Entries for How would you teach AI to be kind?

Abstracting a script and labeling the narrative with human emotional responses by observing facial expressions of people watching the movie.

Games have been used to accelerate breakthrough in genetics research like with Phylo. Why not have a game of a bunch of ethical puzzles?

The proposed mechanism uses a feedback loop and the Decision Balance to program machines to implement more appropriate actions.

We introduce EmpathyNet, a dataset of video/image, audio and text cues, which can be easily compiled from existing video repositories.

A dataset built based in characters of books and movies can help to identify adjusted good behaviour by time and place.

We have to work to make artificial intelligence not just intelligent, but focused on the right data sets to be ethical and good. –Mo Gawdat

A little know science: Axiology. This science is based on discoveries 50 yrs ago that linked Value Judgments with a mathematical system.

The idea is to use crowd intelligence to build the dataset of human ethics behaviors.

The data sets of AI pro-social behaviors, needs a structure, a way to be rated and, they need to be usable for the largest number of cases.

A AGI that leans to be nice form A nice person. By way of switching position. That uses virtual reality to swap position with teacher.

We envision EthicsNet as a framework in which developmentally-situated AI systems can learn from humans about behavioral norms.

This solution allows anyone to quickly and easily contribute to AI ethics, and the rise of kind AI systems: Concept Reference Model

Toulmin arguments that use Aristotle's pathos, ethos, and logos can choose social norms and laws via semantic graph similarity.

The Golden Rule (Law of Reciprocity) as higher level variant of mutual cooperation in The Prisoner's Dilemma game; a basis for AI ethics.