No ChatGPT if you work for NWO

Reviewers and employees involved in the allocation of research funding are not allowed to use AI programs such as ChatGPT and Perplexity. This is stated in a new guideline issued by the Dutch Research Council (NWO).

Last year, NWO already drew a line in the sand: jury members, referees and committee members were temporarily prohibited from using generative AI. This ban has now been made permanent.

Tempting

Researchers spend a great deal of time applying for funding from NWO, despite having little chance of success. Reportedly, this is one of the factors contributing to the high workload at universities.

Moreover, evaluating all these applications is largely a labour of love. NWO asks a panel of researchers to rank applications based on their quality, relying on expert referees from around the world. It is a time-consuming and demanding process.

So, it may well be tempting to take shortcuts. Generative AI can produce texts that sound plausible by using language models trained on vast amounts of data.

Risks

However, there are several pitfalls associated with its use. This is why NWO decided to draft a guideline.

According to the guideline, applicants themselves can decide whether to use AI. NWO permits the use of AI in funding applications, provided that applicants are aware of the risks, such as fabricated references, embedded biases and plagiarism. Transparency is key, the guideline states.

For reviewers of funding applications, however, the use of ChatGPT and similar programs is strictly prohibited. Even uploading an application would violate confidentiality, according to NWO. Moreover, the council questions whether assessments made by AI can be considered reliable.

Approved tools

NWO staff members (lawyers, policy officers, etc.) are only allowed to use pre-approved AI programs, excluding ChatGPT, Gemini and MidJourney.

Even approved AI tools may only be used to a limited extent. For example, they cannot be used to draft preliminary versions of justifications or defence documents in objection procedures.

Currently, hardly any AI programs have been approved. According to the NWO website, there is only one: “At this moment, NWO staff may use the DeepL translation application.”