SAFE Working Paper No. 335

Artificial Intelligence, Ethics, and Intergenerational Responsibility

In more and more situations, artificially intelligent algorithms have to model humans’ (social) preferences on whose behalf they increasingly make decisions. They can learn these preferences through the repeated observation of human behavior in social en- counters. In such a context, do individuals adjust the selfishness or prosociality of their behavior when it is common knowledge that their actions produce various externalities through the training of an algorithm? In an online experiment, we let participants’ choices in dictator games train an algorithm. Thereby, they create an externality on future decision making of an intelligent system that affects future participants. We show that individuals who are aware of the consequences of their training on the pay- offs of a future generation behave more prosocially, but only when they bear the risk of being harmed themselves by future algorithmic choices. In that case, the external- ity of artificially intelligence training induces a significantly higher share of egalitarian decisions in the present.