(88) Identifying games to promote discussions on environmental issues using natural language processing
Date:
Contributors: Edlinger, M. , Huber, S. E., Kogler, M., & Ninaus, M.
Venue: EARLI 2025, Graz, Austria, August 25-29, 2025
Abstract: Video games are increasingly recognized as effective tools for educational purposes and discussing societal issues, including climate change. However, among the thousands of games published each year, the identification of potentially useful games is difficult. Thus, this study explored the use of player-generated reviews to identify games that stimulate dialogue on climate-related matters and could be useful in both educational settings and research. Utilizing natural language processing (NLP) techniques such as zero-shot classification and embedding-extended dictionary-based analyses, we compared player-generated reviews from 45 so-called eco games to 45 action-adventure games, totaling 300,754 reviews available on Steam. The results of the zero-shot classification showed that eco games prompt more climate change discussion than action-adventure games, yet the frequency varies widely among eco games. While the topic of climate change constituted 2.2% of discussed topics across eco games, in certain eco games the level of discussion of climate change-related topics revolved around 9 to 11%. The dictionary-based analyses corroborated these findings, detecting 1104 climate change-related terms in the reviews of eco games, compared to 14 terms in the reviews of action-adventure games. Taken together, player-generated reviews seem to be a powerful resource for identifying games that encourage debate on environmental issues. The NLP techniques used provide a scalable and objective method that could significantly assist educators and researchers in identifying promising video games to promote discussions on climate change without needing extensive gaming expertise. Further research is needed to validate these approaches through, for instance, media comparison studies.