Robert Joel Deacon. 2024. ChatGPT and Human Agency: A Synthesis of Form and Meaning. PDF
Writing is often associated with thinking. Since Large Language Models (LLMs) such as ChatGPT allow writers to bypass much of the traditional writing process, there is concern this technology will diminish critical thinking development – particularly in student writers. Hegelian synthesis writing (dialectic writing) promotes critical thinking by challenging writers to see more than one perspective and to create novel solutions. The ability to steelman alternative positions and negotiate a novel outcome involves higher-order thinking. LLMs like ChatGPT mimic the form of dialectic writing with increasingly impressive results. This paper, however, shows how ChatGPT3.5 fails to match language form with actual higher-order thinking, and how ChatGPT version 4o fails to truthfully evaluate texts. ChatGPT3.5 appears incapable of choosing a winner when the debate is controversial, and while version 4o shows significant improvement in decision-making, it fails to produce truthful textual evaluations when ranking essay revisions. In such cases, only human writers/thinkers can evaluate texts truthfully and human thinkers are needed to create specific, novel solutions. Instructors can discourage academic dishonesty by focusing on higher-order thinking and by being aware of output patterns to discern LLM output from student output. Examples illustrating ChatGPT’s weak performance in evaluation, decision making, and thesis creation are provided. These results suggest writing instructors should challenge students to evaluate multiple drafts of the same essay and to add specific claims to LLM synthesis conclusions.
Keywords
academic writing, critical thinking development, AI, ChatGPT, dialectic writing
AWCT 2024