Do Vision Language Models Understand Human Engagement in Games?
This work addresses the problem of automated engagement inference for game design and player-experience research, but it is incremental as it highlights limitations in existing VLMs rather than proposing a new solution.
The study evaluated whether vision-language models (VLMs) can infer human engagement from gameplay video, finding that zero-shot predictions were weak and often outperformed by simple baselines, with improvements from retrieval-augmented prompting limited to pointwise prediction and theory-guided prompting ineffective.
Inferring human engagement from gameplay video is important for game design and player-experience research, yet it remains unclear whether vision--language models (VLMs) can infer such latent psychological states from visual cues alone. Using the GameVibe Few-Shot dataset across nine first-person shooter games, we evaluate three VLMs under six prompting strategies, including zero-shot prediction, theory-guided prompts grounded in Flow, GameFlow, Self-Determination Theory, and MDA, and retrieval-augmented prompting. We consider both pointwise engagement prediction and pairwise prediction of engagement change between consecutive windows. Results show that zero-shot VLM predictions are generally weak and often fail to outperform simple per-game majority-class baselines. Memory- or retrieval-augmented prompting improves pointwise prediction in some settings, whereas pairwise prediction remains consistently difficult across strategies. Theory-guided prompting alone does not reliably help and can instead reinforce surface-level shortcuts. These findings suggest a perception--understanding gap in current VLMs: although they can recognize visible gameplay cues, they still struggle to robustly infer human engagement across games.