because it motivates people to lie, but making the questioning stressful is. Lying is difficult. Lying to people trained to spot lies is even harder. What all these techniques do is spot the questions that the person had the most difficulty with. The polygraph detects heightened stress. But facial cues detect emotional responses like duping delight and scorn and surprise. Body cues can spot when someone has no confidence in a response.
You identify the questions that caused the most difficulty and then you drill down on those. A really good, well trained liar can be very convincing for three or four hours. But after 10 hours they can wear down. You can make them tell their story in reverse, more difficult for all but the best liars. You can verify parts of their story. Then you can ask questions that they didn't think of and didn't prepare for.
Even if after 24 hours you can't find the lie, no biggie, you have their entire testimony recorded. So it may be in two or three months you uncover something that does not align with their testimony.
What AI does is allow you to raise the level of all your interrogations to the height of the best interrogators in the world. What it also does is make it so that all of your interrogations are raised to this level. You may bring in someone you think was simply a bystander, question them and release them. But because AI was focused on this person as well as all the other pieces of evidence it can flag that person as having lied or more likely as "having hid something".
The best interrogators are running about 60% in their accuracy. But that is when questioning a single person without any other information. The more information the interrogator has the better the odds.
Imagine a police department that has thousands of cameras and video footage from around the city, they have microphones up with these videos capture information. They also have machines reading license plates and putting them into a spreadsheet with time and location, these are attached to every cop car in the city. They also have gps information from phones. That is a lot of information that a lie is going to have to align with.
Then the AI can feed the questions. Knowing what to ask questions about is big, also knowing what questions to ask is big.
Finally, if you can catch a person in a lie it is big because that will greatly help you in identifying their tell.
Personal story: We set up a challenge in my school and I put up $500. The challenge was to come and lie to our class, if my forensic class could not detect the lie you win the $500. First, by offering this to the entire school we were encouraging the people most confident in their ability to lie. Then the competition had three stages. First, if you win, you get $5. Round two you put up your $5 and if you win you get $50. Third, if you make it to this round you bet your $50 and if you win you get $500. The reason for this is that the higher the stakes the harder it is to lie. Also, for some losing money is higher stakes than winning money.
Also, we were videotaping each round. At the end of each round we would know that the person was lying so we were studying them for the tell.
Everyone in my class had trained on micro expressions for about a month. They would question the person, I did not question. However, I would then help them in class go over the video footage. In the first round only one person deceived them. But then we saw her tell and they all detected her lie in the second round (we would take a vote on whether or not we thought the person was lying so there was a certain level of uncertainty, but in that second round it was unanimous that she was lying). She was probably the best liar in the school of 400 teenagers.