AI agents are attending students' online courses, and universities are finding it difficult to prevent this.

      Technology has evolved from merely assisting with homework to effectively replicating students in virtual classrooms.

      AI cheating has progressed from simply requesting a chatbot to write an essay to employing tools that can make AI-generated content appear human. We've discussed AI tools that facilitate student cheating, the increasing issue of AI detectors struggling to keep pace, and how frequently students are already utilizing AI for their assignments. The technology has advanced beyond merely aiding students; in certain instances, it can actually act as the student.

      Simply instruct the AI to attend the class

      According to a New York Times article by Dana Goldstein and Alan Blinder, students can take it a step further by delegating an entire online course to an AI agent. The report states that AI agents can be given basic commands such as “log in and take my quiz,” allowing them to navigate online learning platforms. They are capable of watching pre-recorded lectures, completing tests, writing papers, and even engaging in conversations with other students.

      This is particularly concerning because online learning already relies on digital, often asynchronous coursework. Students can complete lectures, assignments, and exams remotely, sometimes without any direct interaction with their professors or peers. The report notes that more than half of American college students enrolled in at least one online course last year.

      Additionally, the AI doesn’t need to reveal its identity. The Times evaluated top AI tools such as ChatGPT, Gemini, and Grammarly and found that none declined requests to write papers for students. Furthermore, AI companies have not restricted agents from accessing platforms like Canvas, Brightspace, and Blackboard, although these services are exploring ways to detect agent usage.

      The challenge of enforcement grows even greater because these agents can perform tasks beyond simple assignments. Online learning platforms can look for indicators like the duration a student spends on an assignment, but the New York Times highlights that there isn’t a foolproof way to prevent agent-based cheating at this time. Some educators have even requested AI companies to require their agents to identify themselves within online courses, but these companies have resisted this request due to privacy and security concerns.

      The issue extends beyond just identifying cheaters

      This presents a particularly awkward dilemma for universities. AI detection tools can produce inaccuracies, making it difficult for professors to conclusively prove a student has cheated, while administrators may hesitate to impose penalties without solid evidence. Consequently, responses can be inconsistent; some professors may introduce in-person exams, while others may simply permit the use of AI.

      In response, some educators are designing online courses that focus on work that is harder for AI to replicate, such as personal reflections and discussions about what students found challenging or engaging. Others are reinstating certain exams into physical classroom settings.

      However, this raises a larger question regarding the essence of online education when an AI agent can attend classes in place of the individual paying for the education. The concern is no longer limited to whether AI can assist students in cheating on assignments; it now concerns whether a student can earn an entire degree with minimal actual learning. For colleges that have long argued that online education enhances accessibility to higher education, this presents a particularly uncomfortable dilemma.