New guidelines for the use of artificial intelligence in peer reviews
Reviewers may use AI to assist them in evaluating grant proposals, but it should not replace their own judgment. Confidential proposal data must be protected at all times. Accordingly, the SNSF has updated its guidelines.
Researchers seeking SNSF funding must submit a grant proposal. These proposals are evaluated by international experts, who then submit written recommendations. Artificial intelligence (AI) can assist with this process by streamlining tasks like writing, translation and research. To ensure that reviewers leverage this option responsibly, the SNSF has updated its guidelines.
Human expertise remains crucial
The core principle is that reviewers evaluate proposals based on their personal expertise and experience. While AI can support their work, it cannot replace it.
First, the reviewers form their own independent opinion. They evaluate both the quality of the proposal and the qualifications of the applicant. Subsequently, AI tools can be used to polish the text of the written recommendations or to assist in literature searches. Reviewers have sole responsibility for their written recommendations.
Confidentiality and data protection
Because grant proposals contain the applicants’ original ideas and intellectual property, reviewers must treat all content as strictly confidential.
The SNSF does not require international reviewers to use AI tools. However, they must implement technical or legal measures to ensure that grant proposal content is not reused – such as for training AI models, which could make it accessible to third parties.
When utilising AI tools, reviewers are required to configure the settings to ensure that data is neither permanently stored nor used for training purposes. They must also disclose their use of AI to the SNSF when submitting their written recommendations.
These guidelines are reviewed on an ongoing basis
The rapid advancement of artificial intelligence brings new possibilities to research and evaluation. To harness these opportunities while mitigating risks, the SNSF continuously reviews and adapts its regulatory framework.
The evaluation process remains unchanged
Neither an individual written recommendation nor an artificial intelligence tool has the authority to determine funding outcomes. The SNSF continues to base its decisions on at least two external expert assessments. Members of the relevant evaluation panel evaluate these and form their own conclusions. The panel then convenes to deliberate and finalise funding decisions. In this way, the SNSF ensures that decisions continue to be based on human expertise and accountability.