A Definition of Happiness for Reinforcement Learning Agents
This work addresses a conceptual problem in AI and psychology, but it is incremental as it adapts an existing concept to a new context.
The paper tackled the problem of defining happiness for reinforcement learning agents by proposing temporal difference error as a formal definition, which satisfies most desiderata and aligns with human empirical research.
What is happiness for reinforcement learning agents? We seek a formal definition satisfying a list of desiderata. Our proposed definition of happiness is the temporal difference error, i.e. the difference between the value of the obtained reward and observation and the agent's expectation of this value. This definition satisfies most of our desiderata and is compatible with empirical research on humans. We state several implications and discuss examples.