Munoz-Avila captures NSF CAREER award for “Digital Thinking”

In the late 1990s, IBM unveiled a computer system dubbed Deep Blue that became the first machine to win a chess tournament against a reigning world champion, Garry Kasparov.

It was a notable achievement for computer science. Deep Blue’s winning game strategy and ongoing modifications since that time are a testament to the computer’s ability to sift quickly through huge amounts of information and calculate a successful course of play, using a complex set of algorithms pre-ordained by the machine’s development team.


Hector Munoz-Avila

But what if a computer could be taught to make decisions based upon “experience,” asks Hector Munoz-Avila, assistant professor of computer science and engineering.

What if a computer could be taught to take a set of inputs and evaluate options based on past outcomes of the same, or similar, circumstances?

Recently, the National Science Foundation (NSF) granted a five-year CAREER award to Munoz-Avila in recognition of the potential of his research to help computers “learn” in this manner.

“Many cognitive scientists believe that humans begin acquiring knowledge by learning simple skills, which then are combined to learn more complex ones,” says Munoz-Avila. “For example, we learn to play chess by first learning how to move individual pieces. Then we learn simple strategies, such as opening moves and basic positioning. We continue to build incrementally on our expertise until we master complex game-playing strategies based on the integration of basic skills.”

The NSF’s Faculty Early Career Development (CAREER) program supports teacher-scholars who effectively integrate research and education across a broad set of disciplines. Munoz-Avila’s is the third NSF CAREER Award granted recently to an assistant professor in his department, following Brian Davison’s award in 2006, and Jeff Heflin’s award in 2004.

In his research, Munoz-Avila is pursuing a “unified architecture for automated learning of skill hierarchies” from a collection of examples – with one example being a set of previously played chess games and their outcomes.

“The architecture incrementally learns and uses skills from examples,” he says. “Initially, as a relatively small number of examples are provided to the computer, it will learn simple skills by closely mimicking these examples. As more examples are given, it will be capable of learning complex skills that build upon the simpler skills it learned previously. The goal would be, after many, many examples, for the computer to be able to generalize these inputs into abstract, complex concepts and then apply them to new situations.”

Munoz-Avila believes that this approach – a departure from applying pure computational horsepower and predetermined algorithmic computations to a game or other mathematically-based problem – is potentially much more powerful than the search/retrieve/compute power of systems such as Deep Blue.

“Deep Blue and other search-intensive approaches develop strategies in a different manner from humans,” he explains. “Our goal is to build algorithms that resemble the way humans learn and solve problems. This line of research will be capable of developing effective strategies and will also explain the reasoning mechanisms behind them. The latter is of crucial importance in areas such as teaching and decision support, where providing justification for the solutions is as important as providing the solutions themselves.”

Part of Munoz-Avila’s NSF award will support a project with Glenn Blank, professor of computer science and engineering, which provides local middle- and high-school students with hands-on experience in the basic concepts of artificial intelligence as part of Lehigh’s broader efforts to promote science and technology education in the Lehigh Valley.

“Gaming strategy in general is a good example to use to explain this research and other initiatives in the same field,” says Munoz-Avila, “and there are certainly future applications for this type of technology in the area of computer gaming. However, the broader impact will become evident when the technology is developed enough to support military applications, security and other civil concerns, and even project management in terms of industrial production and service delivery.”