Reading time: four minutes
Artificial intelligence (AI) has been more prevalent in our lives than ever before - self-driving cars, automated clinical diagnosis and credit card application approval all incorporate a significant degree of AI.
Its apparent “omnipotence” makes me wonder:
Most AI applications are designed for particular tasks (such as AlphaGo for playing Go) requiring mostly definitive answers (which move can maximise my probability of winning the game?). Such is not the case in law. Ambiguity, challenges and debates are inherently pervasive in legal practice and academia.
To understand how AI might help us, we must first define what AI is and what it encompasses.
First, we need to accept that the current state of AI is far from general human intelligence - it might beat us at Go, but it cannot “feel” emotions, “act on a whim” like we sometimes do, or even butter a toast unless it is explicitly programmed to do so.
What an AI programme can accomplish might seem impressive, but in reality, it merely depends on massive sequences of programmed, elementary calculations - what computer scientists call algorithms.
Machine learning is one of the most significant branches of AI that specialises in extracting patterns from extensive collections of data. A simplified example is determining the probability of a criminal reoffending by comparing them to a collection of past criminal offences in the same geographical area.
Next, we will look at some current implementations of AI in law and analyse their advantages, risks and limitations.
We will start with using AI as legal practitioners, more specifically, solicitors and barristers if we are in the UK. One example is predictive coding models, which are trained on keywords and documents relevant to the litigation at hand and subsequently capable of automatically identifying other similar ones like legislation and law reports among databases containing millions of entries.
Such document review efficiency is impossible for humans; as a result, more time is saved so that lawyers can focus on other tasks such as strengthening client relationships.
Recently, lawyers’ interest in AI for predicting case outcomes has been constantly growing. Traditionally, lawyers weigh the significance and persuasiveness of all their arguments to determine the best course of action for their clients.
Now, they can consult past cases and base their evaluation on enormous amounts of historical data rather than their instinct alone. It is important to stress that AI cannot replace lawyers’ years of experience and extensive understanding of their clients; therefore, they cannot rely on it too much. But how much is “too much”?
Now, we will explore the application of AI for legal administrators. They are different from solicitors and barristers because their primary responsibility is to make decisions; therefore, judges and public authorities all fall within the definition of “legal administrators”.
Imagine that you were a judge who had to decide whether to release an alleged criminal on bail pending trial.
During your risk assessment, you might consider the following questions:
No one can give a definitive, binary, “yes-or-no” answer to all these questions - except, perhaps, an AI tool. Typically, it can learn from a massive amount of similar cases in the past and give a quantitative probability to each question.
Although judges are in no way bound by the results from these calculations, they could be subconsciously affected. Maintaining the independence and impartiality of the judiciary in the democratic society will once more become a challenge when AI tools are implemented more widely in courts.
We see that AI tools can certainly enhance the work efficiency of both legal practitioners and administrators to the next level.
However, careful and extensive regulation is necessary to ensure that humans still have the final say and that we continue to uphold the rule of law, one of the core values of a democratic society.