Skip to Content
logologo
AI Incident Database
Open TwitterOpen RSS FeedOpen FacebookOpen LinkedInOpen GitHub
Open Menu
Discover
Submit
  • Welcome to the AIID
  • Discover Incidents
  • Spatial View
  • Table View
  • List view
  • Entities
  • Taxonomies
  • Submit Incident Reports
  • Submission Leaderboard
  • Blog
  • AI News Digest
  • Risk Checklists
  • Random Incident
  • Sign Up
Collapse
Discover
Submit
  • Welcome to the AIID
  • Discover Incidents
  • Spatial View
  • Table View
  • List view
  • Entities
  • Taxonomies
  • Submit Incident Reports
  • Submission Leaderboard
  • Blog
  • AI News Digest
  • Risk Checklists
  • Random Incident
  • Sign Up
Collapse

Report 6291

Associated Incidents

Incident 12451 Report
Norwegian Student Reportedly Used AI-Generated Deepfake Videos in Spanish Coursework at University of South-Eastern Norway

Loading...
Student tried to fool examiner with fake video: — Special and disturbing
khrono.no · 2024

In the one-year Spanish program at the University of Southeast Norway, students are required to submit self-produced videos on a given topic, in which they speak Spanish.

--- Occasionally there are some technical challenges, such as videos freezing. But this one stood out a bit, says subject manager Hans Jacob Ohldieck.

The Norwegian student had perfect Spanish pronunciation, but she occasionally made completely elementary mistakes. There was a clear difference in level within the oral presentation. She systematically pronounced some words with an English accent. And the image of the candidate had a "blurry and artificial feel".

In total, the student submitted three different videos in the fall of 2023. There were two work requirements, as well as part of an exam, in two subjects.

All of them had such strange features. There was a lack of correspondence between speech and mouth. Despite her high level, the student pronounced well-known references such as Buenos Aires and the name Jorge Luis Borges --- a famous writer --- completely wrong.

 --- This is a completely new type of case for us, we have never had similar cases, says Ohldieck.

Had never met the student 

The university's unit for digitalization and educational quality took a look at the videos. They also had no experience with such issues, but thought the videos could indicate cheating.

The one-year study is an online and meeting-based study. Since the student had not shown up for the meetings, the lecturers had not met her in person.

They called her in for an interview to check whether she spoke Spanish at the level she gave the impression of, but did not get an answer.

Was found guilty of cheating 

The case was eventually treated as a cheating case, and she ended up being found guilty of cheating by the Student Affairs Board.

The board believed the videos were fake.

Such so-called “deepfake” videos, in which a fake version of a person’s voice and face is created, abound on the internet, and there are a number of free tools for creating them.

One such video, in which a fake presenter Fredrik Solvang tricked viewers of Debate on NRK, attracted attention earlier this year.

Banned for one year 

The student's reaction was the cancellation of two work requirements and an exam, as well as a ban for both semesters in 2024.

The board writes:

"The board assumes that there is objectively cheating in this case, and refers to the examiner's statement, which states that it has been revealed in the three answers that the student has an "artificial and mechanical voice and pronunciation", that there is a "lack of correspondence between speech and mouth movements", "the face looks manipulated", and the student pronounces familiar words incorrectly, and that the answers have "a general level that is typical of AI-generated texts".

According to the case papers, she has also admitted to using artificial intelligence (AI) when preparing the answers, without this being further elaborated on in the case papers.

It says that she has not submitted a written statement to the board, and she has withdrawn from the study program

--- Not prepared for


Khrono has not seen the videos herself, the university does not want to share them.

--- What did you think when you saw the videos?

--- I thought it was special and disturbing, says subject manager Hans Jacob Ohldieck.

He says that they have experience with artificial intelligence, and that this is high on their agenda.

--- But that we received a manipulated video, we were not prepared for that. It is also a kind of warning. Now this was not so sophisticated that you could not detect it. But this form of cheating will become more sophisticated. That makes it challenging to use these videos as part of the assessment basis.

He says that such videos have been useful, among other things since it has been easy to give feedback on language and pronunciation. 

They use videos largely as a kind of training for students in the run-up to the exam.

Will change the exam format 

Due to the rise of artificial intelligence, they are now likely to change the exam format.

--- We now have to tighten up, I think, both with the help of school exams and oral exams in real time, says Ohldieck.

When it comes to mandatory submissions, including videos, the university is in the think tank.

--- We are going to do something. But this type of practice assignment is valuable for the students, so I am a little unsure of how we should change it. 

In Spanish, they experience that many students use AI in written submissions.

--- In the written submissions last year, I would estimate that between a quarter and a fifth have clear use of artificial intelligence, to a very extensive extent.

--- But aren't that many people caught cheating?

--- No, it is a big challenge to prove cheating in this context, and this is a challenge that we are also discussing with colleagues from other subjects. Due to the extensive use of AI, one solution could be to make the mandatory submissions less open and more tied to specific syllabus texts, while it varies from subject to subject how appropriate and feasible this is.

Inga Strümke: --- A bit naive


Artificial intelligence expert Inga Strümke is not surprised that fake videos have reached higher education.

--- It's very easy to use, especially if you're willing to spend a little money. So it's not surprising that they get to use it. But it surprises me that once they spend time studying, they choose to cheat their way through it, says the NTNU researcher.

She thinks it seems a bit naive of the university to ask students on an online course to submit videos.

--- A couple of years ago, it would have been a perfectly normal way of evaluating. Now, of course, people are more likely to cheat. It's probably a bit naive to ask students to submit pre-recorded videos. Then you can tinker with it and regenerate it until you're satisfied, she says.

Read the Source

Research

  • Defining an “AI Incident”
  • Defining an “AI Incident Response”
  • Database Roadmap
  • Related Work
  • Download Complete Database

Project and Community

  • About
  • Contact and Follow
  • Apps and Summaries
  • Editor’s Guide

Incidents

  • All Incidents in List Form
  • Flagged Incidents
  • Submission Queue
  • Classifications View
  • Taxonomies

2024 - AI Incident Database

  • Terms of use
  • Privacy Policy
  • Open twitterOpen githubOpen rssOpen facebookOpen linkedin
  • e1b50cd