Bruk av KI-verktøy i forskning - Ansatt

There are many different AI tools that can be used in research, for example in data analysis, literature searches and writing. AI tools that are available via OsloMet and assessed as safe to use for researchers: SIKT KI-chat (sikt.no), Copilot (microsoft.com), Keenious (oslomet.no) and Autotekst (uio.no).

Accountability and transparency

As a researcher, you are responsible for the content you deliver, regardless of whether you have used AI or not. If you use AI in the work of text development, you are, for example, responsible that the information you reproduce is correct and that any origin is credited (referred to). Remember that it is also considered plagiarism to publish a text written by a machine as your own.

Being transparent is another key principle in research. You must always be able to show what you have done during a research process (method), also when it comes to the use of AI. See, for example, the Norwegian APA manual, page 63 (sikt.no) or the APA style blog (apa.org) for more on how you can refer to AI tools such as SIKT KI-chat.

OsloMet's AI policy also has several good principles for AI use.

Tips for researchers

It is up to you as a researcher to assess whether the way you use AI tools is legitimate and in line with current regulations and legislation.

Check the journals' AI guidelines

When you know which journal you want to publish in, check whether the journal or publisher has guidelines for the use of AI. If you have used AI in the development of your work, you may, for example, have to sign a declaration that is published together with the article, or you must attach documentation on how you have used AI during the research process.

Assess AI tools before use

Training data and sources: What is the tool trained on and/or which sources does it get information from? Is it, for example, like SIKT KI-chat, trained on texts from the internet up to a certain point in time? Or is it a tool like Keenious that finds literature from selected databases, where certain fields of study may be underrepresented?
Characteristics of the source data: What characterizes this type of data? Could there be any weaknesses or biases in this source data that you need to be aware of when using the tool? If the training data are, for example, texts from the internet, then these are rarely completely neutral, which will be reflected in the result the tool produces.
Functionality of the tool: How does the tool work? How are the answers produced? It may seem good, but what does the tool actually do? What are the tool's limitations? Is it, for example, a language model and chatbot like SIKT KI-chat that generates answers based on a “prediction algorithm”, which compiles words that are likely to be related based on the words used in the question or instruction the tool received?
Data security: Who is behind the tool? What do they write about how they use your information and data/texts you share with them (for example, read the terms of use)? How do they relate to legislation on personal protection (GDPR), copyright and the like? What do other websites say about the tool? SIKT KI-chat (sikt.no) is the service OsloMet recommends using if you are going to share copyright protected work or other confidential content that an AI tool should not have access to or be able to train on.
Privacy: If you are going to share personal data (datatilsynet.no, page in Norwegian) with AI, this must be stated in the research project that is registered and assessed and approved by SIKT. Informants must also be informed about the use (link to more information about this to come). If you have this in place, you can even share yellow data with SIKT KI-chat. Feel free to ask your privacy contact before using anonymized data, as separate rules may apply.

Tools with AI components

You can also find that traditional software for text analysis has been extended with an AI component as a separate module, such as NVivo and ATLAS.ti.

Tools where you do not share your data on the internet, but can operate locally (offline), are also more and more common. LM Studio is one such example. What these have in common is that they usually require a powerful workstation or server. The advantage is that you have more control over security.

Questions?

The University Library

Helps with the use of AI tools in various parts of the research process (assessment of usefulness, quality, functionality and so on):

R&D IT

For help with access, acquisition, risk assessment

R&D IT (e-mail) or
The web page "Request for a new IT service"