Misunderstanding AI: What machine-learning really means for the law office of the future

  • November 29, 2018
  • Sean Lynch

Lawyers are bombarded with commentary that artificial intelligence (AI) is going to either: 1) eliminate the need for many lawyers; or 2) make life so much easier for lawyers that their workloads will be drastically reduced. It smacks of the early advertising campaigns for smartphones: no need to be in the office at all; you can do all of your work from your phone while on the golf course or at the cottage, and so on.

Dr. Paola Cecchi-Dimeglio, behavioural scientist, Senior Research Fellow at Harvard Law School and chair of its Executive Leadership Research Initiative for Women and Minority Attorneys at the Center for the Legal Profession, summarizes that “much of what constitutes AI remains somewhat misunderstood” and that this misunderstanding is problematic as it may limit the benefits that AI has to offer.[1]

As with the hyperbole that accompanied the smartphone, the reality of AI is far from both the rosiness and the disaster theories being flung out into the legal market consciousness. Part of the problem with understanding AI is definitional. What, exactly, is AI?

Many people, if not most, have heard of analytics. Analytics is essentially a way to organize data in a way that allows for the discovery of patterns within that data. We, as consumers, contribute to the cloud of data that analytics parses everyday, from rating a show on Netflix to swiping a points card at a store. This data is analyzed to identify behavioural patterns. These patterns are later employed to suggest Netflix shows that you may enjoy or to show advertisements that dovetail with your shopping habits. To be clear, the use of analytics here is not “predictive”. Rather the technology determines what is “likely”.

So, is this AI? Some would say yes, others would say no. Any advanced analytics system that provides information or insight from data, with limited human involvement, is now being referred to as Artificial Intelligence. The problem is that all of the currently available “AI” systems necessitate human interfacing to provide those insights. Humans set the parameters and teach the system what is up, down, good, and bad. These systems cannot respond to a question unless you have already told it the answer.

Take our Netflix example above. The system is fed user data whenever a rating is applied. The system was programmed to recognize that a five-star rating signifies enjoyment, while a one-star rating signifies dislike. Each Netflix film or TV show is assigned many data points, including theme, length, and actors. By taking these attributes into account, the system identifies content resembling the highly-rated shows or movies. With each rating provided by a user, the system identifies content with similar attributes that, based on the ratings, the user may like or dislike. Netflix will then promote these similar shows because it presumes that this content will be enjoyed by the user. In the end, this is a simple mathematical relationship: the system prioritizes shows with data points in common to any show rated highly by the user. Such a process is not typically regarded as “intelligence”. Rather, this process that involved the “Teaching” of a system is called Machine Learning.

Machine learning – the training of a computer analytics system by a human being – is at the core of AI. After completion of the training process, which typically requires the input of data and human feedback, the system applies this training to make inferences about new data. In the legal context, these inferences translate to significant cost savings. For example, these systems hasten due diligence exercises that arise with many transactions or, with regard to litigation, allow for the effective, and defensible, reduction in the number of documents that need to be reviewed by counsel.

What is critical to acknowledge about AI and the legal industry is that, while AI is already in use, the technology is in its infancy. It will evolve along with the legal marketplace. Current lawyers, as well as future ones, should educate themselves as to what this technology can offer, how it can assist clients and to what extent individual firms are adopting it. Advanced analytics/AI are not going to take legal work away from lawyers. The technology will allow lawyers, and firms, to do high-value legal work and bolster client satisfaction. In the end, the goal is to provide the best results for the client at the lowest cost. AI technology will support that goal and usher in the next iteration of legal professional services that see technology as a companion, not a threat.

 

About the author

Sean Lynch, B.A., J.D., LL.M. is director, review services, at Ricoh eDiscovery.

 

[1] Thomson Reuters Legal Insights, “Ask Dr. Paola: how is AI changing the legal industry?”, Thomson Reuters, https://blogs.thomsonreuters.com/legal-uk/2018/02/07/ask-dr-paola-ai-changing-legal-industry/, (11 October 2018).