Artificial Intelligence (AI) is poised to reshape the world in countless ways and in almost every field. Algorithm-based decision-making is increasingly used in the United States. It appears in pretrial risk assessments as a tool for calculating a defendant’s risk of recidivism.
Proponents argue that AI would remove bias from criminal justice, such as police, judges or prosecutors, but the theory is also contested.
A new paper by Neha Chugh, a criminal defense lawyer and member of the Board of Governors of the Law Commission of Ontario, Canada, questions the use of AI in criminal justice. Chugh argues that AI risk assessments, while not yet used in Canadian courts, raise several question marks for the judiciary to answer. Chugh says Indigenous defendants, who are already overrepresented in the criminal justice system, are particularly vulnerable to this tool’s shortcomings, according to Phys.org.
Artificial Intelligence would reflect human biases
According to lawyer Neha Chugh, using AI to perform risk assessments would simply mean that people’s biases are transferred to algorithms created by them.
“Proponents of using AI in this way shift responsibility to the algorithm’s designers,” the lawyer said.
AI is already being considered for use in some Canadian courts. Chugh admits he has reservations about how the commission has considered the use of AI in matters such as administrative judicial proceedings or by police as investigative tools.
Subjectivity, considered to be a necessary tool in the judicial system
One of the main problems Chugh identifies with over-reliance on AI for risk assessment and other considerations is the absence of subjective discretion. This, she notes, is a key pillar of an independent judiciary. The laws provide the parameters within which judges can act and which leave them some leeway while taking into account relevant factors such as individual history and circumstances.
“We call ourselves judges and decision makers based on their knowledge of the community. Do we need to outsource this decision-making process to a broad and generalized system? Or do we want to rely on a system where we have one-on-one conversations with criminals? I prefer the second option because I think courts can have a big impact on individuals,” explained Neha Chugh.
Chugh insists that he is not completely against the use of AI in the judicial system, just that more research is needed.