Pavel Kucherbaev

I am a Postdoctoral Researcher within Web Information Systems group at TU Delft. I got my PhD at University of Trento in Italy doing research of Quality Control strategies in Microtask Crowdsourcing platforms.

Research

Microtask crowdsourcing is a powerful approach allowing to solve challenges, which machines are not yet capable of doing well. Still the actors involved in the approach, requesters and workers, face serious problems. Requesters have hard times trying to get results with expert level of quality, and workers struggle to have a comfortable environment to perform well tasks they like. I aim to address these problems focusing on quality control and human computation interfaces aspects of microtask crowdsourcing.

Quality Control

I believe that requesters should not be domain experts in microtask crowdsourcing to run tasks on platforms effectively, so as people browsing websites should not be experts in TCP/IP and UDP protocols. I believe that requesters should have black box solutions assuring close to 100% level of quality. To bring this vision to life I develop various automatic techniques to predict low quality assignments during runtime, as well as various intervention strategies to address the assignments predicted as low quality.

Human Computation Interfaces

I believe that the way microtask crowdsourcing works nowadays is very old-fashioned, making it complicated to publish tasks, to perform them and to integrate results with third party services. To address this challenge I investigate new type of worker-platform interaction in a form of a dialogue with a chatbot, instead of filling up webforms. I also research how various automatic chatbots could be augmented with microtask crowdsourcing, to make them more human-like, and research how automatic chatbots could assist crowdsourcing workers to perform tasks well and with more fun. 

More

More information is available at kucherbaev.com. Feel free to contact me at pavel@kucherbaev.com.

Pavel Kucherbaev