Should artificial intelligence be used in the justice system, and if so should it apply the letter or the spirit of the law? While there are no plans for AI judges yet, this is a question that the UK government is already wrestling with as it considers the . Despite this, lawyers and computer scientists are warning that current systems can鈥檛 handle the ambiguity and nuance often required in legal situations.
For example,聽, formerly at Stanford University and now working at Amazon, and her colleagues assessed how large language models (LLMs), of the type that powers ChatGPT, would judge people carrying out , such as stealing food to feed starving family members. While people may be split on whether or not poverty or need was mitigation in these scenarios, the AIs had none of the understanding or empathy we might expect from human decision makers and almost always found the hypothetical lawbreakers guilty 鈥� which could be disastrous in the context of the legal system.
鈥淭he leeway that judges have where they can say 鈥極K, this is wrong, but in this current situation this is justified鈥� is not something we can really expect an LLM to do as well as a human judge,鈥� says Jain.
Advertisement
But the time-saving potential of AI is a tempting proposition for cash-strapped public services. at the UK鈥檚 AI think tank and research group, the Alan Turing Institute, and her colleagues, have put out a quantifying how many interactions between citizens and the state 鈥� of which there are an estimated one billion each year in the UK 鈥� could be farmed out to AI. The group estimated that 84 per cent were 鈥渉ighly automatable鈥�, although crime and justice was deemed to have a 鈥渓ow potential鈥� for AI automation due to its complexity.
Hashem says that AI could be a huge timesaver in courts in limited roles, such as transcribing proceedings, but not to make decisions. 鈥淚 think that even if the technology itself exists and performs relatively well, it鈥檚 still a question of whether we should or shouldn鈥檛 be implementing it in areas where the consequences are high,鈥� she says.
at the University of Edinburgh, UK, says that there is a concept in legal studies of Judge Hercules 鈥� a superhuman judge who knows all case law, all previous decisions and all the facts of the case, and can arrive at a perfect legal decision. This sounds like what we might want from an AI judge, but Schafer is sceptical that existing technology can step up to the job.
Schafer says human judges can display spontaneous acts of empathy, or a discretionary bending of the rules when it right or just, but that this door swings both ways. And while human judges have individual reputations for leniency or strictness, even for certain types of crime, an AI model would at least be uniform, he says. 鈥淲e want cold logic, that is liberating,鈥� says Schafer. 鈥淩ules liberate us, ultimately.鈥�
But this cold logic brings its own problems, like the so-called in the US that can see a person with two previous convictions caught stealing a loaf of bread face a mandatory life sentence. Schafer says that under that sort of system human judges essentially become computers, who could easily be replaced by AI. Neither option is desirable, he says.
鈥淲hat we normally see in most legal systems is that both elements coexist in an uneasy tension. And as humans we are good at living with these ambiguities,鈥� says Schafer.
