Improbable Icon


SpatialOS as a playground for testing ethical models?



We’ve heard about AI ‘playgrounds’, but what about having a similar setting for testing ethical theories?

This weekend I was having a think about how we could objectively test ethical theories such as Utilitarianism and Consequentialism. Could we have scenarios within an MMO, whereby we test the different outcomes of making decision X over decision Y?

Let me know your thoughts!


Interesting question…

I will maybe take the time later to elaborate further on this topic but in short my main two thoughts are:

  • Encoding what is considered Utilitarian / Consequantialist / other-ethical-theory as behaviour can be done on a case-by-base basis but it’s much harder to generalize as their objectives / decision-making process are not based on mathematical formulas (at least not entirely) so you would be hard-pressed as a developer to provide a sandbox world which tests the global impact of each approach. To do so you would need to make a “Utilitarian” decision which has N consequences to which you would need to react with a “Utilitarian” response, etc.

  • On a more meta-level point: how do you “test” the different theories, or put differently, how do you evaluate them beyond making purely factual observations based on your simulations. Such observations would of course be (very) helpful for decision-making processes but I’m not sure whether it would allow you to achieve a more general conclusion on one theory versus the other. From that perspective I think that testing the theories is actually similar in results to “just” running simulations based on different decision trees as it will generate the same information: i.e the addition of “ethical theories” in the mix does not change the experimentations and the analysis.

Just my 2 cents and personal opinions… and I would definitely need to take some more time to do a more profound pass over the above two points in order to modify / strengthen them.