Whodunit — Case Study

AI-Powered Murder Mystery Chat Game

Role

Co-Developer (Team of 3)

Timeline

Winter Semester 2023/24

Platform

Android (Mobile App Development course)

Stack

Flutter logo
Firebase logo
OpenAI
The App

Whodunit is a murder mystery chat game based on Agatha Christie's And Then There Were None. Players interrogate AI-powered characters through dynamic conversations, gathering clues across chapters until they're ready to accuse the killer. Each character responds in-persona, changes status as murders unfold, and stops responding once they're dead.

My Contributions

This was a closely collaborative project — game flow, design decisions, scripting, debugging, and user testing were shared across the team. My individual contributions focused on the narrative architecture, core game mechanics, and key screens:

Narrative & Game Structure

  • Extracted and structured profiles for 10 characters directly from the book.
  • Structured the final game flow into chapters with objectives, context, and plot points.
  • Wrote the Introduction and Reveal Story narratives.

Core Mechanics

  • Implemented the Guess Button with string-matching logic and overlays.
  • Prompt Engineered the dynamic context that adapts AI instructions as the story progresses.
  • Developed Game Over, Congratulations, and Game State logic.
Whodunit Features — Annotated Home Screen and Character Profiles

Annotated home screen diagram labeling game mechanics and character status trackers.

Character Profile — Detective's Dossier view
Challenges

Backend Switching

Initially used GetStream.IO but switched to Firebase mid-project to better integrate with ChatGPT.

Prompt Engineering

Refining OpenAI instructions to keep characters in-persona consistently across long, unpredictable conversations.

Flutter Learning

Navigating the Flutter/Dart learning curve as a team with no prior deep experience in the framework.