Artificial intelligence systems will soon be able to predict – and decide – the outcomes of trials, raising the possibility of replacing human judges with computers, a new research paper has suggested.
Artificial intelligence systems will soon be able to predict – and decide – the outcomes of trials, raising the possibility of replacing human judges with computers, a new research paper has suggested.
The paper was written by Professor Elliott Ash of Warwick University and published by the Social Market Foundation think-tank and the ESRC’s Centre for Advantage in the Global Economy.
“It is straightforward, in principle, to automate legal decision-making, and there may be some advantages to doing so. But the questions raised by the use of machines in the justice system should be the subject of democratic debate,” Ash said.
Ash is an expert in legal systems who has advised the US Federal government on bias in the criminal justice system. He warned that without careful management, “robot judges” could amplify human judges’ biases against, for example, black defendants.
Machine-learning techniques already allow computers to predict the verdicts that human judges and lawyers will reach in the majority of cases, the paper shows.
Studies based on US legal data show that computers have successfully predicted 88% of prosecutor decisions, 82% of verdicts in asylum cases and 70% of US Supreme Court rulings. Even using “minimal case information”, computers accurately predicted 67% of bankruptcy rulings, Ash found.
These results were achieved using “off the shelf” open-source software, suggesting that dedicated systems developed for court cases could easily achieve much higher accuracy.
Ash said that AI in the courtroom could be a powerful tool for human judges, who could use “robot clerks” to help them reach decisions.
“This is an app that takes in evidence data, runs the numbers, and then produces a prediction about what previous judges would have likely decided. Human judges would then use this prediction as an input into their own decision, which could be based on a wider range of factors,” Ash said. “In uncertain cases the decision would still be made wholly by the human judge.”
Ash said that while computers can help improve judicial decision-making, the use of AI in dispensing justice raises significant questions that need to be answered by politicians.
Machine-learning systems work by assimilating previous human decisions, which means that any bias or prejudice in those decisions will be echoed in the machine’s conclusions. “This matters for the robot judge because any automated decision system that is trained on biased data will also be biased”, Ash said.
For instance, Ash said that US criminal case history suggests that judges are more likely to convict black defendants and give them harsher sentences than white defendants. “This matters for the robot judge because any automated decision system that is trained on biased data will also be biased”, Ash said.
Another worry about the use of AI in justice would be the transparency of decision-making. If the algorithms that an AI system used to make predictions and decisions were not public, its decisions could not be fairly scrutinised or challenged. But if the algorithms were open, it might be possible for lawyers to “game the system” and fool the machine into deciding in their favour.
“The decisions of current algorithms are a black box and cannot be explained to legal participants or the public. There is no assured way to purge pre-existing bias from the machine-predicted decisions. And there are major technological and political challenges to developing and implementing a system that would read the laws and try to implement socially optimal policies,” Ash said.
“Nevertheless, the brave new world of legal automation is upon us. These developments are both thrilling and unsettling because they attack the core of our humanity: is not justice what distinguishes man from machine? These types of questions should be the subject of democratic debate.”
Notes:
Ash’s paper is called Judge, Jury, and EXEcute File: The brave new world of legal automation. It was presented at the Social Market Foundation in Westminster and is published at https://www.smf.co.uk
For more information, contact director@smf.co.uk or call 0207 222 7060. Elliott Ash is @elliottt on Twitter and his website is http://elliottash.com/
About the SMF:
The Social Market Foundation (SMF) is a non-partisan think tank. We believe that fair markets, complemented by open public services, increase prosperity and help people to live well. We conduct research and run events looking at a wide range of economic and social policy areas, focusing on economic prosperity, public services and consumer markets. The SMF is resolutely independent, and the range of backgrounds and opinions among our staff, trustees and advisory board reflects this.
About CAGE:
Established in January 2010, CAGE is a research centre in the Department of Economics at the University of Warwick. Funded by the Economic and Social Research Council (ESRC), CAGE is carrying out a ten year programme of innovative research. The centre’s research programme is focused on how countries succeed in achieving key economic objectives such as improving living standards, raising productivity, and maintaining international competitiveness, which are central to the economic wellbeing of their citizens. Its research analyses the reasons for economic outcomes both in developed economies like the UK and emerging economies such as China and India. CAGE aims to develop a better understanding of how to promote institutions and policies which are conducive to successful economic performance and endeavour to draw lessons for policy makers from economic history as well as the contemporary world. Research at CAGE examines how and why different countries achieve economic success. CAGE defines ‘success’ in terms of well-being as well as productivity. The research uses economic analysis to address real-world policy issues. The centre is distinctive in providing a perspective that draws on economic history as well as economic theory and is applied to countries at various different stages of economic development.
