In an academic paper published this week, our founders explore Oliver Wendell Holmes, Jr.’s idea that “for the rational study of the law, the black-letter man may be the man of the present, but the man of the future is the man of statistics and the master of economics.” This was written in 1897, but represents the evolution currently underway in law. Statistical methods based on machine learning allow us to use data from past cases to make predictions about how a court might decide in a new scenario.
This week’s newsletter draws out some key points of interest from the paper. We outline foundational AI concepts, how we apply machine learning here at Blue J Legal, and pose thoughtful questions as to what AI means for law in the future. If you’re interested in reading more, the recent paper and other thought-provoking resources are listed below.
What are natural language processing and machine learning?
While “artificial intelligence” is a nebulous concept that changes over time, two specific subfields or applications of AI are easier to define:
|Natural Language Processing||Machine Learning|
|What does it do?||Improves retrieval and extraction of information by drawing connections within and across language||Improves predictive accuracy regardless of underlying theory|
|How does it work?||
|What was the traditional method?||Keyword searches, which are literal in their approach and look for exact words or phrases||"Brute force" computation in which the programmer specifies the relationship between variables (i.e., linear regression)|
|What is the advantage over the traditional method?||Allows the user to identify relevant materials, even if they do not contain exact keywords||Leverages connections between and among references, even if they are implied rather than expressed|
|What are some common applications?||
How does Blue J Legal apply machine learning?
Our Classifiers apply machine learning to predict how courts would decide legal questions. Our data derives from published Supreme Court of Canada, Federal Court of Appeal and Tax Court of Canada decisions, in which the court makes a binary determination for a specific legal question. A brief overview of how we use machine learning to develop classifiers is as follows:
- Identify a fact-intensive question of law.
- Determine the factors most relevant to the court when deciding the question.
- Code every published decision in accordance with these factors. This process turns unstructured data (the text of judicial opinions) into structured data (discrete information from the factors in the form of variable data).
- Apply machine learning technology to generate a predictive algorithm. The algorithm identifies connections among the variables that are difficult to specify using more traditional statistical techniques.
What does the future of law look like? Short term (evolution):
- Increased transparency
- Increased efficiency
- Deeper expertise
- Broader expertise
- Greater access to justice
- More value for clients
Long term (revolution): Will AI change the way the law is produced?
- Will AI change the way law is consumed?
- Will AI increase access to civil and criminal justice?