Eternagram

Sustainable behaviors to ameliorate climate change consequences are difficult to elicit in people, because they do not always align to people's intrinsic motives. Targeted interventions require knowledge of the background and personality of individuals we want to affect in climate change education. In order to measure individual propensity to climate communication, we created a text-adventure game that allows players to explore a non-Earth world guided by a ChatGPT-based in-game character. We collected in-game dialogues of players conversing freely with the character, and player choices from designed game mechanics, to explore correlations with player personality and climate attitude. We found that while players' political and climate beliefs affect their conversational content in the game, their actions in the game are pro-climate-leaning despite these differences. Our work illustrates the application of ChatGPT-based generative tools to interactive systems designed for positive social purpose, particular the assessment of participant attitudes for more specific interventions.

Publication: Proceedings of the CHI Conference on Human Factors in Computing Systems (CHI'24).