U study aims to recommend books

The University has paired with a German university to up diversity in recommendation lists.

Mehgan Lee

Thousands of Web sites such as Amazon.com offer products tailored to individual users, based on their prior purchases.

But the lists these sites offer are too generic and narrow for Cai-Nicolas Ziegler, a graduate student at the University of Freiburg in Germany.

“The whole list itself is not very exciting to me,” he said. “I’d like a list that reflects my whole plethora of interests.”

So Ziegler traveled across the Atlantic Ocean to investigate whether other people feel the same way.

He came to the University to conduct a study with the GroupLens Research Project, a group of University faculty and graduate students in the computer science and engineering department.

“They were one of the first groups to conceive of recommender systems, and have a high reputation worldwide,” Ziegler said.

The group hopes to show that people want more diversity in the lists offered by recommendation systems.

“Our goal is to show that diversification makes sense,” Ziegler said.

Bookcrossing.com users were invited to participate in the study, according to the Web site.

Members read books, tag them with a tracking number and then leave them in a public facility. Whoever finds the book is encouraged to read it, then access the Web site and write a journal entry on what he or she thought of it.

Ziegler’s study developed complex software that catalogs those journal entries, he said. Then they developed a new recommendation system that bases its suggestions on people who have similar interests, he said.

Study participants were asked to rate lists given to them based on their journal entries, Ziegler said. Some of the lists had a wide variety of diversity and some did not, he said.

So far, more than 2,000 Bookcrossing.com users have participated in the study.

The results to date show readers prefer lists that encompass a broad variety of their interests rather than narrowly tailored lists, said Joseph Konstan, a member of GroupLens involved with the study, and professor of computer science and engineering.

The recommendation system might be a feature Bookcrossing.com would be interested in adding to its site some day, said Ron Hornbaker, the Web site’s founder and editor.

Morgon Mae Schultz, a journalism senior, has encountered recommendation systems on Amazon.com while buying textbooks and compact discs, she said. But she has never bought any of the products they have tried to sell her in hundreds of e-mails, she said.

“I’m more of a proactive consumer,” Schultz said.

But she said she is interested in peer recommendations, such as those developed by the study.

“I trust my peers more than I trust someone that’s trying to sell me something,” she said.