Postgraduates work in progress

Speaker: Nick Novelli

Title: Beyond the Turing Test: How to identify morally-relevant mental properties in A.I.

Abstract: As Artificial Intelligence technology advances, it becomes increasingly important to know when these machines have attained mental properties that would grant them moral standing. The best-known test for A.I. mental properties is the Turing Test, but the property it purports to test for, intelligence, is not particularly important in any commonly-accepted moral theory. Rather, phenomenological properties such as capacity for pleasure and pain, emotions, and desires are more likely candidates as morally significant mental attributes. We have evidence that these states perform the role of enabling complex learning and social/moral reasoning in humans and certain other animals. The task, then, is to devise tests where these properties are crucial to the success of humans and whichever non-human animals that can pass, and where passing in a way that does not involve these properties would be sufficiently impractical. In this way, we can at least have enough guidance to use in our practical moral decision-making. I will outline a number of such tests and their advantages and drawbacks.