When artificial becomes real
One of the technologies that will profoundly change the Justice System over the next ten or twenty years is already with us. Often we do not notice it, sometimes it disappoints, occasionally it is the cause of serious concern. The technology is called artificial intelligence (AI) - the ability of technology to display human-like intelligence.
Artificial Intelligence has been with us for many decades. As an academic discipline it can trace its roots back to 1955 and since then the development of AI has waxed and waned. It is only since the start of this millennium that the availability of significant computing power has enabled AI to evolve to the point that it is promising to change so many aspects of our lives; from self-drive travel – arriving in a limited form in the UK later this year, through to facial recognition giving us access to our bank accounts.
AI has already been used in the Justice system. At its simplest level it is behind ‘optical character recognition’ (OCR) – that technology trick to turn scanned or photographed documents into editable text. We don’t notice OCR. We probably don’t notice or think about the technology behind ‘chatbots’ either - again another use of AI that has blended into the background. We do, however notice the more ambitious uses of AI and in doing so we see start to see one of the constraints holding it back and that is the acceptance of its use. For example, facial recognition, the ability to spot a known face in a crowd, is being used by some law enforcement agencies. This use has not been universally accepted and some argue that it is an invasion of privacy and prone to racial bias. There was a similar reaction in the US when AI started to be used to assess the likelihood of re-offending. (But this has not stopped Estonia extending the idea to try and develop a ‘robo-judge’).
The holy grail of artificial intelligence in the judicial system is AI that is better informed and less biased than we are. This will be challenging, but over the next ten years or so will be achievable. There are obstacles that we need to overcome. Firstly, we need a sound but practical ethical framework that defines the uses of AI, gives assurances against bias, and ensures transparency and accountability of decision making. This is a significant challenge, but a number of organisations have developed ‘AI Manifestos’ as a first step towards this, including an EU one specific to the judicial system. Our own House of Lords has also been pro-active in calling for an ethical approach to AI development which it sees as an enabler of development and the government has created a Centre for Data Ethics and Innovation.
Secondly, we need quality information. Information, I was taught, is data with meaning. We have an abundance of data, but AI is only as good as the information it is given. “Rubbish in, rubbish out” is an old IT maxim, which could perhaps be updated to “bias in, bias out” for AI. We need to resist the temptation to throw whatever we have into AI and hope it will make sense of it, and be more sensitive to the context in which the data was gathered, and constrained in the meanings that can be drawn from this. This is a significant task, but one that will become easier as computing power increases, which is where the third requirement comes in.
We have seen that the rapid advance of AI was enabled by the arrival of significant computing power around the start of this millennium. We have become used to processing power increasing exponentially, but we are on the cusp of an even greater computing capability. Simply put, quantum computing will give us astonishingly computing power inconceivably faster than classical supercomputers. This in turn will enable us to enhance and enrich our AI models.
The case for the future of artificial intelligence in the judicial system is enticing and compelling. In law enforcement, facial recognition – with public acceptance - will be a powerful way to deter or solve crimes. Within the courts system, demand can be reduced by using AI to provide tailored legal advice that may stop cases reaching the courts altogether. For complex cases, AI can be used to search through voluminous evidence and AI generated legal assistance to advise and guide on all aspects of case preparation. In turn both these initiatives (filtering demand and better preparation) can improve the quality of submissions to court. By contrast simple ‘predictable’ cases can be scheduled efficiently through courts and standard outcome documentation produced automatically. After the verdict, pre-sentencing reports that are informed by unbiased reviews of all the available information can reduce workloads, and prison education could be tailored to the recipient thereby reducing the likelihood of re-offending. The list, as they say, is endless.
Artificial Intelligence has been with us for a long time and its potential to transform the judicial system over the next ten or twenty years is profound. We need to embrace this change, manage it, and make it work for all of us.