Skip to main content

École normale supérieure

Organization2 reports

2 reports, presentations and talks by École normale supérieure in the Game Industry Library, the newest “DEEP: A Model of Gaming Preferences Informed by the Hierarchical Nature of Goal-Oriented Cognition” (2025).

Documents

Cover of DEEP: A Model of Gaming Preferences Informed by the Hierarchical Nature of Goal-Oriented Cognition
Whitepaper67 pages

DEEP: A Model of Gaming Preferences Informed by the Hierarchical Nature of Goal-Oriented Cognition

The DEEP model establishes a comprehensive framework for understanding gaming preferences by mapping player agency onto the hierarchical nature of goal-oriented cognition.

  • The DEEP model categorizes player engagement into four hierarchical dimensions—Discovering, Experimenting, Expanding, and Performing—which map directly to the cognitive architecture of human goal-setting.
  • Empirical analysis of 16,598 video games confirms that 90% of titles utilize this nested hierarchical structure, aligning game design with fundamental cognitive processes.
  • Player behavior is predictable based on these dimensions: those who prioritize 'Discovering' and 'Experimenting' engage with a wider variety of titles, while those who favor 'Performing' focus on mastering fewer, specific games.
Game DesignGame StudiesPlayer Behavior+1
Institut Jean NicodFeb 2025
Cover of Video Games as a Corpus: Sentiment Analysis using Fallout New Vegas Dialog
Whitepaper6 pages

Video Games as a Corpus: Sentiment Analysis using Fallout New Vegas Dialog

This research explores the utility of video games as a rich, professionally annotated corpus for natural language processing (NLP) tasks, specifically focusing on multilabel sentiment analysis.

  • Video games serve as a high-quality, professionally annotated corpus for NLP, as demonstrated by a new dataset extracted from Fallout New Vegas containing over 53,000 lines of dialogue.
  • The dataset supports cross-lingual sentiment analysis across five languages—English, Spanish, German, French, and Italian—with eight distinct sentiment labels including anger, disgust, fear, happy, neutral, pained, sad, and surprised.
  • Multilingual BERT models achieved a peak accuracy of 54% when predicting sentiment in Spanish, demonstrating the ability to transfer sentiment understanding to languages not included in the training set.
AIGame StudiesRPG+2
University of HelsinkiSep 2022

Do you work at École normale supérieure? Get a free report of how these documents are read and cited here.