Professors Arrested After AI-Generated Apology Emails
Two professors from the University of Illinois at Urbana-Champaign found themselves in a tricky situation recently. They arrested after numerous students sent identical apology emails generated by artificial intelligence, following accusations of cheating on assignments.
Professors Carl Flanagan and Wade Fagen Ulmschneider, who are well-known in the field of data science, confronted students for cheating and for tampering with attendance records in an introductory course. Their suspicions were raised when they received many nearly identical emails of apology from the accused students, which turned out to have been created using AI.
This incident took place on October 17th. During a large lecture, the professors read some of the apologies to the class and projected the emails on a screen. It was clear from the laughter in the audience that students recognized the use of AI in crafting those messages.
The introductory course has about 1,200 students split into two sections and employs a tool called Data Science Clicker for tracking attendance and participation. Students are required to log onto their devices and respond to multiple-choice questions in a limited time frame when prompted with a QR code. Participation is valued at 4% of the final grade, especially critical for first-year students.
Earlier in October, the professors became aware that many students marked as absent were somehow still managing to submit answers. After a closer look at website activity, IP addresses, and server logs, they realized that students were getting advance notice about the questions.
Following this discovery, the professors emailed over 100 students, letting them know their deception had been uncovered. Although the university has policies against cheating and plagiarism, there aren’t explicit guidelines about the use of AI, meaning students involved won’t face disciplinary action as their actions weren’t explicitly against any stated rules in the syllabus.
Former students expressed frustration over the current situation, mentioning that the professors worked hard to make the subject matter engaging and easier to grasp. They felt particularly disheartened, given the introductory nature of the course, that peers resorted to AI-generated shortcuts and inauthentic apologies.
Additionally, a company has developed a tool aimed at detecting AI-assisted cheating, having reported that it processed 22 million papers, with many written by AI, from college students.
The company launched an AI writing detection tool last year, trained on essays from high school and college students, alongside known AI-produced content. It claims that its tools currently have a false positive rate of less than 1% when analyzing entire documents. According to their findings, 11% of papers submitted appeared to contain at least 20% AI-generated text, with 3% being composed of 80% or more AI-written material.
