Defense: Telltale Hearts: Modeling Player Response to Choice-Based Cinematic Adventure Games

Speaker Name: 
John Murray
Speaker Title: 
PhD Candidate
Speaker Organization: 
Computer Science
Start Time: 
Friday, June 1, 2018 - 8:30am
End Time: 
Friday, June 1, 2018 - 10:30am
Engineering 2, Room 215
Noah Wardrip-Fruin

Abstract:  Dramatic choices represent the core of the content found in contemporary adventure games like those produced by Telltale Games and the Life is Strange series by Dontnod Entertainment. These consist of emotionally-charged and dramatically contextualized player choices rather than strategic or skill-based challenges. Current approaches to understanding and assessing player experience in these games include surveys, interviews, think-aloud and various biometric measurements. These measures, taken alone, are useful for understanding traditional gameplay challenges such as platformers or puzzles but fall short of assessing how emotional content is designed to influence player experience or in charting relationships between content structure and variations in player responses. Instruments such as surveys and interviews suffer from post-experience effects or require interruptions, while traditional content analysis struggles with the dynamic nat! ure of the content.

I collaboratively conducted a study of six players playing the first episode of The Wolf Among Us (TWAU) by Telltale Games, recording gameplay and player videos and recorded heart rate and skin conductivity. The gameplay is representative of the type of emotional storygame content in the genre. To compare and analyze player responses and classify content features, I developed a formal model and used it to annotate player traversals with feature locations and player choices. These align player response data to content for comparison and analysis. Researchers can use the model to compare multiple paths through the game that can span hours of gameplay. I also developed a web-based visual data mining tool, Sherlock, to aid in applying the model for annotating and analyzing player experience datasets. I assess a prototype of the system through a preliminary user study of game researchers focusing on the visualization and content alignment features. Researchers found the approach ! promising, and many indicated they would incorporate the tool into their work. I compare the encoding process by applying two other formal models, story intention graphs and choice poetics, to a segment of gameplay from TWAU. The formal model describes content patterns that lead players to care about story values through controlling player decision context. Applying it to player traversals also highlighted the role that dependent content, or “payoffs,” played in making players feel responsible for their decisions in spite of severely restricted agency. The general approach of using sets of player responses and in particular facial expressions to understand storygames and player experience provides a promising use of the multitude of player experiences published on streaming platforms such as Twitch for game research.