Aarhus University Seal / Aarhus Universitets segl

Respect for Simulated Persons

Frodo Podschwadek, University of Glasgow, UK

Strong AI capable of meaningful social interaction with humans is, as of now, an unrealized possibility.  Chances are good that this will not change, as the trend away from general purpose computing and toward either highly specialized and/or distributed systems might not offer sufficient incentives to develop strong general AI with social capacities.

However, there remains a plausible version of AI capable of interaction with human beings in socially meaningful ways, mainly because they were constructed to do exactly that. AI could be employed, for example, as simulations of existing (or once existing) persons.

With sophisticated versions of such simulated persons, it would be difficult to draw a principled line between simulated and natural, or “real”, persons in any other way than on the basis of the physical substrate of their personhood. One of the questions implied by this partial indistinguishability between simulated and natural persons would be: should society assign a moral status to simulated persons, either as moral patients or even moral agents?

In this paper I am arguing that there are good pragmatic reasons to assign simulated persons at least a limited moral status by sketching the conditions under which society should be willing to accept artificially intelligent entities as members of the community of social morality.