ChatGPT: The Future of Queer Play?

This paper examines whether large language models such as ChatGPT can resolve a tension between rule-based, heteronormative video game structures and the desires of queer players, who often seek to destabilize normative systems through play. Drawing on scholarship on queer play, the author distinguishes invasive strategies that alter game logic, such as cheating and modding, from non-invasive strategies like imaginative play and role-play that operate within existing rules. The central claim is that LLM-driven non-player characters can expand the space for queer play by allowing players to speak freely rather than choose from scripted dialogue options.

The empirical evidence comes from a playtest of a ChatGPT-powered murder-mystery prototype with 36 participants, of whom 20 (56%) engaged in off-task, exploratory queer play. This finding supports the hypothesis that ChatGPT can facilitate queer play. A second study compared queer and conventional play groups using the miniPXI. Overall player experience did not differ meaningfully between groups, but queer players reported significantly higher perceived autonomy and a stronger intention to break the game. Both groups rated ChatGPT's roleplaying performance similarly across sensibility, naturalness, and perceived human-likeness.

The scope is narrow: the samples are small, the queer-play categorization rests on the author's own interpretation of player inputs, and results are not broken down by identity group, despite 31% of participants identifying as queer or LGBTQ. Data were collected in November 2023. The author concludes that queer play does not harm player experience and that LLMs can support diverse play styles, though autonomy did not significantly predict overall satisfaction, leaving the second hypothesis only partially supported.