Otago University’s artificial intelligence research will help shed light on the effect of AI innovations on law and public policy in New Zealand, including a look at some potentially harmful implications from the powerful and fast-developing technology.The Otago initiative, Artificial Intelligence and Law in New Zealand, is funded in part by the Law Foundation and is a three year multi-disciplinary project involving the Law Faculty and the Philosophy and Computer Science faculties.

Artificial Intelligence & The Law: Will It Take Us To Dark Places Too? — Associate Professor Colin Gavaghan (centre) is joined by Associate Professor James Maclaurin (left) and Associate Professor Alistair Knott (right).

While there is breathless excitement over the potential for AI to transform our lives, including a raft of legal applications that increases daily, the ethical and other implications are also important.

Project team leader is Law Faculty Associate Professor Colin Gavaghan (pictured above and right) who recognizes the fascinating practical and ethical implications of artificial intelligence, and the potential dangers they pose.

The Public Use of AI

Part of the focus of the research is to examine the use of predictive analytics or algorithms in the public sector, which are already being used in the New Zealand criminal justice system, immigration, IRD and ACC.

Their use and continued development are some of the issues to be examined and Professor Gavaghan is quick to point out that they are not developments that he is opposed to, noting that such tools are particularly well suited to processing large quantities of data in a manner well beyond what most human brains can accomplish.

“But there are also concerns about the use of algorithms in these contexts. Some of these relate to the possibility of bias creeping into the system – either through the use of biased training or input data, or by some other means.

“This wouldn’t need to be deliberate, but without detailed and fairly regular checking, it’s a legitimate concern. Just think of the concern in the US about the use of the COMPAS tool in criminal justice decisions, and the revelations that it was far more likely to be predict that black prisoners would re-offend than their white counterparts,” he told LawFuel.

Transparency Issues

Transparency is another issue of concern to the team. “How can you challenge a decision if neither you nor anyone else understands the basis on which it was made?

“This might not be such an issue with the sorts of tools being used at the moment, where the variables and weightings are fairly visible, but as we move down the road into the machine learning proper, it could become extremely difficult for anyone to really understand what’s going on inside the ‘black box.'”

The question of how to handle these ethical and legal challenges are those confronting other jurisdictions. The research has seen the team speaking to leading figures in AI issues in the EU and US as to what they suggest should be done to handle them.

“The challenge, then, is how to get the best out of using these tools while avoiding the possible pitfalls.” Among the suggestions –

A public register of goovernment algorithms;

A ‘right to an explanation’ of algorithmic decisions made about you (in NZ, the Official Information Act already contains something like this, but it isn;t quite clear how it would capture ‘black box’-style algorithmic decisions);

A right to have a human involved in any decision that affects us;

A policy decision not to automate certain kinds of decisions.

“We think there are probably pros and cons for all of these measures,” Professor Gavaghan says.

“For instance, the idea of always having a ‘human in the loop’ might sound reassuring, but there are doubts as to how well human beings can operate alongside high-tech systems. Will we become subject to ‘automation bias’, deferring to the machine rather than properly scrutinising it? Will we suffer from ‘decisinal atrophy’ when we’re only rarely called upon to exercise certain skills?”

The issues may be vexed, but they are also vital. A ‘solution’ of some kind is necessary. “Our present suspicion is that some combination of these measures might be needed to alleviate the concerns around automated decisions, but we’re still a bit away from making our final recommendations.

We’ve been very pleased, though, to be approached by Government with a view to providing expert advice on this issue.”

The discussion on artificial intelligence and its ethical and other issues will continue to exercise human brains for some time – maybe until the robots do take over.