National Science Foundation Honors UVA Computer Scientist for his Human-Centric Vision for Big Data
Imagine you’re a carpenter, and you want to shop online for new tools. Enter “files” in Google, Bing, Yahoo!, or any other search engine, and you’ll get a mix of returns from the X-Files to the latest software. Chances are, anywhere you go on the Web for the next few days, you’ll be pursued by an annoying ad for the TV show. The problem is that the big data systems behind the search engines and ad placement services don’t know you, said Hongning Wang, an assistant professor of computer science in the University of Virginia School of Engineering and Applied Science.
Wang’s goal is to improve the usefulness of these systems while protecting users: “I want to find a way to connect people and big data so that it is truly useful for individuals,” he says.
Wang has proposed creating what he calls a human-centric system for knowledge discovery and decision optimization. This is not simply a set of algorithms that predict a user’s behavior based on actions of similar users, but an ecosystem in which individual human beings and automated knowledge services interact. Wang calls this approach “humans in the loop.”
The National Science Foundation thought enough of Wang’s approach that they presented him with a prestigious Faculty Early Career Development (CAREER) Award. The five-year award is designed to give promising young faculty members like Wang an opportunity to “build a firm foundation for a lifetime of integrated contributions to research and education.”
A Framework for Human-Centric Knowledge Discovery and Decision Optimization
Wang does not hope to flesh out this entire human-centric system in just five years, but he does intend, in the spirit of the award, to build a framework that he and others can expand. Wang will focus his energy on four interrelated projects. The first two projects are about building a system that is more attuned to human needs. “The underlying assumption that guides my approach is that human behavior is rational, that it serves to achieve our own objectives, though they may be latent,” he says.
Wang will begin by building models that combine analyses of the text users generate online—for example, their tweets and posts on Facebook—and the kind of online behavior epitomized by their searches. His goal is to establish a better understanding of their preferences and interests.
The second, related project is finding a way to discern the common thread in what might appear to be a random series of online acts. His goal is to identify the underlying tasks that unify them. This knowledge provides a richer sense of how users go about pursuing information or shopping for a product, an essential prerequisite to creating a truly responsive system that supports their decision-making. “The system is constantly learning from the user rather than applying some predefined algorithm,” Wang says.
But in Wang’s vision, humans are not passive recipients of support and information from a system that could be considered intrusive. The third project is to develop mechanisms that deliver explanations of how the system works and how it reaches specific conclusions about user intent. This would allow the user to play a role in regulating the system, potentially declaring certain categories of data off limits.
Finally, Wang will devise a process to prototype and test his system. He has close relationships with the research teams at Google, Bing, and Yahoo! They have expressed a willingness to collaborate with him as long as the privacy of their users is protected. “The likely scenario will be that I send them my algorithms, which they will run on their system, and they will send me the results,” he said.
Wang believes that the framework he will develop over the course of the next five years could have wide application. “My hope is that it can be applied wherever machines and humans interact to help people discover the knowledge they need and make better decisions,” he says.