4/8/2016 Emily Scott
With another CAREER Award, ISE's Great Eight have become the Notorious Nine!
Written by Emily Scott
Assistant Professor Sewoong Oh is the latest ISE faculty member to win a CAREER Award from the National Science Foundation. His award will benefit his research in social computing and his efforts to improve social computing systems.
“What I want to do broadly is bring mathematical tools and algorithms to make such systems more reliable and more efficient,” Oh said.
Social computing has two sides: aggregating information from individuals and using these contributions to benefit society, such as in the case of Wikipedia, or using people to compute things, such as in the case of crowdsourcing to complete simple tasks on a large scale.
Oh’s proposal for the CAREER Award involves research specifically related to crowdsourcing,
a labor market increasing in popularity that involves using many people to complete large tasks.
“The idea is that, there are a lot of people who have lots of free time, and who would like to contribute to a lot of things, but don’t know how to,” Oh explained.
Companies or individuals may need to complete different simple tasks on a large scale, so they can divide it into small parts and distribute the tasks among many people.
“In crowdsourcing, you have many tasks that you assign to many people,” Oh said. “The question we ask is, how can we assign these tasks to the crowd so that when we aggregate the information, we get the most information out of it?”
He said he is working on developing algorithms, and borrowing techniques from information theory and coding theory, to design an efficient task assignment system.
Another aspect of Oh’s research is recommendation systems, which traditionally use ratings from people to make predictions. These systems can be used to rank results from search engines or give recommendations to users on websites like Netflix.
“By just looking at what you’ve done or what websites you’ve browsed or what movies you like, I can’t tell much about what you’ll do next . . . but by aggregating all this information over the whole crowd in the society, we can try to come up with a way of using the information from other people to tell something about you,” Oh said.
But instead of using a rating system — such as rating a movie on a five-star scale — Oh suggests that a comparison system may work better.
“Ratings are not reliable in the sense that my rating scale is different from your rating scale,” Oh said. “My rating also might be subjective (because of) the time of when I’m rating, and also how I feel at the time.”
Instead of using ratings, Oh said comparing two items would be more effective because this process is scale-free. If a person says they like one movie better than another, it’s exactly the same as someone else saying they like one movie better than another.
“And the question is then, if people give you comparisons, then how do you aggregate it to come up with a recommendation system based on comparisons?” Oh said.
This kind of recommendation system could have many applications — it could help websites decide what advertisements are best to display, it could assist retail websites like Amazon make recommendations to customers, and much more.
Oh said many traditional systems within social computing work well, but one of the important challenges is making algorithms that are based on data that is constantly changing.
“A lot of algorithms that are developed are based on static data,” he said. “But in real industry, one of the most important challenges is doing it dynamically.”
Another challenge is developing these systems in a way that is efficient, without sacrificing performance or accuracy.
Oh said the CAREER Award will help him address these challenges while giving him confirmation of the importance of his research.
“It’s a very important step for a young researcher like me to be acknowledged . . . that it’s not just me who thinks it’s important, but it’s other peers and the research community also thinks that it’s something that should be done,” he said.
READ THE FULL AWARD HERE: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1553452&HistoricalAward...