Meeting the Ethical Challenge of AI

For AI to be the revolution we want, we need ethics by design

One of the buzz-phrases in the current privacy debate is “privacy by design” or even, “privacy as a design experience”. The essence of this, logically enough, is that privacy should be “designed-in” to a service or product from the outset, rather than have to be bolted on afterwards once the implications have become clear. Although this is, in one sense, obvious, it does require a change of mindset; it is not instinctive for designers focused on user experience to build in anything that might compromise that same experience.

Now Artificial Intelligence is coming at us carrying some very similar baggage, and with wider societal implications. How does a healthcare system value a particular treatment for different individuals? How does a driverless car balance the risks to passengers against other road users? Machine learning might enhance efficiency and effectiveness, but, as AI experts point out, pattern matching is not the same as reasoning. In areas like this, “ethics by design” will be not a buzz-phrase, but a necessity.

Automotive engineers and roboticists are already collaborating with philosophers to build “ethical” algorithms for driverless cars. If AI is to change the world in the way that we may hope, rather than fear, then the question of ethics must be answered across every aspect of its use. In both the commercial and public square, policymakers are going to have to get used to incorporating it. Time to brush up on your moral philosophy!