dku-comsci-econ206-2026/commitment-bond-trust-game

Commitment-Bond Trust Game

An interactive COMSCI/ECON 206 PS1 teaching demo with two modes. A persistent Player A (you) versus Player B (programmed AI) board makes the turn sequence and active rule visible while you play. The one-shot study is the proposal's clean cheap-talk-versus-bond comparison. The game states a 50/50 honest/exploitative programmed-receiver prior before forecasts, treats investment as a reliance decision, and reports forecast calibration separately from investment. The game runs entirely in the browser and stores no data.

The eight-round extension contains five cheap-talk rounds followed by three transparent-bond rounds. This fixed order illustrates learning and is not a causal comparison: a future study would randomize or counterbalance the order. The model is a programmed benchmark, not evidence about human participants or deployed LLM behavior.

Project links

• —GitHub: https://github.com/TemurAkhtamjonov/ps1-overleaf-template
• —Google Colab: https://colab.research.google.com/drive/1EwIGvrd73aRVu7Y1Rci-PaBxlrlTwzzC
• —Interactive demo: https://huggingface.co/spaces/dku-comsci-econ206-2026/commitment-bond-trust-game