Beyond Asking: A Pipeline for Personalized Game Generation that Reads Players from Behavior
2026-08-17 • Artificial Intelligence
Artificial IntelligenceHuman-Computer Interaction
AI summaryⓘ
The authors study how to understand a player's unique style and skill by analyzing their gameplay using large language models (LLMs). They create a special set of artificial players with known traits to test how well different methods can guess these traits just from actions in the game. To improve accuracy, they introduce a way to tell if a player had the chance to do something but chose not to, separating preference from opportunity. The authors find that while LLMs do better than some simple methods, traditional supervised models still perform best. They also show that using inferred player profiles can help adjust game difficulty and test if these methods work with real human players.
personalized game generationlarge language modelsplayer profilingbehavioral inferencesynthetic player populationtrait inferencedecision-moment representationdifficulty adaptationfew-shot learningsupervised regression
Authors
Yifan Lu, Xiaopeng Yuan, Haohan Wang
Abstract
Personalized game generation requires inferring a player's abilities and behavioral style from how they play. Large language models have made this inference more attainable than ever: an LLM can read a raw gameplay transcript and produce a fluent, plausible profile of the player. Plausible, however, is not verified, and verification is precisely what the field lacks: latent traits are unobservable; questionnaires provide noisy proxies and become circular when self-reports are used to validate behavior-based inference; and behavior itself is ambiguous without context -- a player who never collects an item may not want it, or may never have had the chance. We address both problems. First, we construct a synthetic player population whose traits are ground truth by construction: each trait is an explicit bot parameter, accepted only after controlled manipulation produces consistent, trait-specific behavioral change. Unlike prior parameter-recovery work that inverts a known decision model, our benchmark evaluates policy-agnostic inference from behavioral transcripts alone. Second, we introduce an opportunity-aware decision-moment representation that disentangles preference from the chance to express it; ablating it selectively degrades opportunity-dependent traits. On this benchmark, few-shot LLM inference outperforms embedding- and rule-based baselines on most traits, though feature-based supervised regressors remain stronger overall. Finally, we close the loop: inferred profiles drive difficulty adaptation, evaluated against ground-truth references and mismatched-profile controls, and an exploratory human study examines whether these findings transfer to real players.