Herself’s Artificial Intelligence

Two Hungarian researchers have come up with a cunning way to create the most frustrating computer game imaginable.  Laszlo Laufer and Bottyan Nemeth at the Budapest Univesrity of Technology and Economics have discovered that a gamer’s button presses can be predicted 2 seconds before they make them, through measurements of skin conductance.. . .

Laufer and Bottyan had volunteers play a simple computer game called YetiSports JungleSwing, which involves controlling a swinging yeti as it leaps up through the branches of a tree, while measured players’ heart rate, skin conductance the electrical activity in their brain. It’s a pretty fun game, but its surprisingly tricky to time the leap from branch to branch correctly. You can try it for yourself over here.

The researchers then used neural networks to analyse the biofeedback signals and input records, to see if they could predict the moment that a player would click the jump button in the game. To their surprise, they found that skin conductance alone is enough to predict a jump up to 2 seconds beforehand.

. . .

Skin signals betray a gamer’s moves

More information:
Yeti Sports

Social scientists have really found a gold mine in using artificial communities to simulate real world events. This time we have ‘Corrupted Blood’ rampaging through ‘World of Warcraft’.

When an estimated 4 million people encountered a deadly epidemic called Corrupted Blood that left the landscape strewn with corpses, scientists were intrigued, not horrified. Similarly, when Whypox, a measles-like disease, was unleashed into a community of over 1.2 million young adults, researchers sat back and took notes.

That’s because these infections were running rampant through the virtual worlds of massive online communities such as the World of Warcraft and Whyville.net. Serious academic researchers, from epidemiologists to economists, are beginning to think online games and virtual worlds can be new laboratories to observe behavior and test theories they can’t experiment with in the real world.

“There’s 9 million people playing World of Warcraft every day — think of the insights you could gain,” said Dr. Ran D. Balicer, an epidemiologist at Ben-Gurion University of the Negev in Israel. He normally studies pandemic preparedness but recently wrote a paper in the scientific journal Epidemiology on the virtual virus Corrupted Blood in World of Warcraft, a fantasy role-playing game where players do battle with swords, maces, and other weapons. “This is a new evolutionary step in infectious disease modeling.”

. . . Online gamers become guinea pigs

More information:
Modeling Epidemic Spread in Synthetic Populations – Virtual Plagues in Massive Multiplayer Online Games ( pdf )

More information
An Agent Based Modeling for Pandemic Influenza in Egypt

Could a computer predict your next move in a game of strategy based only on observations of your past behaviour? The US Defense Advanced Research Projects Agency (DARPA) certainly hopes so. It has funded a project led by computer scientist H. Van Dyke Parunak to explore the idea.

Artificial intelligence experts normally approach this problem by assuming that the inner state of a person (or an “agent” in AI parlance) can be described in terms of beliefs, desires and goals. Since these goals determine an agent’s actions, it should then to be possible to use this knowledge to make predictions about the agent’s future actions.

Parunak claims to have used these ideas with some success in making predictions about future behaviour. He says his simulation works in relatively complex environments, making predictions in real-time.

Stategy predicting software

Prediction of human behavior through simulations has been a busy topic of late. It was used, mostly unsuccessfully to predict how fighter pilots would react to stressful situations. It has been used in the corporate world to predict consumer behavior and more recently in creating game characters that are believable. Success so far has been limited.

More information:
Geosimulation: Modeling Crowd Behavior
US Patent Application
Action-Reaction Leaning: Analysis and Synthesis of Human Behavior ( simpler interactions are discussed in this thesis )
They Got Gameplay – Interactive Pretending: An Overview of Simulation (pdf)
Egress Reseach used after the Rhode Island fire to predict how and why crowds move as they do under duress.