Eternagram: Probing Player Attitudes Towards Climate Change Using a ChatGPT-driven Text-based Adventure

Title of the Paper

Eternagram: Probing Player Attitudes in Alternate Climate Scenarios Through a ChatGPT-Driven Text Adventure

Paper Information

• Subject Area: Examining player attitudes toward climate issues using gamification and a GPT-driven conversational system
• Keywords: Climate attitude assessment, gamified testing, natural language processing, human-computer interaction, personality-behavior correlation, democratic values, data analysis, environmental psychology

Research Background and Problem

• Identified Problems:

  • Traditional methods for assessing climate attitudes rely on simple questionnaires, which fail to fully capture genuine attitudes toward climate issues.
  • Current approaches are susceptible to social desirability bias or other external influences.
  • The complexity of attitudes toward climate issues cannot be comprehensively addressed using a single tool.

• Significance of the Problem:

  • The global impact of climate change necessitates a more accurate understanding of individual attitudes.
  • Encouraging public or group awareness and action on climate change requires more engaging and immersive assessment methods.

• Research Motivation and Related Work:

  • Gamified Behavioral Assessment (GBA) enables participants to experience immersive scenarios while naturally collecting data on climate attitudes.
  • With advancements in artificial intelligence, particularly NLP, chatbot technologies can dynamically adjust interactions, providing a more natural conversational environment for testing.
  • This study aims to establish an innovative assessment method by combining game narratives with a ChatGPT-driven conversational model.

Proposed Solution

• Proposed Method:

  • Develop the text-based adventure game Eternagram, embedding a ChatGPT-driven conversational system within a future climate crisis scenario.
  • Design detailed narrative arcs and trigger mechanisms within the dialogue prompts, allowing players to drive the game progression through free-text input.
  • Use measurement tools (questionnaires and behavioral analysis) to correlate in-game and real-world attitudes.

• Innovations:

  • The first integration of an open-ended conversational system (GPT) with interactive narrative to assess climate attitudes.
  • Creation of a social media-like interface (e.g., Instagram) to enhance immersion and realism.
  • Systematic incorporation of "future scenario forecasting" to quantify and analyze shifts in player attitudes toward climate issues.

• Implementation Steps and Key Techniques:

  • Worldbuilding: Employ the One Hour Worldbuilders method to construct a future climate crisis scenario for the text-based game.
  • Game Mechanics: Focus on open-ended text-based interactions, with key trigger points controlling narrative progression.
  • Character Design: Use GPT-4 to develop the core character Ryno, with manually designed prompts to shape character dialogue traits.
  • User Evaluation and Data Analysis: Collect three sets of survey data (pre-game, in-game, post-game) using mixed methods, and analyze correlations in user attitude changes.

Research Outcomes

• Key Findings:

  • Attitudes expressed in the game were strongly correlated with post-game attitudes but not with pre-game attitudes, indicating a homogenizing effect of the game scenario on player attitudes.
  • Personality traits such as "openness" and "agreeableness" were significantly correlated with climate attitudes. Additionally, participants who supported democratic values were more likely to report positive climate attitudes.
  • In-game attitudes and behaviors were shown to effectively complement traditional attitude assessments (e.g., questionnaires or observed behaviors).

• Advantages Over Existing Solutions:

  • Introduced immersive interaction and real-time dialogue into attitude assessment, making the game a more flexible and realistic evaluation tool.
  • Avoided social desirability bias inherent in traditional questionnaires and achieved contextualized natural interaction.

• Experimental/Evaluation Results:

  • Spearman correlation analysis revealed significant relationships between attitude changes and various personality traits and political orientations.
  • Data indicated that supporters of democratic values and players willing to engage in collective action exhibited more positive attitudes toward climate issues.

• Limitations and Future Directions:

  • Limitations:
    • Participant demographics were primarily 18-34-year-old Asian students, limiting generalizability and excluding non-native language speakers.
    • Variations in game questionnaire design may introduce quantification bias.
    • Occasional technical instability in the game interface.
  • Future Directions:
    • Optimize dynamic scenario generation tools and interactive experiences.
    • Further explore attitude expressions in natural user dialogues.
    • Expand the use of game-based interventions and educational tools.