Personalized and Diverse Task Composition in Crowdsourcing

摘要: We study task composition in crowdsourcing and the effect of personalization diversity on performance. A central process is assignment, mechanism through which workers find tasks. On popular platforms such as Amazon Mechanical Turk, assignment facilitated by ability to sort tasks dimensions creation date or reward amount. Task improves producing for each worker, a personalized summary , referred Composite (CT). propose different ways CTs formulate an optimization problem that finds most relevant diverse . show empirically workers’ experience greatly improved due enforces adequation with skills preferences. also formalize various diversifying CT. grounded organization studies have shown its impact worker motivation  [33] Our experiments contribute improving outcome quality. More specifically, we while throughput retention are best ranked lists, crowdwork quality reaches diversified requesters, thereby confirming look expose their “good” work many requesters.