Presenter Information

Lucille NjooFollow

Event Website

https://docs.google.com/presentation/d/19X53U3WeAY6f4u_TSBFB1kJHTw1B9fpWB5BljXR2whs/edit?usp=sharing

Start Date

12-12-2018 12:05 PM

Description

Recent research into artificial intelligence (AI) in games has found that AIs have learned to use unconventional strategies that humans would not expect to be effective. Recent poker AIs, for example, deliberately use to their advantage certain poker plays that humans would deem unwise or useless [1]; AIs are becoming “smarter” at game strategy than humans themselves through these unconventional strategies [2] [3]. At the same time, researchers also aim to create AIs that are more humanlike, exhibiting emotion and personality [4]. Thus, these recent AI developments raise the question of how unconventional strategies might affect user experience of a computer game. I intend to explore this idea by investigating how unconventional strategies like randomization might affect user experience of an AI’s personality and opponent difficulty for the card game “Big 2.” The results of this study may have implications for our understanding of good strategy and could improve a wide variety of computer game AIs by offering alternative decision-making algorithms that better user experience. I propose to study this by programming a 2-player “Big 2” computer game with four different AI versions, each with different decision-making algorithms, employing play testers to play each version through Mechanical Turk, and gathering user experience feedback.

Share

COinS
 
Dec 12th, 12:05 PM

User Experience of Unconventional AI Algorithms in “Big 2”

Recent research into artificial intelligence (AI) in games has found that AIs have learned to use unconventional strategies that humans would not expect to be effective. Recent poker AIs, for example, deliberately use to their advantage certain poker plays that humans would deem unwise or useless [1]; AIs are becoming “smarter” at game strategy than humans themselves through these unconventional strategies [2] [3]. At the same time, researchers also aim to create AIs that are more humanlike, exhibiting emotion and personality [4]. Thus, these recent AI developments raise the question of how unconventional strategies might affect user experience of a computer game. I intend to explore this idea by investigating how unconventional strategies like randomization might affect user experience of an AI’s personality and opponent difficulty for the card game “Big 2.” The results of this study may have implications for our understanding of good strategy and could improve a wide variety of computer game AIs by offering alternative decision-making algorithms that better user experience. I propose to study this by programming a 2-player “Big 2” computer game with four different AI versions, each with different decision-making algorithms, employing play testers to play each version through Mechanical Turk, and gathering user experience feedback.

https://digitalcommons.lmu.edu/honors-research-and-exhibition/2018/section-02/10