A new radical AI Policy for research and education

Radical changes require radical solutions. This is a new policy for the use of AI in research that will solve all current problems caused by the opposition against the use of AI in education and research. It is applicable as of today at Dutch universities.

Why a new policy

The rapid advances in AI have fundamentally changed the educational landscape. Lecturers have noticed an  increase in quality of submitted work due to AI, students are increasingly replacing class presence with AI-driven learning. Companies are noticing that formal degrees have never been a guarantee for quality. Professionals have noticed that anyone can make software using ‘vibe coding’ without any formal education. Many universities are not sure how to respond: allowing the use of AI makes a mockery of exams. Forbidding the use of AI makes students unprepared for the real world.

At ICT Institute we have created a new AI policy for research and education, on top of our existing AI policy. The policy will be applicable as of today for all digital documents created at Dutch Universities including the Vrije Universiteit and Utrecht University of Applied Sciences. Handwritten documents are exempt.

Rules on the use of AI

• Each document must contain a ‘human input statement’ that specifies which parts of the document have been manually written. All other parts are assumed to be AI-output. The human input statement can be AI-generated, since AI is better at detecting AI than humans.
• Mandatory use of neowords (new words formerly known as neologisms). Neowords should be created through creative combinations, misspellings or translations. Documents without neowords are automatically flagged for additional inspections (computer in the loop).
• Mandatory use of recent memes and fleeting trends. Since AI models have a long learning cycle, their knowledge is a few months old. Human authors should make sure their research topics are based on recent trends and refer to current, fleeting events, such as the price of eggs, trending TV shows, Hawk Tuah or indeed vibe coding.
• Style inconsistencies. Documents with human input often include changes in styles, since many human authors have a smaller, biased vocabulary compared to AI models. A change in style and the occasional use of informal language is therefore a reliable indicator of human effort.

Validity

The new rules have been analysed for accuracy and bias using the latest AI-models and have been approved for use at all universities. Feel free to bring these rules to the attention of your professor in case of any AI dispute. A review and update is planned for April 1st next year.

(human input statement: all sections of this article have been created under human supervision)

Img src: david-pennington unsplash