In the view of the DFG Executive Committee, AI technologies are already changing the entire work process in science and the humanities, knowledge production and creativity to a significant degree and are being used in various ways in the different research disciplines, albeit for differing purposes. In terms of generative models for text and image creation, this development is still very much in its infancy.
"In view of its considerable opportunities and development potential, the use of generative models in the context of research work should by no means be ruled out," says the paper: "However, certain binding framework conditions will be required in order to ensure good research practice and the quality of research results." Here, too, the standards of good research practice generally established in science and the humanities are fundamental.
In terms of concrete guidelines, the DFG Executive Committee says that when making their results publicly available, researchers should disclose whether or not they have used generative models and if so, which ones, for what purpose and to what extent. This also includes funding proposals submitted to the DFG. The use of such models does not relieve researchers of their own content-related and formal responsibility to adhere to the basic principles of research integrity.
Only the natural persons responsible may appear as authors in research publications, states the paper. "They must ensure that the use of generative models does not infringe anyone else’s intellectual property and does not result in scientific misconduct, for example in the form of plagiarism," the paper goes on.
The use of generative models based on these principles is to be permissible when submitting proposals to the DFG. In the preparation of reviews, on the other hand, their use is inadmissible due to the confidentiality of assessment process, states the paper, adding: "Documents provided for review are confidential and in particular may not be used as input for generative models."
Instructions to applicants and to those involved in the evaluation process are currently being added to the relevant documents and technical systems at the DFG Head Office.
Following on from these initial guidelines, the DFG intends to analyse and assess the opportunities and potential risks of using generative models in science and the humanities and in its own funding activities on an ongoing basis. A Senate Working Group on the Digital Turn is to address overarching epistemic and subject-specific issues in this context. Any possible impact in connection with acts of scientific misconduct are to be addressed by the DFG Commission on the Revision of the Rules of Procedure for Dealing with Scientific Misconduct. The DFG will also be issuing further statements in an effort to contribute to a "discursive and science-based process" in the use of generative models.
For the text of the statement, see the DFG website here