Title: Planning under Ignorance

Agents currently excel at making decisions under risk (aka, measurable uncertainty) which is when probabilities can be calculated or sampled. However, I am interested in agents that make good decisions under ignorance (aka, unmeasurable uncertainty) which is when the available information is too scarce to be aggregated by probabilities, e.g., due to a capricious environment. My thesis is that planners for agents under ignorance should produce strategies that conform to given criteria of good decision making from Decision Theory. In this talk, I will describe one such approach called ``best-effort planning'' that returns a non-dominated strategy, and thus, intuitively, returns a strategy that achieves its goal if the environment does not display worst-case behaviour, e.g., if the environment makes a mistake at a crucial time.