DECEMBER 29, 2023
Editors' notes
AI reliably detects emotions based on facial expressions in psychotherapeutic situations
The face reflects a person's emotional state. The interpretation of facial expressions as part of psychotherapy or psychotherapeutic research, for example, is an effective way of characterizing how a person is feeling in that particular moment. In the 1970s, psychologist Paul Ekmann developed a standardized coding system to assign basic emotions such as happiness, disgust or sadness to a facial expression in an image or video sequence.
https://medicalxpress.com/news/2023-12-ai-reliably-emotions-based-facial.html
Ekman and others who can do this reliably command top dollar consulting fees. However, AI has several key advantages. First it can be much faster than most people who need to take a video and then go through the video slowly, repeatedly.
Secondly, the AI might identify three thirty second clips in a 30 minute video where it wants a second opinion. So instead of hiring Ekman and his team to do everything you simply pay them to consult on three thirty second clips, something they could do in 30 minutes.
Now imagine you focus this on the jury in a trial. If it is hard to go through a thirty minute interview with one person, imagine going through 8 hours a day of a trial trying to evaluate 12 jurors and three alternates. This is one type of lawyer, a guy who specializes in evaluating a jury. One skill is reading the jury which AI would be able to do as well if not better, but certainly much faster and cheaper than the best teams out there.
A second skill is jury selection. AI is studying Facebook and social platforms to understand the psychology of people and predict their political ideology. This can be very useful in picking a jury. Also, while questioning a potential juror there are very clever psychological tricks that can be used to determine their biases by reading micro expressions.
My point is this, if AI can perform better than a legal consultant whose team can make over a million dollars on a big trial, then why isn't that considered "intelligence"? This is where AI is today, with recursive self improvement AI can study the results of a trial. If it had predicted a juror would vote a certain way and got it wrong it can go back and analyze. Perhaps it was right that the juror was leaning one way, but now needs to pick up on the psychology that this person is a follower and will acquiesce to the more dominant members of the jury. But since AI is already at the level where it can be very good, the recursive training could make it spectacularly good in a matter of a few months.