Machine learning has aided legal prediction for over fifty years

Evidence

Machine learning has been used for legal prediction for more than fifty years (see, for example, Mackaay & Robillard, 1974). While most of the uses have been academic or have involved private sector applications, there is one major area of concern in public sector use: statistical algorithms that have been employed in risk assessment. For instance, in our home state of New Jersey, there is a pretrial assessment system designed to algorithmically generate “risk” scores that indicate whether a defendant should go free or be detained in jail while awaiting trial (Hanna, 2018; O’Dea, 2018). — Joseph Avery, Jim Sidanius, Joel Cooper, Gregory Davis, Amanda N., Bias in the Law, 1163

Categorization

• QUESTION: What did Mackaay and Robillard’s 1974 system actually predict, and how accurate was it? Early-history angle on legal AI worth chasing.