Enabling educators to assess authentic student work in the age of Generative AI

When I was a teacher with around 25 students, I could often tell if students handed in work that was not original based on my knowledge of their writing style. When in doubt, I had a conversation to measure the students’ understanding of what they wrote, and this is how I validated authenticity before I assessed the students’ work.

With the introduction of Generative AI, it is difficult to determine the author of any text. While the above process has proven effective and is used by teachers daily to confirm the authenticity of a student’s work, it is not scalable and is open to human bias. Norvalid saw an opportunity to solve the challenge of determining authenticity in the age of AI by making the above approach scale. We do this by extracting the student’s linguistic fingerprint from a writing sample, which we compare against submissions, and we confirm the student’s knowledge by asking automatically generated questions from the content they submitted.

Generative AI is similar to ghostwriting.

Traditional text-matching has played a significant role in combating plagiarism after the internet came along in the late 1990s. While “detection” effectively identifies copied text and can be used to combat plagiarism, this is not a viable strategy related to generative AI. Research on AI detection is clear; it is unreliable and does not produce any evidence. Generative AI can be compared with the age-old problem of ghostwriting and contract cheating, as neither are original students’ writing.

Verify the absence of other writers without detection.

As we need a new approach to assessment security in the age of AI, we also want to change the narrative. Instead of detecting cheating, we are moving towards a positive strategy of validating authenticity, thereby confirming honesty. By validating a student’s original writing, we verify the absence of others’ writing, such as copying, collusion, contract cheating and the use of generative AI. Moving from cheating detection to validating authenticity allows organisations to promote a viewpoint that students are rewarded for submitting original work.

Authenticity when it matters

At Norvalid, we aim to enable educators to assess authentic student work. We ensure that universities know their graduates are capable of what their degree requires. We enable educators to embrace generative AI, since it is a critical workplace skill, while at the same time ensuring that they are assessing original student writing.

To learn more about our approach, please do reach out to contact@norvalid.com.

Written for OEB Global 2024 by Garnet Berry (Garnet@norvalid.com), Norvalid (Formerly Nor.Education).