Paper Reading #2: Protecting Artificial Team-Mates: More Seems Like Less

Reference Information

Tim Merritt, and Kevin McGee. "Protecting Artificial Team-mates: More Seems like Less." Protecting Artificial Team-mates. National University of Singapore, 05 May 2012. Web. 03 Sept. 2012.

Authors

Tim Merritt - Pursuing his PhD at the NUS Graduate School for Integrative Sciences and Engineering. 

Kevin McGee - Researching the development of partner technologies that increase human participation in life's interesting and important challenges.

Article Summary

Photo Credit: StockPhotos - (http://www.ehow.com/article-new/getty/xc/86531465.jpg)

This article looked into the relationship between people's behavior toward human and AI in cooperative gaming. Both human and AI team-mates' beliefs about their behavior differed depending on who they thought they were playing with. In the past, it has been suggested that humans and AI agents in games both had different impacts on humans when being played cooperatively with.

After beginning with this initial idea, they recruited 32 participants and examined logged data of game performance when playing with an AI subject along with a human. Additionally, the emotions and playing style differed depending on who they were playing with as well. Taking all this into consideration, the team of two were trying to see if the participants could be tricked into thinking they were playing with an artificial teammate.

In the end, it was interesting to see that even when playing with an artificial teammate, as long as they think they were playing with a real human, the results would mimic that of a real player.

Related Work

When looking at work related to this specific idea, there were a few other pieces of work that were similar in nature. They focused on behavior and the human response in a cooperative interaction. Most of these studies also tried to deceive humans a bit to get the data they needed. Here are a few of the related work I found:

• Blascovich, J. A theoretical model of social influence for increasing the utility of collaborative virtual environments. 
• Nass, C., Fogg, B.J. Can computers be teammates?
• Oviatt, S. Designing and evaluating conversational interfaces with animated characters.
• Abraham, A.T. AI for dynamic team-mate adaptation in games.
• Kiesler, S. A prisoner's dilemma experiment on cooperation with people and computers
• Merritt, T. Did you notice? Artificial team-mates take risks for players intelligent v. agents
• Merritt, T. Choosing human team-mates: higher enjoyment and preference for humans
• Miwa, K. Analysis of human-human and human-computer agent interactions...
• Weibel, D. Playing online games against computers vs humans. Effects on presence...
• Williams, T. Agression, competition and computer games: computer & human opponents.

Most of the above work didn't expressively look into whether humans preferred AI or human teammates, but instead focused on thee effect it had on their playing styles.

Evaluation

The two authors evaluated the work with a slight bias. They had to deceive the participants a bit to make the players they were instead playing with someone of a different nature than themselves. They used quantitative words that played with emotions and described how the participants felt and behaved instead of numbers. They looked at the study group as a whole and went further in depth depending on the type of opponent they thought they were playing with. Both ways were combined and appropriately measured.

Discussion

This paper is important because it shows that perception has a lot to do with how people think they act. When being deceived, some users had the same reactions as they had when they thought they were playing with a real player. I think the analysis done on this subject is interesting and their evaluation was appropriate for this nature of work.

If human AI can eventually be at the level humans are at, then I believe that perception will play an important part in how humans respond to the game. I learned from this that although humans played better with AI, they had more fun with humans because of the peer aspect towards cooperation.