In this keynote lecture at the Symposium on Digital Transformation of Research (DiTraRe), I will address two questions. First, what fundamental epistemological differences exist between human research strategies and research shaped by algorithms. Second, what are consequences for the objective phrased in the Reflection and Resonance dimension of DiTraRe “to establish trust in research that is shaped by algorithms, [where] it is vital to ensure transparent, reproducible practices and clear communication of findings.”

To explore these questions, I will draw on previous philosophical work that emphasizes the crucial role of descriptions of phenomena and conceptual models in human research strategies (Boon, 2015; 2020a). In contrast, ideas about AI-based scientific research suggest that phenomena and models are redundant once large amounts of data are available (Boon, 2020b).

In this lecture, I will use examples from experimental and empirical (data generating) practices in the natural, engineering and social sciences to explain how humans generate and use data, descriptions of phenomena and conceptual models in their scientific reasoning to arrive at a (scientific) knowledge.

My objective is to provide a deeper (epistemological) understanding of scientific reasoning strategies that align with human cognition and to invite critical reflection by the audience on how this compares to AI-based research.

Mieke Boon bio

Mieke Boon is professor of philosophy of science in practice at the Department of Philosophy of the University of Twente. She holds a PhD in chemical engineering and biotechnology. In 2006, she established a new movement, dubbed the philosophy of science in practice. Her research in this field focuses on epistemology, with particular interest in how human reasoning (i.e., their reasoning strategies) enables the construction of epistemic results (theories, models, concepts and phenomena, laws) that meets stringent epistemic and pragmatic criteria. Topics she has been publishing on are scientific models, representation, scientific instruments, phenomena, paradigms of science, scientific methodology, and epistemological responsibility. Current topics include AI, interdisciplinarity, and science education.

References and recommended reading

Boon, M. (2020a). The role of disciplinary perspectives in an epistemology of scientific models. European journal for philosophy of science, 10(3), 31.

Boon, M. (2015). Contingency and inevitability in science-Instruments, interfaces and the independent world. In Science as it could have been: Discussing the contingent/inevitable aspects of scientific practices. L. Soler and E. Trizio (ed.).  University of Pittsburgh Press: 151-174. 

Boon, M. (2020b). How Scientists Are Brought Back into Science—The Error of Empiricism. In M. Bertolaso, & F. Sterpetti (Eds.), A Critical Reflection on Automated Science: Will Science Remain Human? (Vol. 1, pp. 43-65). (Human Perspectives in Health Sciences and Technology; Vol. 1). Springer. https://doi.org/10.1007/978-3-030-25001-0_4

Boon, M., & Van Baalen, S. (2019). Epistemology for interdisciplinary research–shifting philosophical paradigms of science. European journal for philosophy of science, 9(1), 16. https://doi.org/10.1007/s13194-018-0242-4

Giere, R. (1996) Understanding Scientific Reasoning Wadsworth Publishing

Kwa, C. (2011). Styles of knowing: A new history of science from ancient times to the present. University of Pittsburgh Press.

=================

Bailer-Jones, D. M. (2009). Scientific Models in Philosophy of Science. Pittsburgh: University of Pittsburgh Press.

Bogen J. and J. Woodward (1988). Saving the phenomena. The Philosophical Review, 97(2): 303-352.

Hacking I. (1983). Representing and Intervening: Introductory Topics in the Philosophy of Natural Science. Cambridge: Cambridge University Press.

Glymour, B. (2000). Data and Phenomena: A Distinctions Reconsidered. Erkenntnis, 52: 29-37.

Massimi M. (2008). Why There are No Ready-Made Phenomena: What Philosophers of Science Should Learn From Kant, Royal Institute of Philosophy Supplement 83(63): 1-35.

Massimi, M. (2011). From data to phenomena: A Kantian stance. Synthese, 182(1): 101-116-71.

McAllister J.W. (1997). Phenomena and patterns in data sets. Erkenntnis, 47, 217-228.

McAllister W. (2011). What do patterns in empirical data tell us about the structure of the world? Synthese, 182(1): 73-87. 

Woodward J.F. (2011). Data and phenomena: a restatement and defense. Synthese, 182(1): 165-179.

Universities need more than powerful AI models—they need a people-centered way to weave AI into every workflow. Drawing on the newly launched AI & Design Thinking Lab (AIDT-Lab) at CUC Communication University of China, Beijing, this talk outlines a “living-lab” approach: empathize with students and staff, prototype AI solutions in Design Thinking sprints, and treat the campus as a continuous testbed. By synchronizing policy drafts with software releases and embedding upskilling tracks from day one, we sketch a reproducible playbook for turning any institution into a resilient, AI-ready ecosystem.

The digital transformation blurs the boundaries between research 
activity and infrastructure in the humanities. Data‑driven methods and 
AI are gaining importance, yet their outputs hinge on the quality of 
underlying data and metadata. Initiatives such as the German National 
Research Data Infrastructure (NFDI) and the European Open Science 
Cloud (EOSC) foster federated data spaces that enable large‑scale, 
cross‑modal data retrieval and analysis. Potential gains include 
accelerated discovery of cultural artefacts, reproducible workflows and 
new insights through federated data corpora. Risks appear in data 
fragmentation, algorithmic bias, dependence on evolving platforms, and 
the need for sustainable digital publication and development of new skills.

Core competences of the humanities — close reading, archival work and 
hermeneutics — still remain essential. The Digital Humanities extend 
this repertoire with computational text processing and annotation, 
historical network analysis, computer vision, interactive visualisation 
and digital knowledge representation.

This impulse talk examines how these new methods integrate into everyday 
scholarship and teaching in the humanities, supported through federated 
infrastructures such as NFDI4Culture. It closes by asking how digital 
methods can strengthen, rather than replace, humanities research while 
safeguarding data quality, data sovereignty and long‑term availability 
of data and results.

This presentation will examine the critical components that drive AI progress, with a focus on the need for large-scale, machine-interpretable data that adapts to the unique demands of different sectors. Key topics will include transparency and reproducibility in AI, as well as the importance of making essential resources—publications, data, models, and code—accessible, interconnected, and scalable.

A robust data infrastructure is essential for the AI community, supporting every stage of the data lifecycle—from collection and creation to processing, analysis, publishing, archiving, and reuse of resources. This talk will highlight the foundational role of these infrastructures in AI-driven initiatives, showcasing several ongoing infrastructure initiatives and illustrating how they contribute to a modern data ecosystem.

TBA

As digital tools—from AI to big data platforms—reshape how research is conducted, traditional legal norms often lag behind rapid technological advances. This presentation explores how ethical principles can fill the gaps where legal clarity is lacking, and how the two frameworks can work together to guide responsible innovation.

Research in Chemical Informatics, in particular in Deep Learning, has long been hampered by a lack of free and open datasets, let alone large ones. In my statement, I will highlight examples for promising machine learning in chemistry and describe efforts to build a large corpora of chemistry data in the context of NFDI4Chem and elsewhere. 

Research e-infrastructures such as EUDAT, EGI, GEANT and OpenAIRE provide universal access to data and computing resources spanning multiple research domains. They operate IT infrastructures that serve as a model for the EOSC EU node, and they are closely connected to research needs via their membership organisations. Sharing services with, or offloading them to, the established e-infrastructures would reduce operational costs and complexity by, for example, enabling deduplication and standardisation. This would likely benefit research data management and accelerate the development of national data infrastructures.

During the coming decades we will see a massive evolution of large, generic and „small, specific“ AI models for all kinds of purposes which will especially be true when we have solved the data sovereignty challenge, i.e. when data and model providers can be sure to get the credits for what they are offering. To a large extent these models will cover human knowledge and heritage requesting a completely different attitude to data as we are used to. Persistence and clear identification will be a requirement to exactly know which model is meant and how it was generated. Transparency is of great relevance to know where data came from and who processed them. Accountability of all actors involved will become a requirement, since humans will be increasingly dependent on the correctness of models. Proven and lean data infrastructures with certified processes willneed to guarantee accountabiility. FAIRness in the sense of machine actionability will remain important. FAIR Digital Objects will be the basic building block in these infrastructures and dataspaces that have a chance to improve sovereignty, transparency, accountability and persistence. 

*Anne Krüger and Ingmar Mundt will give a joint presentation.

Since the release of ChatGPT, there is an ongoing debate about the consequences of generative AI models on the production of scientific knowledge. Currently, genAI is either discussed as a promise for enhancing productivity and efficiency or viewed as a potential threat to human expertise. Beyond such very general discussions, it becomes clear that the scientific field comprises highly diverse application contexts due to various disciplines which means that genAI can be integrated into the research process in completely different ways ranging from being used as a mere tool to being treated as a “critical friend”. However, it also becomes clear that there are increasingly gray areas where the use of genAI is possible but raises important questions about what truly constitutes science and where the boundaries of applicability lie. This applies to questions about the necessity of human creativity in scientific knowledge production or with regard to authenticity, i.e., the traceability and accountability of scientific results. Connected to this is also the question of the corresponding skills that researchers must possess to use genAI adequately. These include not only skills in applying the tools but also knowledge of how these tools actually function in order to gain insights into possible sources of error. In our contribution, we explore these three aspects of scientific knowledge production--creativity, authenticity, and skills--and discuss the extent to which these aspects are influenced by genAI.

*Anne Krüger and Ingmar Mundt will give a joint presentation.

Since the release of ChatGPT, there is an ongoing debate about the consequences of generative AI models on the production of scientific knowledge. Currently, genAI is either discussed as a promise for enhancing productivity and efficiency or viewed as a potential threat to human expertise. Beyond such very general discussions, it becomes clear that the scientific field comprises highly diverse application contexts due to various disciplines which means that genAI can be integrated into the research process in completely different ways ranging from being used as a mere tool to being treated as a “critical friend”. However, it also becomes clear that there are increasingly gray areas where the use of genAI is possible but raises important questions about what truly constitutes science and where the boundaries of applicability lie. This applies to questions about the necessity of human creativity in scientific knowledge production or with regard to authenticity, i.e., the traceability and accountability of scientific results. Connected to this is also the question of the corresponding skills that researchers must possess to use genAI adequately. These include not only skills in applying the tools but also knowledge of how these tools actually function in order to gain insights into possible sources of error. In our contribution, we explore these three aspects of scientific knowledge production-creativity, authenticity, and skills - and discuss the extent to which these aspects are influenced by genAI.

Digital transformation is fundamentally reshaping scientific publishing. Today, generative AI tools support nearly every stage of the publishing process – from discovering relevant content and assisting with text generation and translation, to automating quality checks and streamlining peer review.

How can these technologies help make scientific content more rapidly produced, accessible, and user-focused – thereby accelerating knowledge transfer and expanding the reach of research? What new challenges do they pose for integrity, transparency, and the role of human expertise? And how can generative AI be integrated into publishing workflows responsibly – as a tool to support, not replace, human judgment?

Recent technological breakthroughs in the field of generative AI (GenAI) have led to its rapid adoption in all areas of scientific research, including scientific policy advice at the intersection of science and politics. In the field of technology assessment, a special form of policy advice focusing on innovation and technology, GenAI is not only a subject of research, but has also become a tool in the daily practice. However, its use in scientific policy advice raises specific challenges.It can affect the scientific credibility of the advice, its legitimacy and its relevance in positive, but also in negative ways. At the Office of Technology Assessment at the German Bundestag (TAB), we have been working intensively on these issues for the past two years. On the one hand, we have investigated and reflected on the use cases for which generative AI is expected to have the greatest value. On the other hand, we have developed living guidelines to ensure the responsible use of generative AI in scientific advice to the Parliament.

In this presentation, we will firstly provide a brief summary of the literature on the potential and challenges of digital science and generative AI for scientific policy advice. Secondly, we will present the central use cases for generative AI in scientific policy advice that TAB has focused on, from identifying emerging issues, exploring the state of knowledge on issues of interest to the parliament to ensuring the quality of the advice and enhancing target-group-specific communication. We will then describe the central challenges and tradeoffs that we have faced, in terms of skills and capacities, but also tradeoffs between efficiency and ethical considerations. After that, we will discuss how the use of generative AI is changing advisory practices. Finally, we will reflect on the specificities of parliamentary advice and whether the findings may apply to other situations of scientific policy advice.

*No abstract is required, as Dana will join the session to provide comments.