The crowdsourcing movement has spawned a host of successful efforts that organize large numbers of globally-distributed participants to tackle a range of tasks. While many demand-driven crowd marketplaces have emerged (like Amazon Mechanical Turk, often resulting in workers that are essentially replace-able), we are witnessing the rise of supply-driven marketplaces where specialized workers offer their expertise. In this paper, we present a comprehensive data-driven measurement study of one prominent supply-driven marketplace -- Fiverr -- wherein we investigate the sellers and their offerings (called "gigs"). As part of this investigation, we identify the key features distinguishing "super sellers" from regular participants and develop a machine learning based approach for inferring the quality of gigs, which is especially important for the vast majority of gigs with little feedback.

Recent work on planning in incomplete domains focuses on constructing plans that succeed despite incomplete knowledge of action preconditions and effects. As planning models become more expressive, such as in temporal planning, the types of incompleteness may not only change, but plans become more challenging to evaluate. The primary difficulty to temporal plan evaluation is accounting for temporal constraints that may not be satisfied under all interpretations of the incomplete domain. In this work, we formulate incomplete temporal plan evaluation as a generalization of the temporal consistency problem, called partial temporal consistency. We present a knowledge compilation approach that is combined with symbolic constraint propagation and model counting algorithms for counting the number of incomplete domain model interpretations under which a plan is consistent. We present an evaluation that identifies the aspects of incomplete temporal plans most impact performance.

Engineering complete planning domain descriptions is often very costly because of human error or lack of domain knowl- edge. Learning complete domain descriptions is also very challenging because many features are irrelevant to achieving the goals and data may be scarce. We present a planner and agent that respectively plan and act in incomplete domains by i) synthesizing plans to avoid execution failure due to ignorance of the domain model, and ii) passively learning about the domain model during execution to improve later re-planning attempts. Our planner DeFault is the first to reason about a domain’s incompleteness to avoid potential plan failure. DeFault computes failure explanations for each action and state in the plan and counts the number of interpretations of the incomplete domain where failure will occur. We show that DeFault performs best by counting prime implicants (failure diagnoses) rather than propositional models. Our agent Goalie learns about the preconditions and effects of incompletely-specified actions while monitoring its state and, in conjunction with DeFault plan failure explanations, can diagnose past and future action failures. We show that by reasoning about incompleteness (as opposed to ignoring it) Goalie fails and re-plans less and executes fewer actions.