Can we teach an algorithm to make a human judgement of guilt in court?
Do defendants have the right to a human trial?
Criminal trials in Scotland use an adversarial model, meaning that trials are between the accused and alleged victim, for the public interest, with the jury representing this public interest, and making the ultimate verdict.
This speculative design proposal aimed to provoke debate around the justice system.
Imagine a courtroom without a jury that can be swayed, without personal biases and human error.
How would that work?
People judge whether others are truthful through assessing eye movement, and perspiration along with body language and tone of voice. The “Jury Service” would learn from human decisions to make its own human-like decisions based on measuring both the defendant and witness’ physical data. To train this algorithm there will be a team of professional jurors that will pass verdicts on criminal trials through a mobile application.
Before making any verdicts the users will first be classified into 6 different personality types, based on how they make decisions. They will also have to undergo specific training modules for each type of case. They will make their verdict of ‘guilty’, ‘not guilty’ or ‘not proven’ based on a police report by the procurator fiscal, with his recommendation, as well as video clips of witnesses, accused and alleged victim giving evidence in court, as well as any other evidence.
The data will be collated by personality type and checked by a team of a psychologists and data scientist, this will generate one verdict per personality type for the case. Since juries are currently made up of 15 people there will be a secondary algorithm that will randomly weight each personality type and produce a majority verdict that will be presented in court.
At the same time as being video recorded all defendants, alleged victims and witnesses will wear a device on their neck will speaking in court. This device will record the person’s pulse, body temperature, perspiration as well as the tonality and frequency of their voice. This data will be collated with the verdicts given by human participants, meaning that in the future the device alone will be able to predict a human judgement of guilt based on these physical parameters.