Uses WildChat
A paradox of AI fluency
arXiv preprint · First arXiv submission
Relates inferred AI fluency to task complexity and conversational outcomes. The study compares active iteration, critical evaluation, and passive acceptance, and examines how these behaviors relate to visible and invisible failures.
Data used
27,000 English transcripts sampled across 27 months from the non-toxic WildChat-4.8M release, with separate analyses of recurring Midjourney and Blockman interaction clusters.
Findings
More fluent users attempt harder tasks and show more visible failures, yet achieve greater success on complex work; less fluent users more often have failures without overt negative feedback.
Sections 3–5: monthly sampling, cluster handling, model-assisted annotation, and outcome analysis. Read the source.
Dates and versions
First arXiv version. ArXiv version history.
arXiv preprint.
Cite this paper
Christopher Potts, Moritz Sudhof. 2026. A paradox of AI fluency. arXiv preprint.
Show BibTeX
@misc{aifluency2026,
title = {{A paradox of AI fluency}},
author = {Potts, Christopher and Sudhof, Moritz},
year = {2026},
doi = {10.48550/arXiv.2604.25905},
eprint = {2604.25905},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2604.25905}
}