MoralAgreements

As machines are becoming increasingly independent through rapid advances in artificial intelligence (AI) technology, questions arise about how ethical principles should guide their moral behavior. One major challenge involves the collection and assessment of social expectations to how these questions should be solved. In this paper, we propose a framework for collection and measurement of public opinion on ethical dilemmas. Using our framework as a tool, researchers can add their own collections of dilemmas, making it simple to investigate public opinion ondilemmas of their own choosing. In other words, the framework is a plat-form for crowdsourcing both public opinion on moral dilemmas as well as moral dilemmas. We describe how gamification can be implemented to improve the experience of participating and encourage participationin our tool. In addition, we demonstrate a method of assessing social expectations of AI ethics through comparison of collected data with existing moral theories. The capabilities of our framework is demonstrated through the deployment of a working prototype that implements gamification, compatibility with five moral theories, and a set of ten ethical dilemmas.

Prototype: https://moral-agreements.herokuapp.com/ (pin code "0001")