Making machines that reason and act morally on the basis of causation rather than mere correlation
The more important decisions we leave to artificial intelligence, the more important it is that the decisions are sensitive to the moral dimensions involved and that we understand how and why they are so.
Plausibly, real morality deals in causation rather than mere correlation. As a first step, we try to develop systems that can play moral games and solve simple moral problems by virtue of reasoning in terms of causation rather than mere correlation. By utilizing the Tsetlin-machine developed by CAIR, which works on the basis of simple propositional logic, we also aim to make the moral reasoning transparent and understandable to us humans.