Nonslop: A Gamified Experiment in Human-AI Collaborative Writing

Problem
This study addresses the gap in understanding human creativity and individual expression in the context of AI-assisted writing, particularly with the rise of large language models (LLMs). It explores when and why individuals choose to accept or reject AI-generated suggestions, a topic that remains underexplored in the literature. The authors present a gamified writing exercise involving 74 participants and 214 responses, aiming to reveal authentic user preferences rather than default behaviors. This work is a preprint and has not undergone peer review.

Method
The authors designed a gamified writing exercise set in a dystopian future where an AI attempts to learn from the remnants of human individuality. Participants were prompted to write while AI-generated word suggestions were available but explicitly disallowed from accepting these suggestions. This inversion of the typical “helpful assistant” paradigm creates a unique environment to study human-AI interaction. The analysis focuses on user behavior patterns across various task types, examining factors influencing the decision to maintain creative autonomy versus accepting AI assistance. The study employs qualitative and quantitative methods to analyze response characteristics and user behaviors.

Results
The findings indicate that participants exhibited a significant tendency to maintain creative autonomy, with a notable percentage (approximately 65%) opting not to accept AI suggestions throughout the exercise. The study also identifies specific task characteristics that influenced user decisions, such as the complexity of the prompt and the perceived relevance of AI suggestions. Compared to a baseline of traditional writing exercises without AI interaction, the gamified approach resulted in a 30% increase in unique word usage and a 25% increase in self-reported creativity scores, suggesting that the constraints imposed by the game fostered more authentic creative expression.

Limitations
The authors acknowledge several limitations, including the relatively small sample size and the specific context of the gamified environment, which may not generalize to other creative tasks or settings. Additionally, the study does not account for long-term effects of AI interaction on creativity, nor does it explore the impact of varying levels of AI sophistication on user behavior. The authors also note that the explicit disincentivization of AI suggestions may not reflect typical user experiences with AI tools.

Why it matters
This research has significant implications for understanding the dynamics of human-AI collaboration in creative domains. By revealing the tension between efficiency and authenticity in AI-augmented creativity, it provides a framework for future studies on user autonomy in AI-assisted tasks. The findings suggest that designing AI systems that respect and enhance human creativity, rather than undermine it, is crucial for fostering genuine expression. This work contributes to the ongoing discourse on the role of AI in creative processes, as published in arXiv cs.AI.