How Artificial Intelligence Can Improve Human Moral Judgments

How Artificial Intelligence Can Improve Human Moral Judgments

Walter Sinnott-Armstrong (Duke University)

Walter Sinnott-Armstrong argued for the attractive idea that artificial intelligence can be used to help us make better moral judgments. For example, complicated cases can confuse us, and we can be led astray by emotion, bias, order effects, and so on. But what if AI could learn to predict the moral judgments that we’d make if we weren’t limited in this way? Sinnott-Armstrong described a project he’s working on with Joshua August Skorburg and others to design an AI that could do this for decision-making about kidney exchanges. In a kidney exchange case, someone needing a kidney might have a willing, but incompatible, donor and can trade the donor’s kidney for a compatible one. Add enough people requiring kidneys and wanting to make these sorts of trades, and deciding who gets kidneys quickly becomes a highly complex problem requiring AI algorithms to solve. Sinnott-Armstrong described a method for eliciting human judgments about which attributes matter in these decisions (e.g. age, health, etc.), how much weight they should each be given, and how to aggregate and translate this information to improve kidney exchange algorithms. He argued that methods like this like could be used in designing autonomous vehicles and weapons systems, and could also be used to improve human moral judgments in many areas of ethics.Walter Sinnott-Armstrong is Chauncey Stillman Professor of Practical Ethics at Duke University in the Philosophy Department, the Kenan Institute for Ethics, the Duke Institute for Brain Science, and the Law School. He publishes widely in ethics, moral psychology and neuroscience, philosophy of law, epistemology, philosophy of religion, and argument analysis.

This public lecture was held at the Kambri Cinema, ANU on 26 June 2019.