Declaration template
Based on the current praxis at major scientific publishers, we have designed this template for declaring AI usage. The declaration should be placed at the beginning of the assignment so that the reader/examiner knows that AI has been used. [Text in square brackets is a suggestion for prompts and needs to be adapted to individual needs].
If you have used AI in your assignment, use one or more of the suggested templates below:
Declaration of the use of generative artificial intelligence
In this scientific work, generative artificial intelligence (AI) has been used. All data and personal information have been processed in accordance with the University of Oslo's regulations, and I, as the author of the document, take full responsibility for its content, claims, and references. An overview of the use of generative AI is provided below.
(List bullet points with a description of usage, see examples below.)
Information on what it means to process data in accordance with UiO's regulations can be found further down.
Example: use for language editing
The service GPT UiO, developed by UiO IT department, has been used to improve the content of the report/assignment. A first version of the work was pasted in its entirety, and the model was given the prompt [rewrite this text to make the language more lively.] The text was then iterated a few times through the model where new prompts were used to get the correct structure of the text. The final result was cut out, fact-checked, and partly rewritten by the author(s).
Example: use for image generation
The AI service DALL-E 3, developed by OpenAi, has been used to produce the illustration in figure [X]. The specific prompt for producing the image was [Create an image of a generic superhero clearly strapped to a rotating wheel. The scene is charged with energy, with dynamic lines to illustrate the spinning motion. The background is a blur, enhancing the feeling of speed and urgency.] The prompt for producing the image was created through brainstorming with Open AI's ChatGPT-4.
Example: use for brainstorming
The service ChatGPT-4, developed by OpenAI, has been used for brainstorming to find a suitable working hypothesis.
Example: use as an analytical tool
The service GPT UiO, developed by UiO IT department, has been used as an analytical tool to look for trends and patterns in anonymized data. The result was compared with other analytical tools, and the usage is described under the methods section.
Data processing in accordance with UiO's regulations
To use AI services, you need to know which data you are allowed to use where. At the University of Oslo, divide data by color codes - green, yellow, red, and black data, and you as a user of generative AI must be sure not to send data into non-permitted services. More information about different types of data can be found on USIT's websites about classification of data and information.
Green data can be used with any service. For example, you can simulate the motion of a pendulum, print out the data from the simulation (which in this case are green data), send it into ChatGPT, and ask it to describe the motion for you.
If you wish to analyze data that is yellow, red, or black, you must use a service that UiO has a data storage agreement with. More information about which services UiO has agreements with can be found on USIT's websites about IT services.
Not all use is good use
Although this document provides guidelines on how generative AI can be used in scientific work, it is your responsibility to assess whether the use is sensible. For instance, if you use generative AI to generate a response from nothing, you have less control over whether this could be unintended plagiarism (read more in this exploratory article about plagiarism and paraphrasing using GPT UiO). Consult with your supervisor/teacher what is acceptable use of generative AI and document precisely.
Uncritical use of AI can also damage your learning. There is also a risk that AI can generate text that is more or less similar to other text, and it can also make up fictional claims and sources. It is therefore important to carefully investigate all claims, and we do not recommend that you let AI generate large amounts of text for you. If you let AI generate parts of your text, you should carefully work on these parts and make the text your own. To be on the safe side, it might be wise to first write a text yourself and ask generative AI for feedback or to rephrase some parts.
Praxis in the major scientific publishers
The guidelines proposed above are based on the praxis at major scientific publishers. In the table below, you find the guidelines at some of them as they stand per 03.04.24.
| Springer Nature | Elsevier | Wiley | |
|---|---|---|---|
| AI as co-author? | No. Source: Nature portfolio editorial policies on AI (read 03.04.24). | No. Source: Elsevier guide for authors (read 03.04.24). | No. Source: Wiley ethic guidelines (read 03.04.24). |
| Use of AI for text handling | Yes. It is not explicitly stated that Springer Nature accepts text handling with AI, but that the use of AI tools in writing must be adequately documented in the method section. Kilde: Nature portfolio editorial policies on AI (lest 03.04.24). | Yes, but only to improve readability and language, and the use must be declared. Simple use for grammar and reference handling does not need to be declared. Kilde: Elsevier guide for authors (lest 03.04.24). | Yes, but it must be declared. Simple use for grammar and reference handling does not need to be declared. Journals may have their own rules. Wiley ethic guidelines (lest 03.04.24). |
| Use of AI-Generated Images |
Nei. Exceptions apply for services with which Springer Nature has special agreements, or images where AI generation is an essential part of the research - decided on a case-by-case basis. Source: Nature portfolio editorial policies on AI (read 03.04.24). |
No. Source: Elsevier publishing ethics (read 03.04.24). | Possibly. The absence of central guidelines suggests that Wiley in principle allows this, but journals under it often have rules against such images. Source: Information for Authors - ACES journals (read 03.04.24). |
| Use of AI services in peer review | No. Exceptions apply for safe AI services with which Springer Nature has an agreement. | No. Source: Elsevier publishing ethics (read 03.04.24). | Possibly. The absence of central guidelines suggests that Wiley in principle allows this, but journals under it often have rules against this. Source: Responsible use of AI - JMS journal (read 03.04.24). |
Elsevier provides this template for the declaration of AI in their articles.
"Statement: During the preparation of this work the author(s) used [NAME TOOL / SERVICE] in order to [REASON]. After using this tool/service, the author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the publication."
Kilde: Elsevier guide for authors (lest 03.04.24).
Based on this, we have created the variant that is given above.