'Indifference' methods for managing agent rewards
This work addresses control challenges in agent design, but it appears incremental as it builds on existing indifference concepts without claiming major breakthroughs.
The paper tackles the problem of controlling reward-based agents in POMDPs using 'indifference' methods to achieve goals like event-dependent rewards without manipulation, effective disbelief, and seamless reward transitions, presenting several methods with their uses and strengths.
`Indifference' refers to a class of methods used to control reward based agents. Indifference techniques aim to achieve one or more of three distinct goals: rewards dependent on certain events (without the agent being motivated to manipulate the probability of those events), effective disbelief (where agents behave as if particular events could never happen), and seamless transition from one reward function to another (with the agent acting as if this change is unanticipated). This paper presents several methods for achieving these goals in the POMDP setting, establishing their uses, strengths, and requirements. These methods of control work even when the implications of the agent's reward are otherwise not fully understood.