An Academic's Experiment in AI-Assisted Text Generation and Detection

Serdar HocamAuthor & Editor

An academic conducted an experiment in writing an academic article using ChatGPT and an AI detection tool, and evaluated the results.

◉ 0 views
My Adventure in AI-Assisted “Writing” and Detection

Amid concerns that students might misuse large language models, an academic conducted an experiment in writing an article on international criminal law using ChatGPT and AI detection tools like Pangram.

AI Anxiety in the Academic World

One of the greatest concerns for contemporary academics is the misuse of large language models by students in their assignments and theses. Studies suspected of being written by artificial intelligence, particularly in master's theses, are drawing attention.

Academic Tasks of Artificial Intelligence

According to ChatGPT, large language models can be used in many academic tasks such as research assistance, source comprehension, literature synthesis, question and argument development, structural planning, editing, rewriting, and text generation.

Experimental Article Production Process

Following a lengthy report prepared for the Danish Ministry of Defense, the author experimented to produce an academic article that would be perceived as entirely human-made. Using ChatGPT 5.6 Sol, a 6,000-word article was requested based on a 150-word abstract.

Detection Tools and Ethical Boundaries

The text, generated in nineteen minutes, was examined by running it through the Pangram detection tool. After a small section was edited manually, the tool evaluated the text as 100 percent human-made, though the author did not find this ethical.