Novel Methods for Designing Tasks in Crowdsourcing

Crowdsourcing is becoming more popular as a means for scalable data processing that requires human intelligence. The involvement of groups of people to accomplish tasks could be an effective success factor for data-driven businesses. Unlike in other technical systems, the quality of the results depends on human factors and how well crowd workers understand the requirements of the task, to produce high-quality results. Looking at previous studies in this area, we found that one of the main factors that affect workers’ performance is the design of the crowdsourcing tasks. Previous studies of crowdsourcing task design covered a limited set of factors. The main contribution of this research is the focus on some of the less-studied technical factors, such as examining the effect of task ordering and class balance and measuring the consistency of the same task design over time and on different crowdsourcing platforms. Furthermore, this study ambitiously extends work towards understanding workers’ point of view in terms of the quality of the task and the payment aspect by performing a qualitative study with crowd workers and shedding light on some of the ethical issues around payments for crowdsourcing tasks. To achieve our goal, we performed several crowdsourcing experiments on specific platforms and measured the factors that influenced the quality of the overall result.