The study, published in Computers and Composition, was led by Abram Anders of Iowa State University and Emily Dux Speltz of Embry-Riddle Aeronautical University. Anders and Dux Speltz designed an experimental "AI and Writing" course and followed 38 undergraduates from 22 majors over two semesters.
Students completed structured assignments, reflected on their processes, and documented how their thinking changed while experimenting with generative AI tools. Early beliefs such as "AI will do the work for me" were challenged. The researchers describe several threshold insights: writing with AI is experimental and requires testing; AI output can be polished but misleading, a problem they call the "fluency trap"; and AI should augment, not replace, human decision-making.
Students who adopted these ideas reported becoming more reflective, critical, and intentional in their writing.
Difficult words
- experimental — made to test or try something new
- undergraduate — a university student before a degreeundergraduates
- generative — able to produce new content or ideas
- threshold — an important point for a new understanding
- fluency trap — smooth writing from AI that hides errors
- augment — to add to something to improve or help
- misleading — giving a wrong idea or false impression
- intentional — done on purpose with clear plan
Tip: hover, focus or tap highlighted words in the article to see quick definitions while you read or listen.
Discussion questions
- How could the "fluency trap" change the way you check text from an AI tool? Give one example.
- Which course activity (structured assignments, reflection, or documentation) would help you learn to use AI better? Why?
- Do you agree that AI should augment, not replace, human decision-making in writing? Explain your view.
