GPT-4o raised the marketing assignment grades by almost one point on a five-point scale in an experiment involving 1 053 Bocconi University students. The Decoder reports this.
At the same time, the description of the experiment does not state whether researchers checked what the students actually learned. Therefore, the higher score in this case is not equivalent to a confirmed increase in knowledge.
The Decoder also notes other research indicating that completing assignments with the help of AI without independent thinking causes long-term harm. The provided package lacks details about these studies, so this claim requires separate verification.
editorial commentary
Why it matters
A probable consequence is an intensification of disputes about what exactly educational grades measure when using AI. The next observable signals will be studies that compare final scores with independent verifications of knowledge. Substantial uncertainty remains: the provided material lacks experimental methodology, data on long-term learning, and details of the mentioned studies.