The Evidence Engine: A Serious Game for AI Research Literacy

The Evidence Engine is a browser-based serious game for AI research literacy, built by two individuals as part of an OpenAI 2026 hackathon submission. The project claims to use inquiry-based and experiential learning to teach players how to evaluate AI research claims through interactive gameplay. It is described as a tool for educators and researchers to help students understand concepts like bias, data leakage, and model evaluation in real-world scenarios.

The description states that the game uses GPT-5.6 and Codex for development and is built with React, TypeScript, and other web technologies. It includes five complete cases across domains such as healthcare and public safety, and aims to move beyond traditional quizzes by allowing players to make decisions and observe consequences.

Key open questions include whether the project has been tested in real classrooms, how it differentiates from existing educational tools, and what its long-term viability or scalability looks like. The author's own account is self-reported and unverified; no revenue, customer data, or traction evidence is provided.

The description states that The Evidence Engine is a browser-based serious game for AI research literacy. It is described as an interactive tool where players take on the role of research investigators in five different cases involving domains such as healthcare, public safety, environmental monitoring, scientific research, and deepfakes.

Players are tasked with inspecting datasets, reports, citations, audits, and witness statements; questioning a fallible AI assistant; identifying bias, leakage, confounding, and unreliable evidence; reproducing model evaluations in a research lab; building an evidence board; and defending a final recommendation before a council. The game is said to evaluate the quality of players' evidence and reasoning rather than rewarding correct answers.

The product is described as using inquiry-based and experiential learning methods embedded directly into gameplay, with scaffolding provided through tutorials, visual cues, contextual explanations, and guided challenges.

The description states that The Evidence Engine was inspired by the author's experience as a lecturer and researcher at Monash University. It aims to address a gap in how students learn AI concepts—specifically, that they often recognize terminology without practicing how to evaluate actual research claims.

The project positions itself as an alternative to conventional multiple-choice quizzes, focusing instead on experiential learning and inquiry-based approaches. The author notes that the goal was to move beyond simple recognition of concepts like bias or accuracy to applying them in messy evidence environments.

The claim evolution appears to be from a pedagogical problem (students don't apply concepts) to a solution (a game that lets players make decisions and see consequences). There is no indication of prior versions or iterative development beyond this single submission.

The description states that The Evidence Engine is intended for use by students and researchers, particularly in educational settings. It is described as a tool for lecturers and educators to help teach AI research literacy.

It targets those who need to understand how to evaluate AI research claims, especially in domains such as healthcare, public safety, environmental monitoring, scientific research, and deepfakes. The game is designed to be played by individuals investigating AI research problems rather than general audiences or consumers.

The description does not specify whether it targets specific grade levels, academic institutions, or professional training programs beyond being a tool for educators and researchers.

Not evidenced.

The description does not contain any information about pricing models, monetization strategies, or business models. There is no mention of subscriptions, licensing fees, or commercial use cases beyond educational applications.

The description states that the game was built with React and TypeScript using a data-driven case system that allows new investigations to be added without rebuilding the entire game. It uses CSS for cartoon-style interface and animations, and the Web Audio API for interaction sounds directly in the browser.

It is said to use GPT-5.6 as a creative and reasoning collaborator for developing case narratives, research problems, evidence relationships, and explanations. Codex was used for rapid prototyping and development, supporting interface design, gameplay systems, tutorials, animations, audio cues, responsive design, testing, documentation, and GitHub workflow.

The application currently includes five complete cases and a reusable engine for future scenarios. It is described as browser-based and responsive across desktop and mobile screens.

Not evidenced.

There is no evidence of revenue, customers, usage metrics, or adoption beyond the single project submission to a hackathon. The description does not mention any classroom testing, user feedback, or deployment in educational environments.

The author notes that the next step is classroom playtesting with students and researchers, suggesting that real-world traction has not yet occurred.

Not evidenced.

The description does not provide information about competitors, existing tools in the same space, or how The Evidence Engine compares to other educational platforms or serious games for AI literacy. No market analysis or competitive positioning is included.

• Lack of traction: The project has only been submitted to a hackathon and lacks evidence of real-world use or classroom adoption.
• Unverified claims: All descriptions are self-reported and unverified; there is no independent validation of the educational effectiveness or gameplay quality.
• Limited scope: While five cases are included, there is no indication of how easily new cases can be added or whether this will scale effectively.
• Dependency on AI tools: Heavy reliance on GPT-5.6 and Codex raises questions about reproducibility and long-term viability if these tools change or become unavailable.
• Unclear commercialization path: No evidence of a business model, pricing strategy, or monetization approach beyond educational use.

1. Has the game been tested in real classrooms or with actual students?
2. What specific pedagogical outcomes have been observed during playtesting?
3. How does the team plan to scale beyond the current five cases?
4. Are there any plans for monetization or commercial deployment?
5. What are the limitations of using GPT-5.6 and Codex in educational content creation?
6. How is the quality of AI-generated content ensured and reviewed?
7. What distinguishes this approach from other serious games or educational platforms currently available?