by Sofia Baroncini, Constanze Buyken, Judit Garzón Rodríguez, Ian Kisil Marino, Sarah Oberbichler, Cindarella Petz
The 2025 Digital Humanities Conference, held this year in Lisbon under the theme “Building Access and Accessibility: Open Science to All Citizens”, brought together a global community of researchers, developers, and practitioners to reflect on the role of openness, inclusivity, and ethics in digital scholarship. Hosted in the vibrant academic and cultural setting of Lisbon, the conference offered a rich program of keynotes, panels, workshops, and poster sessions, engaging with pressing questions around public participation, digital infrastructure, and the ethical responsibilities of working with data and digital tools in the humanities.
The conference theme built on the legacy of the Budapest Open Access Initiative (2002) and extended the conversation to contemporary challenges and opportunities surrounding Open Science. Contributions addressed a wide range of topics—from citizen humanities and inclusive digital platforms to multilingual practices, the ethics of AI, and FAIR/CARE principles—demonstrating the many ways in which the Digital Humanities community is actively shaping more equitable and transparent research practices.Researchers of our DH Lab participated in the event with several contributions that touched on different facets of this year’s theme. Two posters presented our efforts in promoting critical data literacy and inclusive infrastructures within the humanities. The poster on “Innovative Pathways to Data Literacy: Tailored Formats for Humanities and Cultural Studies” presented by Judit Garzón Rodríguez and her colleagues from the HERMES project, focused on two formats developed in this project: the Bring Your Own Data Lab (BYODL) and the Data Carpentries, designed to support researchers in the humanities and cultural studies in acquiring practical data skills through context-sensitive and collaborative learning. The second poster was developed jointly by members of HERMES and the NFDI4Memory consortium and presented by Constanze Buyken and Judit Garzón Rodríguez. Showcasing the HERMES research studies program and the NFDI4Memory FAIR Data Fellowships, the poster entitled “Small Grants, Big Opportunities: Enabling Inclusivity and Innovation in Digital Humanities” explored how small-scale funding schemes can foster experimentation, interdisciplinarity, and inclusion in DH contexts.
In addition, several researchers of the Lab contributed with conference papers. In her two presentations Sofia Baroncini addressed different dimensions of cultural heritage: one presented an approach to improving access to the interplay between material and immaterial cultural heritage through semantic modeling, while the other analyzed symbolic associations in the arts using linked open data. With their talk on “LLMs as Analysis Tool” Sarah Oberbichler and Cindarella Petz introduced a framework for applying large language models in Digital Humanities research, focusing not only on their potential but also on the importance of evaluation criteria and critical assessment when using these tools. Finally, Ian Kisil Marino participated in a collaborative panel entitled “The Global State of Digital History: Establishing Data Culture(s) in Uncertain Times”, which brought together scholars from different institutions to discuss how digital history is being shaped by regional and institutional contexts, and how data cultures can emerge in complex or unstable environments.
Ethical considerations were a common thread running through all of these contributions —whether in relation to equitable access to digital infrastructures, responsible use of AI, or the design of inclusive learning and funding formats. In the sections that follow, we share individual reflections on selected talks, workshops, and panels we attended and resonated with our work, especially those that raised important questions about ethical dimensions.

Data Ethics at the DH2025: Patterns and Trends
With 117 long papers, 218 short papers, 133 posters, 10 panels, and 17 workshops, the DH2025 conference brought together a large amount of researchers from all over the world presenting new research insights, proposing new methods and fostering discussions. Looking at the abstracts of those contributions, we can see that a considerable part of conference was dealing with questions around data ethics. In order to get a broader insight on the conferences pattern and trends in data ethics, we conducted a quantitative analysis of the contributions relevant to the topic.
We first identified ethics-related themes (e.g., AI and ethics, data sovereignty) and used an LLM (Qwen 3, 235B)[1] hosted by the Gesellschaft für wissenschaftliche Datenverarbeitung mbH Göttingen to parse the conference program into a structured JSON format (data available here). Next, the same model was employed to auto-generate (ethics-focused) keywords for each contribution and add those to the JSON file, which were validated and refined manually. Finally, we curated a subset of 67 contributions (presentations, posters, workshops) based on thematic relevance, which were analyzed and visualized to map ethical discourses at DH2025. While we may have inadvertently overlooked a few contributions, the overall pattern remains clear:

the conference’s overall ethics representation as seen in Figure 2 indicates that data ethics has become a priority within digital humanities. Themes such as AI ethics, bias analysis, inclusive access, community governance, the FAIR and CARE principles, data sovereignty, and the ethics of open access have been part of many presentations and discussions at the conference. In this context, inclusive access and AI ethics as well as FAIR/CARE principles emerged as the conference’s primary ethical concerns. Also critical frameworks where an important topic. AI ethics addresses the responsible deployment of AI in humanities research, while inclusive access focuses on removing barriers to ensure equitable participation in digital scholarship. FAIR/CARE promotes findable, accessible, interoperable, and reusable data practices alongside community-controlled approaches to Indigenous data governance, while critical frameworks provided theoretical foundations for examining power structures and biases in digital humanities work.
The analysis also demonstrates how different contribution formats engage with ethics themes. Panels showed the highest integration rate at 70% (7 out of 10 panels), followed by workshops at 23.5% (4 out of 17 workshops), while short papers maintained a 13.4% ethics focus (29 out of 218 short papers) and long papers showed 12.7% engagement (15 out of 117 long papers).

The geographic distribution of ethics-focused contributions at DH2025 further reveals important patterns, though the collaborative nature of digital humanities means that a considerable part of the contributions involves cross-regional partnerships. Researchers from Europe and North America feature 83.6% of all 67 ethics contributions, whether as sole contributors or in collaboration with other regions. This reflects their larger overall presence at the conference. When examining regional engagement with ethics, the proportional analysis becomes more nuanced. Contributions involving African institutions show the highest focus on ethics, followed by those with South American and North American participation, while proportional, ethics themes were least present in Asian and European contributions.
Although a considerable number of our DH Lab members participated in the conference, the extensive program with multiple parallel sessions made it impossible to attend every session. We therefore provide an overview of the personal impressions we gained from the talks we did attend, without claiming that they constitute a complete review of the themes discussed.
After exchanging our thoughts over some refreshing Portuguese beverages, we identified the subtopics related to data ethics among the talks we attended and which we decided to explore further in this blog post.
AI and Ethics
The DH2025 conference highlighted urgent debates on AI ethics in the digital humanities, focusing on mitigating bias, ensuring responsible AI use in historical and cultural research, and centering marginalized voices through frameworks like FAIR/CARE principles and data sovereignty. Discussions emphasized the need to prioritize interdisciplinary collaboration, transparency, and community-led governance to address systemic inequities and ethical challenges in AI-driven scholarship. However, we observed a clear discrepancy between theoretical discourse and practical implementation: while many panels and also workshops emphasized themes like responsible AI usage, data literacy, sustainability, and thoughtful model selection, quite some research papers showcased analyses relying heavily on ChatGPT as their primary tool. This preference for accessibility and convenience over critical evaluation of models often overshadowed considerations of local, more transparent alternatives. Privacy concerns were largely overlooked, and the uncritical framing or examination of ChatGPT as a “historian” in some presentations underscored a troubling lack of scrutiny. Additionally, proposed frameworks for AI literacy, although well-intentioned, frequently remained abstract, offering conceptual ideas without practical implementations. In conclusion, the conference ultimately highlighted an urgent need to transition from dialogue to deliberate, responsible application of AI in practice. Encouragingly, a handful of presentations demonstrated promising steps toward this goal, proving that progress is possible when ethical considerations are paired with tangible strategies.
FAIR Principles
The FAIR principles, along with open data guidelines, were an overarching theme of multiple sessions, integrated in projects with different degrees of depth. Whereas the community agrees on the need to adopt these widely acknowledged principles, the discussions revealed the complexity that arises when it comes to properly applying them. In recent years, much research has been conducted on the topic. Also, frameworks and tools designed to integrate these principles in research or to evaluate the FAIRness of data have been proposed. Nevertheless, the evaluation tools seem to provide different assessments depending on the criteria used, and the current frameworks are not widely and thoroughly applied. Indeed, the level of detail reached by the available support tools requires careful selection and study to avoid a biased evaluation or a simplistic application of the principles. Furthermore, the need for a thorough terminological literacy has emerged, as terms such as “open” and “accessible” were used indiscriminately, despite the research field providing clear definitions: while “open” refers to providing the data with a clear and open license, which allows its non-commercial reuse (e.g., through Creative Commons licenses), the FAIR principle of “accessibility” refers to the possibility to reach and access such data made available online. Both accessibility and openness are crucial for researchers. However, as licensing becomes increasingly urgent to protect data from being scraped for AI model training, we may need to step back from complete openness when it comes to research data reuse. Also, the availability factor needs some critical reflection. Not all sources, especially from under-represented communities, have the community approval to be openly shared, which – as one of the papers at the conference highlighted – raises the question of “slow research” as a possible valuable contribution.
Discriminatory language in databases, under-representation of communities, data sovereignty
Several sessions of the conference delved into a discourse concerning critical cataloging, specifically the approach of considering metadata and data modeling structures as expressions of power dynamics and specific viewpoints possibly subject to discriminatory practices. Handling data in research means dealing with bias that may occur in every stage of the data production life cycle, as it can affect the sources, the cataloging metadata, or, in the case of using AI, the algorithms selected. Indeed, many sources from the past contain racist and sensitive terms, which are in contrast with the current ethical principles. The problem arises especially when data are made publicly available, with an urge to find solutions. Among the strategies adopted, some projects propose to hide discriminatory terms at first glance, and to provide the full expression only upon an explicit request of the user. This approach ensures that access to historical information is maintained, without promoting or publicizing discriminatory content that could affect users who come across it unexpectedly. Further solutions to support the catalogers and their reparatory activity leverage the use of ontologies and Knowledge Graphs that represent the guidelines provided by museums on the topic. Nevertheless, as all produced data reflect a partial perspective, it is not always possible to mitigate bias. Another solution proposed at the conference is the identification and annotation of bias, according to a defined vocabulary, in order to make it explicit and allow further analysis of the bias.
Further talks addressed the issue of doing (digital) history when the only known or consulted sources come from former colonial states, or, generally, from archives held in the so-called “Global North”. In this case, a practicable solution is more challenging if no further sources are available. Therefore, the main approach lies in making the bias explicit. Some projects estimated the representation of communities in art databases, and considered which perspective such cataloguing practices reflected. They also explored how to enrich datasets to balance the under-representation of communities in data (e.g., through ingesting data from further sources). One good example that we could perceive at the conference relates to the CARE principles, developed for promoting indigenous data governance, sovereignty in data management, and accessibility to the indigenous community to which the data relates. In particular, the theme of accessibility to the target community was discussed, to underline that the simple availability of data online does not imply that the data is effectively accessible to the community. Furthermore, the challenges for implementing and practicing decolonized terminology and viewpoints in museums remain significant, especially in the case of European and North American museums. This proves the need for further collaborations of museums with local communities to reach concrete, respectful solutions. Positively, many talks on the topic were led by speakers from Asia, Africa, South America, and the Caribbean, showing a possible path of fruitful collaboration and a possible future of concrete data sovereignty for the so-called “global peripheries”.
Despite the positive practices and solutions proposed, the discourse remains challenging, given that labeling a community continues to be a possibly discriminatory practice. The proposal of technical solutions needs to be grounded in a critical humanistic discourse, involving the voices of scholars and communities from the so-called “Global South”, and capable of treating human local knowledge in a sensible manner.
Our key takeaways
The presence of ethics at the DH 2025 Conference was sound. As our brief report shows, there are many shades of ethics being discussed in DH scholarship worldwide. As one presenter remarked during a workshop, ethics is more a topographic than an ontological question: it is a complex dilemma that cannot have a universal answer but requires adaptation, care, and genuine attention. From our perspective as historians, this makes sense: we are used to dealing with multiple pasts that we constantly create and recreate depending on our social context, methods, theoretical paradigms, archives, and sources. These different pasts, continuously debated by diverse people in various contexts, from North to South, and from East to West, represent different ethics. Perhaps the most significant takeaway from the so plural presence of ethics at DH 2025, in accordance with Stefan Berger, is that ethics in history can encompass many things, but it is certainly not a monologic issue [2].
Another important takeaway is that we are still far from implementing AI in an ethical way. AI ethics involves ethical tool transparency and (self-) critique, as well as awareness of how these tools function and shape our work. But the lack of AI tools that are ethical, accessible, and of good quality lead many researchers to the use of proprietary platforms owned by a handful of Big Tech billionaires, which highlights a broader ethico-political turn – bringing issues of colonialism, imperialism, racism, and sexism into focus within a technological landscape marked by inequality, uncertainty, and ubiquity. In addition, or as a reaction to those issues, we also believe that future discussions need to move beyond critical assessment of computational tools and Gen AI functionalities. Instead, we should ask: how can we, as DH researchers, actively shape the development of AI to be more ethical and better aligned with humanities values? Rather than remaining passive users of proprietary platforms, the DH community has the expertise and responsibility to contribute to creating AI systems that embody the critical, contextual, and research-centered approaches that define our field.
One important aspect of the humanities’ involvement in AI developments is the focus on knowledge graphs (with a reference to neuro-symbolic AI). Knowledge graphs and ontologies will enhance AI while they can also help in mitigating bias, through 1) ingesting more information about under-represented groups, 2) gathering domain information about discrimination and bias that can support practices of critical cataloging, 3) in defining taxonomies of bias, to make it explicit when it is not possible to mitigate it. The importance of neuro-symbolic AI and knowledge graphs is also evident in the way they are foregrounded as a relevant topic at the upcoming TPDL 2025 conference, where two IEG researchers (Sarah Oberbichler and Johanna Mauerman), are presenting their co-authored paper on model design bias evaluation.
In conclusion, with the pocket full of challenges, but also paths for solutions, we are especially looking forward to continue the conversation in the upcoming Data Ethics for Historical Research conference, organized by the IEG in collaboration with the Academy of Science and Literature, and supported by NFDI4Memory, in which the authors of this post will have an active role.
References:
[1] Accessed via the ChatAI interface of GWDG, https://www.uni-goettingen.de/de/686446.html.
[2] Berger, Stefan. “History Making and Ethics—an Integral Relationship?” History and Theory 62, no. 1 (2023): 161–73. https://doi.org/10.1111/hith.12294
Please cite as follows:
S. Baroncini, C. Buyken, J. Garzón Rodríguez, I. K. Marino, S. Oberbichler and C. Petz (01.08.2025): From Access to Ethics. Data Discussions at the DH2025 Conference in Lisbon. DH Lab https://dhlab.hypotheses.org/7355.
The suggested citation reference below is generated automatically and does not take multi-authorship into account.
Featured Image: Portuguese azulejos, from the Museu Nacional do Azulejo in Lisbon (picture taken by S. Baroncini; the artwork itself is in public domain).
OpenEdition schlägt Ihnen vor, diesen Beitrag wie folgt zu zitieren:
Sofia Baroncini (1. August 2025). From Access to Ethics: Data Discussions at the DH2025 Conference in Lisbon. DH Lab. Abgerufen am 23. Januar 2026 von https://doi.org/10.58079/14gdf