Sign In

Communications of the ACM

ACM TechNews

Researchers Advance Artificial Intelligence For Player Goal Prediction in Gaming

An image from the Crystal Island game.

New artificial intelligence software developed at North Carolina State University improves the ability to predict player actions in video games.

Credit: IntelliMedia Group/NC State University

North Carolina State University (NCSU) researchers have developed artificial intelligence (AI) software that greatly improves the technology's ability to predict player actions in video games, which they say is a key to enhancing player experiences in both recreational and educational games.

NCSU Ph.D. student Wookhee Min says current AI software is able to correctly guess players' intentions 48.4 percent of the time, slightly worse than random chance. However, the new software has an accuracy rate of 62.3 percent. The researchers accomplished this by incorporating deep-learning techniques similar to those used in computer-vision and natural-language processing; these techniques enable the software to make sense of massive data sets of player activity logs tracking a player's actions over time. As a result, as a player plays a game more the software's ability to predict their goals will improve, to a point.

The researchers developed the software for use in educational games and tested it in an educational game called Crystal Island, using it to track and predict the actions of 137 players.

"This is a key step in developing player-adaptive games that can respond to player actions to improve the gaming experience, either for entertainment or--in our case--for education," says NCSU professor James Lester.

From NCSU News
View Full Article


Abstracts Copyright © 2014 Information Inc., Bethesda, Maryland, USA


No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account