AI is changing the research world – What does that mean for open access?

What do developments in the area of artificial intelligence (AI) mean for open access and open research? (FUßNOTE - Definition: Concepts and practices of open and sustainable access to research results and materials or to the entire research process. Open research is closely related to open science as defined in the UNESCO Recommendation on Open Science. However, the term allows for the inclusion of practices in the fields of humanities and cultural studies.) In a series of blog posts, the open-access.network project wants to deal with recently emerging questions and discussions about the relationship between these thematic areas. The series begins with a blog post that provides a general introduction and contextualisation and opens up a space for questions from the community.

What's it all about?

Artificial intelligence (AI) refers to algorithms that are capable of imitating human thinking and learning. It includes machine learning (data identification and sorting based on repeated data inputs) and neural networks (data identification and sorting based on merging data from different databases). At present, AI is used among other things for the purpose of image recognition and creation, speech recognition, text generation and optimisation, pattern recognition, and process optimisation (see Fraunhofer Institute for Cognitive Systems, 2024).

While open access increases opportunities for educational equity through barrier-free access to scientific information, the use of AI also makes an important contribution to this development. For example, language barriers can be removed, thereby making information outside a person's own education system more accessible (see Kasneci, 2023).

Open access and AI tools in research

Many AI tools are used in science, too. But what points of contact does AI have with open access (i.e. barrier-free access to scientific information on the Internet)?Here, we can look on the one hand at AI tools that play a role in the publishing process, and on the other hand at tools that reduce access barriers in science. AI tools help in formulating texts or searching for suitable reviewers. In addition, they can be used to improve the accessibility of scholarly publications, for example by creating automatic translations, converting text to speech, describing images, or automatically generating subtitles for videos. In this way, people can get access to scientific information irrespective of specific requirements. Researchers who are not English native speakers can also benefit from AI because tools based on large language models (LLMs) help them to translate or generate their scientific texts.

Thanks to the aforementioned AI applications, scientists (supposedly) have more time for researching, planning, and implementing research projects or for teaching (see Staiman, 2023). While open access increases opportunities for educational equity through barrier-free access to scientific information, the use of AI also makes an important contribution to this development. For example, language barriers can be removed, thereby making information outside a person's own education system more accessible (see Kasneci, 2023).

However, at this point, two limitations on the use of AI should be noted, taking ChatGPT as an example: The use of that tool depends on the quality of the formulated prompts (instructions/queries). It is important to refine queries further, and that takes some practice. Moreover, the various versions of the software differ in terms of the underlying dataset and the accuracy of the outputs. Version 4.0 is based on a larger dataset, but the pro variant is subscription-based. There is thus a danger that the quality of the responses will depend on a person's level of income, and that a form of inequality will therefore be created.

Open access publications not only provide legal protection for authors when reusing content, they also do a lot more. For it is (still) difficult to write complex scientific works with the help of algorithms. For example, references provided by the AI application are sometimes "hallucinated" – that is, the sources cited by the AI application are sometimes non-existent, or the content generated is invalid (Alkaissi & McFarlane, 2023).

Copyright and AI

One much discussed aspect of AI is the question of the legal conditions under which publications may be used to train LLMs, and whether different regulations exist – or should exist – in the case of usage for commercial and economic purposes (see Brehm, 2022).
Regarding copyright, further regulations and stricter legislation are needed to reduce existing grey areas. According to one view, crawling data that are freely available on the Internet is permitted by virtue of the limitations on copyright for text and data mining (Sections 60d and 44b of the German Copyright Act [UrhG]; see Rack, 2024). In future, providers of AI services that are trained on large amounts of data will be required under EU law to disclose the data used for training purposes (see Baumann, 2024). It is standard practice that works published in open access are made available under a Creative Commons licence. Thus, the reuse of the content is regulated by the respective licence.
Open access publications not only provide legal protection for authors when reusing content, they also do a lot more. For it is (still) difficult to write complex scientific works with the help of algorithms. For example, references provided by the AI application are sometimes "hallucinated" – that is, the sources cited by the AI application are sometimes non-existent, or the content generated is invalid (Alkaissi & McFarlane, 2023). The availability of texts in open access plays a very important role in validating artificially generated texts and their sources because the references and cited texts and data are publicly accessible and verifiable. This helps to expose invalid research or "hallucinated" texts.

AI, and what now?

AI has developed rapidly in recent years and has become an integral part of many people's daily working lives. A future without AI is clearly very unlikely. Thus, the question is not whether but rather how we handle the new tools. International economic policy discussions  are being conducted on how to deal with the new developments (e.g. Manola, 2023).

Those who work in the areas of infrastructure (e.g. libraries) and teaching can contribute to improving information competence by offering event and information formats on the critical engagement with these tools (advantages and disadvantages, handling). ). Discourses should also be encouraged on the role of AI not only in producing and distributing but also in recognising "hallucinated" research results or text publications generated by AI within science communication.

Questions relating to quality assurance in research and the benefits and limits of AI for open research will remain relevant in our work.

Those who work in the areas of infrastructure (e.g. libraries) and teaching can contribute to improving information competence by offering event and information formats on the critical engagement with these tools (advantages and disadvantages, handling). Discourses should also be encouraged on the role of AI not only in producing and distributing but also in recognising "hallucinated" research results or text publications generated by AI within science communication.

Outlook: Blog post series

This blog post is the first in a series that will deal at irregular intervals with the topics of artificial intelligence and open access/open research. Critical engagement with the handling of ChatGPT and Co. and the questions that arise empowers people to find their own way of dealing with the developments.

Now it's your turn: Do you have feedback or suggestions for future blog posts on this topic? How do you handle AI? What potential and dangers do you see in relation to AI and open access/open research? You are welcome to use the comments function under this post for a constructive discussion!

1Definition: Concepts and practices of open and sustainable access to research results and materials or to the entire research process. Open research is closely related to open science as defined in the UNESCO Recommendation on Open Science. However, the term allows for the inclusion of practices in the fields of humanities and cultural studies.

References

This article is licensed under the Creative Commons Attribution 4.0 International Licence (CC BY 4.0).

Last updated on

More information on the topics of this page: