022

Title:

Optimization Opportunities in Human-in-the-loop Computation

Discipline: Computer Science

Presenter:

Dong Wei

Abstract:

The rapid development of social networks and online labor markets is making an increasing number of individuals to rely on such networks to find jobs. Online job marketplaces are gaining popularity as mediums to hire workers to perform certain tasks. On those platforms, workers can find temporary jobs in the physical world or in the form of virtual "micro-gigs" such as help with HTML, JavaScript, CSS, and JQuery. Crowdsourcing platforms are a very popular type of online job marketplaces nowadays. These platforms are fully virtual: workers are hired online, and tasks are also completed online. Examples of crowdsourcing platforms are Amazon Mechanical Turk and Figure Eight in the USA. All these platforms could broadly be categorized as "human-in-the-loop” systems.

This dissertation investigates the opportunities and challenges to design the computational framework of a next-generation "human-in-the-loop" systems, especially in the context of online job platforms. Our take is to study the nuances of such a "human-in-the-loop" online platform from the perspective of workers, task designers, as well as from systems. Considering workers’ perspective attitudes, preferences, opinions of the human workers are to be understood, as well as the next generation online platforms need to train workers on the job for human capital advancement. Considering task designers, we study how such platforms can recommend application designers the best deployment strategies, enabling rapid development of human-in-the-loop" applications on online platforms. Finally, from the perspective of systems, we study different alternatives of combining human intelligence with machine algorithms to optimize throughput.

Author(s):

Dong Wei

Funding Acknowledgements:

National Science Foundation