ClueCart: Supporting Game Story Interpretation and Narrative Inference from Fragmented Clues

Research Background and Problem

Issues and Challenges

• The authors identified indexical storytelling in games—a method of conveying narratives through fragmented clues—as a challenging approach. This narrative style requires players to reconstruct the story from puzzle-like clues, but the nonlinear structure and diversity of clues make classification and interpretation complex.
• Challenges include low efficiency in clue classification, high cognitive load in understanding clues and their relationships, the potential loss of hidden clues, and difficulties in integrating information from both within and outside the game.

Significance

• Game narratives are not only a medium of entertainment but also serve as tools for cultural dissemination. Player-generated content and discussions on video platforms and forums have become vibrant forms of community interaction.
• Understanding complex narratives in games can deepen player engagement and immersion. Research and tool development in this field hold potential for advancing cultural exchange and education.

Research Motivation and Related Work

• Current research on game narratives primarily focuses on categorizing narrative elements from the designer's perspective, often neglecting how players (or content creators) interpret clues to form personalized narratives.
• While tools supporting creative practices exist, they are mainly designed for creating new stories rather than interpreting existing complex clues, lacking support for the needs of player-generated content creation.

Solution

Methodology and Innovations

• The authors proposed a two-tier classification system to meet the needs of content creators for organizing and categorizing clues.
• They developed an open-source creativity support tool called ClueCart, which integrates automated clue capture, classification, and presentation, along with an interactive interface to assist players in analyzing narrative clues and reconstructing stories.
• Innovations include an embedded game mod for automatic clue capture and classification, as well as the use of large language models (LLMs) to generate clue summaries and extract keywords.

Implementation Steps

1. Design of Classification System: Based on a literature review and player participatory design workshops, a two-tier classification system was introduced, categorizing clues (e.g., characters, locations, achievements) and elements (e.g., cutscenes, related characters, environments, artifacts, text, dialogue).
2. Development of ClueCart: The tool captures clue names, locations, and achievements in real-time through a game mod and further classifies them using LLMs, generating summaries and keywords for each clue.
3. Testing and Optimization: User studies were conducted to compare ClueCart with a baseline tool (Miro) in terms of user experience and narrative efficiency.

Research Outcomes

Key Findings

• Efficiency Improvement: Groups using ClueCart demonstrated greater flexibility and efficiency in completing narrative tasks.
• Quality Enhancement: Groups using ClueCart significantly outperformed baseline tools in dimensions such as story clarity, organization, and logical structure, highlighting the quality of narrative analysis results.
• New Perspectives: ClueCart's external knowledge integration feature helped users link game clues to real-world meanings, generating new and profound insights.

Advantages Comparison

• Compared to traditional tools, ClueCart offers automated clue classification and real-time content processing. This reduces the manual workload for content creators and enhances analytical efficiency.
• By visualizing relationships between game clues, ClueCart improves creators' understanding and analysis of complex narrative structures.

Experiment or Evaluation Results

• System Usability: In user satisfaction surveys, ClueCart's system usability scores were significantly higher than those of the baseline tool.
• Creativity Support: Users reported feeling more creative and imaginative when using ClueCart, with its features aiding exploration of nonlinear narratives and multidimensional story structures.
• User Feedback: While overall evaluations were positive, some participants suggested adding customization options to the classification system and improving the usability of clue retrieval.

Limitations and Future Directions

• Limited Classification Flexibility: Some participants found the two-tier classification system insufficiently adaptable to diverse clues, highlighting the need for enhanced user customization in the future.
• Platform Constraints: The game mod design is currently limited to open platforms. Future work should explore solutions for closed platforms (e.g., Nintendo Switch, Xbox, and PlayStation).
• Automation Improvements: Future versions are recommended to integrate more LLM-based automated relationship detection features to enhance clue understanding and work efficiency.

Conclusion

• ClueCart provides an effective tool for supporting player creative communities, improving the efficiency and quality of game narrative analysis, and advancing clue-based storytelling methods.
• The project demonstrates the profound impact of game narratives on community content creation and proposes directions for improving creator tools, offering greater possibilities for future cross-media narrative applications.