Playful Interactions with AI

Characterizing Emergent Playful Interactions with ChatGPT

Recent advances in generative artificial intelligence have transformed how people interact with computational systems. Rather than using AI solely as functional tools, users increasingly engage with them in exploratory and playful ways to understand their capabilities, limitations, and implications. Despite the growing prevalence of such behavior, playful interactions with AI systems remain underexamined in Human-Computer Interaction (HCI) research.

We address this gap by investigating how users engage in playful interactions with a widely used conversational AI system, ChatGPT, and what these interactions reveal about emerging human-AI relationships. To explore this phenomenon, we conducted a thematic analysis of 372 user-generated posts collected from the ChatGPT subreddit on Reddit. The analysis first distinguished between non-use discussions, practical interactions, and playful interactions, revealing that a majority of posts (54%) centered on playful engagement with the system. We then performed an inductive analysis of the playful posts to characterize the forms these interactions take.

Our findings identify six major categories of playful interaction: reflecting, jesting, imitating, challenging, tricking, and contriving, each encompassing several subcategories and interaction patterns. Together, these categories form a preliminary framework describing how users playfully interrogate AI systems to probe their reasoning, test their limits, explore creative possibilities, or circumvent constraints.

This work contributes to HCI and CSCW research in three ways. First, it provides an empirical characterization of emergent playful interactions with conversational AI. Second, it introduces a conceptual framework that offers vocabulary for describing these interactions and their underlying intentions. Third, it highlights how playful engagement can reveal users’ perceptions of AI agency and inform the design of future AI systems that better support exploration, creativity, and meaningful human-AI relationships.

Identifying Playful Design Elements that Foster Curiosity and Exploration in AI Interaction (ongoing)

Users often engage with AI systems in unexpected and playful ways, using experimentation, humor, and exploration as strategies to better understand the system’s capabilities and limitations. These observations suggested that playfulness is not merely a peripheral behavior but can serve as an important mechanism for learning, sensemaking, and discovery when interacting with AI technologies. Motivated by these findings, the current project seeks to examine how playfulness can be intentionally designed into interactive systems to support users’ curiosity, exploration, and creative engagement with emerging technologies.

In this project, we currently conducting a literature review to identify playful, gameful, and ludic design elements that foster curiosity-driven exploration. We compiled a collection of relevant academic papers and are analyzing them to extract the specific design features and interaction mechanisms that encourage playful engagement, experimentation, and open-ended discovery. The goal of this analysis is to build a structured understanding of how different design approaches, such as ambiguity, metaphor, social play, and exploratory interaction, can stimulate curiosity and deeper engagement with technological systems.

The expected contribution of this work is a conceptual and practical framework that organizes and synthesizes these playful design elements into a coherent library. Such a library can serve as a resource for designers and researchers developing interactive systems, particularly AI-powered tools, that aim to support exploration, learning, and creative use. Ultimately, this research highlights playfulness as a valuable design strategy for enabling users to better understand and appropriate emerging AI technologies, while also contributing to broader discussions in HCI about curiosity-driven interaction and discovery-oriented design.

Narrative game for studying LLM pitfalls in everyday decision-making (ongoing)

Large language models are becoming part of everyday life. People use them for learning, writing, and planning, often because they are fast and easy to access. However, many users do not fully understand how these systems should be used, where they can be helpful, and where human judgment still matters. This creates an important educational challenge: how can we improve AI literacy in a way that is engaging, memorable, and relevant to ordinary life? This project is motivated by the need to help people develop a more balanced understanding of LLMs, not by rejecting them, but by showing both their usefulness and the risks of overreliance. And we think that game and play are perfect channels for it.

Concretely, we are developing a narrative-driven serious game that explores the role of LLMs in everyday decision-making and social interaction. Rather than teaching through direct explanation, the game places the player inside realistic interactive scenarios supported by conversations, guided investigation, and mini games. A key goal of the project is not only to encourage critical thinking, but also to show that LLMs can be extremely useful when they are used appropriately. The game therefore aims to present a balanced perspective: it educates players about the importance of careful use, while also showcasing situations in which these tools can provide meaningful support and practical value.

The expected contribution of this ongoing project is an interactive educational and research artifact that helps players think more carefully about how LLMs should be used, when they can be helpful, and why human judgment still matters.