I had the pleasure of meeting Ujwal Gadiraju in Heidelberg last week at this year’s Forum. He warned me he could go on forever about his research, and he wasn’t entirely wrong; the only thing stopping him was that I had to leave for another interview before we got finish our conversation! Ujwal was kind enough to follow up with me via email. Here, I share what he had to say to a few of my questions.
Can you tell me more about your research on crowdsourcing?
My main research interests and experience broadly lie in the fields of Human Computation and Crowdsouring, Information Retreival, Data Mining, Web Science, and Social Computing.
My experience with research in the realm of Computer Science began in 2012 in the form of my Master’s Thesis work on “Detection of Duplicate Content on Twitter” at Delft University of Technology, in the Netherlands. Within the scope of this work, we developed a scale that can be used to measure the degree of duplicity within a pair of duplicate tweets. We analyzed the top-k search results on Twitter and observed the extent to which duplicates appear in top-k retrieval.
Over the last few years during my work as a PhD Candidate at the L3S Research Center, Leibniz University of Hannover, I have developed a keen appetite for research in the growing field of human computation and crowdsourcing. The idea of building hybrid human-machine systems to solve problems that go beyond the capability of machines was immediately appealing to me. My work in this regard has mainly covered two pivotal aspects that influence the effectiveness of the paid crowdsourcing paradigm: (i) task design, and (ii) crowd workers’ behavior. Leveraging the dynamics of tasks that are crowdsourced on the one hand, and accounting for the behavior of workers on the other hand, can help in designing tasks efficiently. I am currently working on designing quality control mechanisms that outperform existing state-of-the-art methods across the diverse landscape of microtasks.
With the existing advances in artificial intelligence today, machines are posing a realistic threat of rendering many a human unemployed. I believe that although machines can mimic human actions and behavior to an alarmingly unprecedented extent in the near future, humans shall always hold an edge over machines in certain areas that require intelligence. At this juncture, crowdsourcing can serve as an adequate means of acquiring human input at a large-scale. In many cases as recent works have indicated, braiding humans and machines together in hybrid systems can help to eliminate computational limitations of human-centered systems and enhance the capability of machines. I am highly motivated to contribute towards the promising potential of this line of research.
Do you think that game-based crowdsourcing could work better than other kinds (such as paid approaches), since games are engaging?
I believe gamification plays a pivotal role in crowd worker engagement and retention. Research has shown time and again that adding layers of game elements to workflows improves the response rate and throughput in crowd-powered systems. Often, monetary compensation is a sufficiently strong motivator to attract crowd workers. However, tasks can be made more enjoyable if they can be modeled as ‘games with a purpose.’ The challenge then is to mold monotonous, repetitive and otherwise boring tasks into those with interesting workflows that eventually yield desired results. This of course, is not always easy to do. Quite often the costs required to gamify a crowdsourcing task are disproportionately larger than the benefits. In settings where both game-based or monetary incentives are equally viable, game-based crowdsourcing tend to have the edge.
Forgetting whether it’s realistic at this moment, what problem would you most love to crowdsource?
I would love to crowdsource multilingual personal assistants for older adults around the world. Older adults can benefit from having a crowd-powered personal assistant on demand to help them tackle everyday obstacles that manifest due to challenges ranging from their technological to physical abilities. These crowd-powered personal assistants can be accessed or seemingly summoned using online web application triggers, or simply calling them on telephones. For example, a grandmother who speaks only Telugu wants to Skype with her granddaughter but doesn’t know how do go about it, can summon a virtual personal assistant who can help her accomplish it. The novel contribution of this vision is the potential of providing a diverse range of micro-services via crowd-powered personal assistants on demand.
What’s next for you?
After the doctoral studies, I will have to choose between research in academia or industry. At the moment I am not entirely sure which way to sway, but I’m excited to see where I can make a difference.