How Will AI Affect the Law?

by Tom Beattie

(source: Wiki Commons)

Artificial intelligence (AI) refers to the development of computer systems able to perform tasks that would otherwise require human intelligence. There is an important distinction to be made between two forms of AI. Firstly, machine learning which is responsible for a lot of the computer systems we have in place today. Secondly, true artificial intelligence, whereby a computer or hypothetical machine can evince behaviour at least as skilfully and flexible as humans do. Machine learning refers to the use of algorithms to make a computer learn outcomes and create predictions, this is AI in a basic form and represents a more accurate picture of what we have in today's society. The short term goal of AI, through a corporate lens, is to alleviate repetitive and simple tasks where possible, and to aid humans in order to create a more optimal outcome. The prospect of artificial intelligence in law is as challenging as it is stimulating. The primary goal being to aid in a new, more advanced, form of legal reasoning, however this integration challenge becomes apparent when we consider the attributes or characteristics of the law. This is a subject that entails various forms of reasoning, regulated by traditions involving difficult concepts that are often ambiguous, understanding important and accurate legal knowledge and of course integrating the difficult legal rules and concepts into a new and sophisticated argument. The successful integration of a rigorous and intellectual subject with AI could be one of the greatest and progressive reforms in the field of law today.   

Whilst it is advantageous to consider and speculate possible future applications, dangers and regulations of AI, it is equally important to understand the practical uses in the current legal field. One obvious practical challenge in being a lawyer is the volume of documents, contracts and research needed. AI powered software has the ability to analyse, scan and produce or even critique thousands of documents quicker than any human could. This works primarily via machine learning; that is that once a certain type of document or a piece of content is marked as relevant, machine learning or algorithms can find documents that are similarly relevant. However, where this could be best used is by flagging documents that are questionable and may need a human to verify or check, dramatically decreasing the workload and volume and a useful service in a law firm. Once enough data has been given for the machine to ‘learn’, it is even feasible to suggest that a machine could write a customised and accurate contract autonomously. Much of a working solicitor's time is spent conducting due diligence, the investigation or research that a professional is expected to carry out before entering into an agreement or a contract. This involves validating facts and figures and researching prior cases to give the best counsel to clients. Again AI can aid legal professionals in accurate and fast research in this respect as this task is often mundane predictable and a capable task for machine learning. Law firms are often required to check over contracts and identify possible risks and faults that could have adverse effects for their clients. As mentioned within due diligence, AI can help analyse contracts in large quantities. These are some of the more basic uses of AI today; alleviating monotonous tasks. However, there are more complicated uses of AI that could come to fruition in the future. 

It is possible that after a significant period of machine learning, using previous cases that have been marked as successful or otherwise relevant, machines could start predicting the outcomes of other legal cases. If the machines have been given a large amount of data, then they have the ability to, through algorithms, analyse and predict the outcomes of various cases by using similar cases as a model modelling. Having a repetitive task completed without having to consider the opportunity cost of your own time is invaluable. According to the Huffington Post the typical price of a divorce “in theory could cost less than £1500 in total” however, they also note that “costs will range from say £3,000 to £25,000 in disbursements for average divorces where matters are agreed, but can rise significantly beyond that where financial proceedings are issued and contested”. An independent survey by Aviva Family Finance estimates average marital breakdowns costs at £14,500. “Wevorce” is an online computer powered software that provides couples with guidance to avoid the costly contests. Couples simply define or enter their desired outcome into the system and it will guide them to this optimal outcome. There are also legal experts who can step in if assistance is required, making client acquisition that much easier. 

Highlighted above are some fascinating practical automation within law firms. However, such a dramatic change does come with potential pitfalls. Cyber security and attacks are one of the major threats of modern society and data breaches concerning sensitive information in a confidential and sensitive area such as law has the potential to be seriously damaging. As noted in ‘iberianlawyer’, “there are two types of companies, those that have suffered a cyber attack or those that will''. Seeing this danger, it is imperative that firms create a strong protective structure and employ those who are used to working in conjunction with fast paced technology. They go on to assert that “AI should not be used to replace lawyers” but rather to complement them in decisions, as AI can often identify risk better.

There could also be potential adverse effects on the lawyer-client relationship as clients may find it challenging to interact with or through technology. However, it has been shown that clients are more likely to tell the truth to a machine during an interview as it is not capable of displaying judgement. The use of AI will reduce the time spent carrying out tasks, so not only will clients expect to pay less, they can expect a quicker service. This supports the argument that technology should be seen as a tool to aid lawyers in their work but not replace them. A further barrier to the integration is the demand for and from younger lawyers. Whilst there may be some reluctance from smaller firms to adapt to a new system, many firms have experienced negative effects on their take-up due to law schools failing to teach or at least introduce new technology. This is in defiance of the fact that it is young lawyers who are actively calling and more likely to embrace the change. It is, therefore, imperative that law firms continue to train their current staff on new tools that are available to them. Firms that chose to do this are more than likely going to be seen as more attractive to young lawyers as employers. Finally, there are prominent issues surrounding privacy as advanced AI compromises this.

Technology already tracks and predicts individuals’ shopping preferences, political bias’ and even time spent in various locations. This data is often accumulated and shared between different companies and as a result different platforms. AI is now starting to tackle more controversial issues by predicting sexuality and propensity to commit a crime; is this an invasion of privacy? 

Moores Law was defined in 1965 and says that computing power will double every two years. It has been remarkably accurate to date and every year the world produces more digital data than in all previous years put together. With computer systems developing at a faster rate than society it begs the question, how do we regulate such change? Not only does AI compute faster than society, it also changes faster and soon it could be adapting and evolving autonomously. This means that AI is progressing faster than the government itself and therefore faster than societal infrastructure. It is already true that large companies such as Facebook and Google collectively own more data than the government (UK), and it is certainly true that data is influence and influence is power. So how do we regulate AI? When the first case of this comes around lawyers will enter uncharted territory and may be trying cases in front of judges who possibly do not understand the technology that they are asked to rule on, with no prior cases for reference. Seemingly simple questions in the context of today's legal field will be reinvented with an extra layer of complexity. For example, who is at fault?  Who is responsible for the civil or criminal wrong and was this out of negligence or ignorance? These are all questions fundamental to many types of Law.

Globally governments are under more pressure to operate more efficiently given the increase in population. The concepts that have been applied above to the private sector, show that AI or automation at scale has the potential to help. This does require the government to be willing to embrace new technologies. The McKinsey Global Institute, a technology think tank, stated that AI and big data are not only contributing to the transformation of society but, when compared to the Industrial Revolution, is “happening ten times faster and at 300 times the scale, or roughly 3000 times the impact”.

Automation could create several key advantages that could improve the efficiency of government including much lower operating costs, more efficient processes, and less wastage and errors. McKinsey estimates that “as many as four out of five processes in HR, finance and application processing are at least partially automatable, with the potential to reduce cost by at least 30 percent”. This is a feasible suggestion as the benefits of AI automation can come to fruition relatively quickly. Many solutions or improvements can be built upon current IT systems, it is just a question of scale. Whilst the government has moved several key surveys online now such as “The Census” and UK passport and driving licence application and tax returns, it is now a question of using AI to automate internal processes and create a government that is digital throughout. Whilst there has been digitalisation, automation of more central processes has shown to be more challenging. In Finance, Human Resources (HR) and procurement Mckinsey says that “60 to 80 percent of tasks are automatable, creating potential for net long-term savings (after accounting for implementation and ongoing software costs) of at least 30 percent”. 

Within Finance, the government conducts many of the same finance processes as private-sector firms. For example, they disperse and lend cash, conduct budgeting and complete financial planning and analysis. In the UK 11,000 civil servants work on finance across 25 departments. As outlined, up to 80 percent of finance roles in the private sector have seen at least some level of automation and a similar level can be replicated in government. Mckinsey goes on to say that “one large European utility piloted the automation of new vendor creation... and found the processes were 70 percent automatable.” 

The government is the largest employer in the country. 80 percent of HR processes are at least somewhat automatable, with payroll administration, record keeping, benefits administration, and recruitment administration-functions. In the private sector, one large energy provider was able to automate 90 percent of its on-boarding process, including ordering and delivery of passes, phones, and office equipment ready for day one, leading to more than 20 percent cost savings. 

Finally, the government processes many applications and claims for a range of reasons including payments and services, social welfare, visas and tax returns. Whilst there has been great reform in terms of citizen facing systems they are often met without an out of date back end, hindering efficiency. Automation can help by reading and writing data between applications, checking consistency and completeness, and even sending and interpreting emails. One insurance firm used software to automate subrogation claims processing and reduced time per claim from 10.0 minutes to 3.5 minutes, increasing the volume of claims processed per week by a third. 

Above we have considered how AI could rationally affect the efficiency of the government. As mentioned before, AI could be used in courts, but could it affect high court or even supreme court? Lord Hodge, in “Law and Technological change”, references the first time an English court publicly approved the use of predictive coding software. He says that “the case concerned alleged breaches of directors’ duties in the hotel and leisure industry, where over three million documents had to be considered for relevance and possible disclosure”. He then explains that, for the purpose of disclosure, The High Court considered whether the parties could rely on predictive coding, which would fall under ‘machine learning’ as a form of AI defined in the introduction. It would take data input by humans about what is deemed relevant and then apply it at scale to thousands of documents. As a result, the acknowledgement came that predictive coding software offers greater consistency in the review and flagging of documents than manual review conducted by hundreds of junior lawyers. They also concluded that it was more economically viable. It was estimated to cost in the millions for a full manual review compared to £500,000 for predictive coding software. Therefore, they concluded that it was a suitable case in which to use technology. Indeed the courts have shown a willingness to embrace new technologies, the Lord Chief Justice, Lord Burnett of Maldon, has set up a new advisory body to make sure the judiciary of England and Wales is fully informed about developments in AI that could potentially manifest inside the courtroom.

So what does AI mean for future lawyers? In the near future replacing litigation lawyers seems nearly impossible. AI will simply remove the mundane functions of a lawyer and free up more time for the individual and ultimately the firm to perform more challenging tasks. It is also hard to support the idea that AI will imminently replace advocates in court. The role of a Barrister is  highly personal and involves a clear understanding of specific case details and key personable and argumentative skills. Whilst AI may be able to achieve this in the future, lawyers won’t be entirely replaced soon. Furthermore, law and litigation is based on the English language which can itself be notoriously ambiguous in nature. Whilst the law has an array of built-in mechanisms for removing interpretive disagreements, there is no doubt that words and phrases can have different meanings depending on the context and intention. Despite built-in safeguards, so called rule uncertainty or imprecision in the language of the law can be an instrument of injustice and whilst using the same conventions for ascertaining meaning, the most experienced lawyers and judges can disagree. 

Due to the complexity of modern legislation, it is difficult to believe the suggestion that humans will not be needed in at least some capacity to debate meaning. A good barrister can create empathy with a jury and lead them down a train of thought, arguably a trait that is at the moment unique to humans, therefore making us indispensable to legal deliberations. It is a lawyer's job to complete a legal task as quickly, effectively and inexpensively as possible, and the use of AI will give lawyers the opportunity to satisfy this task to a greater extent. Many have speculated that this will result in lawyers not being able to bill their client for as many hours and on the surface this seems true. However, a quick and effective service is likely to lead to repeat business and in turn more clients. The result is that firms will be able to increase their revenue whilst boosting their client portfolio. Overall, AI will undoubtedly replace jobs, an example being the call for AI to relieve clogged up court systems. AI has been used to determine eligibility for bail by detecting behavioural patterns within an individual and to determine flight risk. This is a task that a judge would traditionally undertake.