AI, texts, and the future of knowledge work

From text generation in academic settings to the ethical challenges of using artificial intelligence in the publishing industry: Professor Christoph Bläsi takes a comprehensive look at AI in book studies – and far beyond. Since 2019, he has been one of three spokespersons for the KI@JGU network.

The world of publishing and artificial intelligence (AI) is a highly dynamic field, with new questions and debates constantly emerging. Christoph Bläsi, Professor of Book Studies at Johannes Gutenberg University Mainz (JGU), is well acquainted with these developments. He has been one of the coordinators of the interdisciplinary KI@JGU network since 2019, alongside his JGU colleagues Petra Ahrweiler, Professor of Sociology, and Stefan Kramer, Professor of Computer Science. The network was initially established in the mid-2010s to identify who at JGU was already using artificial intelligence, in what forms, for what purposes, and across which disciplines.

According to Bläsi, the number of AI projects at JGU was still relatively small when the network was created. "We could count the initial responses on the fingers of two hands." That has since changed enormously. "Over time, more and more institutes have engaged with AI, and we’ve seen the planning and funding of numerous large-scale projects," he recalls. Bläsi points to TOPML as a prominent example of this development. TOPML is a theoretical computer science project sponsored by the Carl Zeiss Foundation that examines so-called non-functional requirements for AI systems, such as transparency and fairness. From the outset, TOPML has actively sought interdisciplinary exchange and collaboration with humanities research on AI.

Preparing for a professional future likely to be heavily shaped by AI

"In the publishing market, AI is often viewed quite broadly as a knowledge tool that helps shape the future of digital publishing in a wider sense," Bläsi explains. Indeed, the publishing industry has long been employing artificial intelligence in many different areas, which has resulted in genuine gains in productivity across the sector. At the same time, however, a critical reflection on the opportunities and limitations of AI and its applications is essential. "It remains a very timely question which tools of artificial intelligence enable people in knowledge work to do exactly what," the book studies scholar adds.

At the same time, Bläsi sees significant potential for the use of AI in humanities and social sciences research, not least in the field of digital humanities, which applies various computational methods and tools, including AI. In book studies research, for example, artificial intelligence is used to analyze large datasets, such as those generated by examining the contents of historical school textbooks. With the help of AI, this kind of research can explore how moral values have changed over time and how these shifts have manifested in school textbooks as a very specific form of publication.

Bläsi advocates a constructive approach to the rapidly evolving technology. For him, this means that "while exercising due caution, we should not focus solely on the risks of AI – like a rabbit frozen in front of a snake." Instead, he argues, we should look for meaningful applications and put them to use in our own work. "Rather than just placing excessive restrictions on the use of AI out of fear of data extraction and misuse associated with large-scale models, most of which are operated by U.S.-based companies, we should also be looking more closely at the opportunities offered by smaller, possibly European pre-trained systems that can be run on one's own hardware," Bläsi says.

He considers the teaching of AI-related competencies particularly important, especially in the field of book studies, because large language models operate on texts – and texts are the fundamental working material of the book industry and book culture. "After all, higher education should serve to prepare young people for a professional future that will, in all likelihood, be heavily shaped by AI," Bläsi emphasizes, pointing to the growing relevance of AI in the media sector. In the publishing industry, AI tools are already being used in many ways, from text generation, 'gatekeeping' and editing to support for strategic decision-making about publishing programs and even pricing.

From Computational Linguistics and Digital Publishing to Book Studies

Born in 1960, Christoph Bläsi brings a decisive advantage to his work, rooted in his background. He initially trained as a computer programmer and system analyst before enrolling in German Studies and Mathematics with a focus on Computer Science in the teacher-training program at the University of Freiburg in 1982. While still a student and after completing his degree, he spent periods abroad studying computational linguistics at the University of Sussex in Brighton and the Istituto di Linguistica Computazionale in Pisa. Digital publishing and information management have therefore been central to his profile for many years, including through his professional work with major German publishing houses such as Brockhaus-Duden, C. H. Beck, and others. During this period, he obtained his doctorate at Heidelberg University in 1999. In 2004, Bläsi was appointed Professor of Book Studies at Friedrich-Alexander-Universität Erlangen-Nürnberg before moving to Mainz University in 2009.

This combination of research and practice continues to shape Bläsi's teaching. In the winter semester of 2023/2024, he and his students produced a concise English-language professional handbook on entering a publishing industry in transition. The book – just over 120 pages in length when combined with its machine-translated German version – was produced largely through automated processes, drawing on state-of-the-art AI technologies as well as the editorial expertise of Springer Nature, an academic publisher active worldwide and based in Wiesbaden and Heidelberg. The book, titled Young Professionals in Publishing, is available from Springer Nature as both a digital product and a print-on-demand edition. A key takeaway for the authors is that although AI – in this case, the second evolutionary stage of ChatGPT – can assist with conceptualization, text production, and translation, writing a book that meets one's own expectations remains a very demanding task. "Proofreading and editing AI-generated text, in particular, are often what can quite rightly be described as a real pain for those involved in the writing process. It was hard to make something enlightening and readable out of the material initially produced by AI," Bläsi says.

Bläsi expects artificial intelligence to have significant implications for academic curricula and assessment formats. Given the capabilities of generative AI, for example, the traditional term paper is likely to lose relevance very quickly. "Sending our students off into the semester break with the task of writing a 20-page paper on a particular topic is no longer a viable form of academic assessment," Bläsi is convinced. Oral examinations may again become more relevant, as may supervised, in-person exams that require focused writing without internet access. "There is an urgent need for substantial changes to curricula and examination regulations in the humanities," he summarizes.

Artificial intelligence and the humanities

For research in the humanities, AI has a particularly immediate relevance, as there are substantial overlaps with issues that affect the everyday lives of many people also outside academia. In this context, Bläsi points to the role of AI systems as communication partners, an aspect that is becoming increasingly important for many users of digital services. In book studies, additional questions arise that are important particularly to authors, such as protecting one's own creative output and engaging with the different ethical requirements of publishing in the age of AI.

Given that more than 80,000 new titles are published in Germany every year, a further increase in productivity in the publishing sector is not necessarily desirable. Instead, AI may be used to make viable those book products that have long seemed attractive but were largely abandoned due to their time-consuming production. This applies in particular to enhanced books and book apps, which – beyond a few promising prototypes – have rarely been realized. Supported by AI, such formats can deliver more informative and immersive narratives by combining text and images with audio, video, and interactive elements.

In conclusion, Professor Christoph Bläsi advises keeping a close eye on the ongoing development of AI systems; this is not least true for book studies students. This does not require in-depth technical engagement with the latest iterations of large language models. "It is already a good step to regularly try out major new models and systems and, where possible, integrate them into one's own work," Bläsi recommends. Doing so helps keep the door open to this important and impactful field of development.