What makes an action sequence enjoyable to watch?

This work will be presented at CogSci 2026. See the project page for an overview.

This repository contains the stimulus-generation code, experiment, and analysis notebooks for our study on what makes action sequences enjoyable to watch. The project uses short Flappy Bird-style gameplay videos as controlled stimuli and asks participants to judge how enjoyable, dangerous, or difficult those trajectories appear.

At a high level, the experiment tests how enjoyment relates to two computationally defined properties of a trajectory:

• dangerousness: how close the observed path comes to failure or risky states.
• map difficulty: how hard a map is for agents to complete.

Repo Structure

• stimuli/: stimulus construction, upload notebooks, and the Flappy Bird stimulus-generation pipeline.
• stimuli/generation_pipeline/: scripts for recording trajectories, generating maps, scoring dangerousness and difficulty, selecting stimuli, and rendering videos.
• experiment/: jsPsych browser experiment plus Node/Socket.IO services for stimulus assignment and data writes.
• analysis/: notebooks for cleaning behavioral data, estimating ratings and pairwise preferences, and relating human judgments to computed scores.
• data/: anonymized analysis-ready CSV files used by the notebooks.
• environment.yml: Python environment for stimulus generation and analysis.