Artificial Intelligence in Education as a Rawlsian Massively Multiplayer Game: A Thought Experiment on AI Ethics

AbstractIn this chapter, we reflect on the deployment of artificial intelligence (AI) as a pedagogical and educational instrument and the challenges that arise to ensure transparency and fairness to staff and students . We describe a thought experiment: ‘simulation of AI in education as a massively multiplayer social online game’ (AIEd-MMOG). Here, all actors (humans, institutions, AI agents and algorithms) are required to conform to the definition of a player. Models of player behaviour that ‘understand’ the game space provide an application programming interface for typical algorithms, e.g. deep learning neural nets or reinforcement learning agents, to interact with humans and the game space. The definition of ‘player’ is a role designed to maximise protection and benefit for human players during interaction with AI. The concept of benefit maximisation is formally defined as a Rawlsian justice game, played within the AIEd-MMOG to facilitate transparency and trust of the algorithms involved, without requiring algorithm-specific technical solutions to, e.g. ‘peek inside the black box’. Our thought experiment for an AIEd-MMOG simulation suggests solutions for the well-known challenges of explainable AI and distributive justice.