Data may be misleadingly considered an objective and impartial product. When doing historical research based on data (re)use, applying an ethical perspective becomes essential. Data is neither neutral or objective, nor are our practices for accessing, analyzing, and documenting it. Any discourse on the production of knowledge or on technology that creates and/or manipulates knowledge implies an ethical discourse and (un)conscious decisions. Ethics imply responsibility, as ignoring ethical issues or not taking action means, in many cases, perpetuating the existing biases and inequalities. The need for scholarly exchange in order to navigate the multifaceted and complex dimensions of research ethics was also reflected in the more than 200 registrations for the conference “Data Ethics for Historical Research in a Digital Era”, about which we would like to share our thoughts in this blog post.1
When discussing data ethics, three dimensions can be identified, which are equally important towards ethical and responsible research. First, the ethics of data, concerning the provenance, curation, and sharing of data; second, the ethics of algorithms, concerning the re-use of data in programming, machine learning, and artificial intelligence; and third, the ethics of practices, raising questions about data sovereignty and the role of actors in data science.2 Above all, the peculiarities of data historical research add another layer of complexity to the already challenging theme. Sources were created in past contexts from communities with a different sensitivity than what we have today, if not from unequal systems, which perspective is still visible nowadays in the languages describing and cataloguing such objects, and in the chain of selections that were performed throughout the past, determining what to retain and what to forget in the path of history.

During the conference3, such aspects were discussed by the participants, who, by simply taking a step out of their research and raising crucial questions, recognized their responsibility as actors who curate, interpret and propagate possible biases and inequalities present in the source data. The sessions of the conference were designed in a way that they reflected the connections among the contributions. These sessions were deeply interwoven. The issues of privacy, consent, community agency, bias, the power of language as a knowledge-shaping instrument, as well as documentation as best practice to re-contextualize data, metadata, and the tools we use in our research – all of these came up in the contributions. But one of the core topics that was present in all the talks and the discussions was responsibility. We thus would like to center our reflections about the exchange we participated in during this conference around this topic.
Responsibility when working with data from colonial contexts
The first session confronted how colonial legacies are embedded not only in the context of historical records but also in the very structures of data, metadata, and institutional practices. Considering the global aspect of both historical practice and digital transformation, the session addressed the difficulties of dealing with present-day inequalities, which are heavily marked by traumatic pasts that recall colonial dynamics of power. Such challenges include, for instance, the practical need to normalize data and metadata properly and the question of who holds the power to make such choices. This challenge is especially significant when dealing with data regarding global peripheries and colonial heritage.
As we learned from the presenters, it is fundamental that one doesn’t impose such categories but puts effort into working collaboratively and listening to communities instead. In order to avoid recreating asymmetrical dynamics in producing historical knowledge, one should be open to rethinking dataset designs and standards and not let them impose preformatted manners of representing and narrating sensitive pasts. Even so, the session invited reflection on how this very fundamental practice can itself be tricky, as one should avoid paternalistic research practices. In other words, responsibility should be taken cautiously, because it can lead to reproducing colonial power dynamics – even behind ethics-oriented discourses. In this sense, the main takeaway from this session is about the voices that speak behind data – and about our capability to truly listen to them, avoiding both the imposition of our own standards and the presumption that we have all the answers to what it means to be ethical when facing the weight of colonialism.
Responsibility in dealing with sensitive data and discriminatory language
This lead directly to the second session, which dealt with ethically and legally sensitive data: from instances of discriminatory language in our sources and issues of perpetuation when transformed to digital formats or when legacy data is re-used. These raised the question: How do we balance scholarly inquiry and the paradigm of open data with respect for privacy, dignity, and community agency?
While some measures are undoubtedly necessary as to mitigate instances of discriminatory language, harmful stereotypes, continued exploitation, and other forms of explicit biases, this session revealed also delicate balances and ambivalences, e.g., how to securely identify our own implicit or internalized bias in our data. The question of how to balance data protection and transparency, was one of the key points of this session. On the one hand, we should strive to lay open our (ultimately always biased) selection decisions: in the transformation of our data to digital formats, and our decisions for further processing within all stages of the research cycle. On the other hand, we need to adhere to the protection of our sources and data: we do have strict legal responsibilities, following general legal frameworks and specific consent-oriented re-use allowances for our data. Sometimes, and especially in the context of sensitive data, these legal dimensions and the quest for full transparency do clash, and decisions for e.g., anonymization can have far reaching implications for the re-usability and linking of research data, and might be even in contrast to what time witnesses would like to have their testimony presented. Reaching out to these witnesses, stakeholders, and communities has been continuously advised. This ties back to the learnings from the first session: the voices behind our data need to be heard.
Responsibility in using AI on historical data and its (re)presentations
Thirdly, as AI reshapes how we access and interpret history, the following session asked: Who controls the narrative? How can we reflect on biases that are encoded in algorithms and models, and how we can use AI ethically? By examining AI implementations in archives and museums, this section not only showed how frequent AI is already integrated in cultural heritage institutions, but also revealed that AI isn’t simply a neutral tool to ease access to cultural heritage. Most of the time, heritage institutions and visitors alike are not fully aware of how complex AI systems are. Moreover, without contextualization, digital reconstructions and archival analyses based on AI models risk being perceived as definitive truths rather than interpretations.
Museums and archives bear significant responsibility when implementing AI to provide access to cultural heritage. Yet they often rely on general-purpose language models, the only readily available option so far, which are poorly suited to the sector’s needs. These models struggle with the complexity and nuance inherent in cultural heritage. Furthermore, when such models are designed to answer broad, general questions but can only access a single institution’s resources, they inherently create or amplify bias. The sector so far lacks clear guidance on appropriate use cases for AI in cultural heritage and best practices for responsible implementation. Domain-specific systems tailored to the cultural heritage sector have yet to be developed, which would promise more reliable results and would be considerably more sustainable.
Despite these gaps, cultural heritage institutions are already deploying AI, as this section has shown impressively, mainly to ease or democratize access (e.g., by providing chatbots). This creates a fundamental tension: the more accessible and user-friendly the technology becomes, the harder it is to ensure users understand the models’ limitations, bias and interpretive nature. Cultural heritage institutions or museums would need to balance accessibility and ethical considerations, which especially includes transparency and documentation. However, safe, trustworthy, and ethical implementation of AI needs resources and expertise, which institutions often don’t have.
Responsibility is also required from users/visitors, who usually trust museums and archives to provide reliable access to cultural heritage. This trust cannot be expanded to AI tools, which introduce new forms of fragility, bias and potential manipulation. AI literacy therefore needs to become an essential skill in a world where AI increasingly mediates access to information and cultural heritage.
Responsibility for data documentation
Ethical practice begins with documentation. That’s why the fourth session was a hands-on workshop guiding through concepts and tools to contextualize datasets, such as the “data envelopes”. It became clear that documentation is potentially unlimited and that determining a sufficient amount and selecting the right type of information is one of the main tasks of an ethical research practice. Since documentation standards and best practices are still being developed and discussed, examples from data sets are a very valuable resource for guidance on how to document your data.
Within our discussions different actors were identified, such as data creators, collection providers and users of collections or datasets. Due to their different resources, capabilities, and responsibilities, the assessment of an ethical research documentation varies for each of them. Together, however, they share different parts of a common overall responsibility. Data creators have exclusive insight into the collection processes and internal data structures, while providers of collections and datasets are able to enrich metadata according to standardized principles and add context and provenance information. Data users must then critically evaluate the existing documentation and provide feedback to providers if documentation is to be practiced collectively. Responsibility for data documentation is closely linked to the respective role in the data lifecycle and first of all serves to make the decisions made within this process transparent.
Responsibility when utilizing data from archives
The fifth and final session turned to archives and archiving, not as static repositories, but as dynamic, contested, and living practices. In today’s digital research landscape, accessing archival materials has become easier, but not its assessment. This became evident when we looked at web archives. As un-curated collections, they are unsystematic snapshots of the public web, captured by automated crawlers in unpredictable ways. They resemble digital ghost towns – filled with unintended, unconsented, and unaccounted content. This raises urgent ethical concerns, especially regarding data subjects. Most individuals who shared content online never expected it to be preserved for decades. Identifying or contacting them is often impossible, making traditional consent unworkable.
The use of archival data is not neutral—it’s an ethical ecosystem where responsibility is shared among archivists, researchers, and technologists. This distributed model shifts the focus from abstract ideals of “accessibility” to the real-world challenges users face. This leads to responsibilities for archivists and researchers: archives and libraries are guardians of collective memory and are responsible for a well-considered collection policy that strikes a balance between preservation and data protection, copyright and confidentiality—especially for digitally created materials. At the same time, we as researchers are not passive users. We must behave ethically: understand the context, respect privacy, and collaborate with archivists. Our work shapes the interpretation of history. For everyone, AI and computer-assisted tools offer powerful opportunities for analyzing large data sets—but also carry the risk of distorting historical representations. As Lise Jaillant warns, “the risk is to bias the historical document and consequently history as well as our collective memory.”4 Archivists and historians must be involved in the development of AI to ensure transparency and accountability.
Ethical use of data is an ongoing negotiation. We need a shared language, empathy, and collaboration among digital humanists, archivists, and researchers. The goal is not to reach perfection, but to follow a more responsible and inclusive approach to memory and access. Ultimately, digital history must be more than technical skill – it must be a practice of ethical responsibility. We must recognize power dynamics, demand institutional accountability, and build solutions collaboratively. In this shared responsibility, no one acts alone. Together, we preserve not just data, but the integrity of our collective past.
Key aspects of responsibility in ethical data handling in the historical domain
First, the issue of ownership of data, concerning the data’s provenance but also what consequences follow when the data is published and publicly available: To whom does the data we are working with in our research “belong”? For whom and about whom are we doing our research?
Second, it is essential to think about the context in which we work and to verbalize the situated-ness of ourselves as researchers, of our knowledge and of our research questions (“situated knowledges”5 ): Who is working on/with the data and why? What are research questions/context in which the study is done?
Last but not least, we want to highlight the aspect of community. This means not only to get in touch with the community related to the research topic or research data, if possible, to ask for their opinions, thoughts and concerns (this may also apply to their descendants), but also the responsibility of our community as researchers.
In the conference, we saw that every use case has its particular features and challenges. So maybe there is not one best practice for applying an overall fitting, general framework for Data Ethics. But still, we all have many of good practices or examples, so getting in touch with each other is more than important: We must acknowledge that Data Ethics are not something which is “nice to have” but are crucial for many aspects of research and data lifecycles.
Featured Image: Conference Panel minutes before the start.
- Author contribution according to CReditT: Writing – original draft: Sofia Baroncini, Constanze Buyken, Fabian Cremer, Ian Kisil Marino, Sarah Oberbichler, Cindarella Petz, Elena Suárez Cronauer, Thorsten Wübbena. [↩]
- Floridi, Luciano, and Mariarosaria Taddeo. “What Is Data Ethics?” Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences 374, no. 2083 (2016): 20160360. https://doi.org/10.1098/rsta.2016.0360. [↩]
- “Data Ethics For Historical Research in a Digital Era. Critical Reflections and Best Practices.” Conference. Organized by Sofia Baroncini, Constanze Buyken, and Elena Suárez Cronauer. Leibniz Institute of European History (IEG), Research group Digitality in Historical Research: Methods and Research Data | DH Lab, in cooperation with the Academy of Sciences and Literature, Mainz and NFDI4Memory. Supported by the German Research Foundation (DFG). Held in Mainz, Germany, November 10-12 2025. https://ieg-dhr.github.io/Data-Ethics/. [↩]
- Lise Jaillant, ed. Archives, Access and Artificial Intelligence: Working with Born-Digital and Digitized Archival Collections. 1st ed. Vol. 2. Digital Humanities Research. Bielefeld University Press / transcript Verlag, 2022. https://doi.org/10.14361/9783839455845, p. 23. [↩]
- Donna Haraway. “Situated Knowledges: The Science Question in Feminism and the Privilege of Partial Perspective.” Feminist Studies 14, no. 3 (1988): 575. https://doi.org/10.2307/3178066. [↩]
Nur der Text ist unter der Lizenz Creative Commons Namensnennung 4.0 International nutzbar. Alle anderen Elemente (Abbildungen, importierte Anhänge) sind „Alle Rechte vorbehalten“, sofern nicht anders angegeben.
OpenEdition schlägt Ihnen vor, diesen Beitrag wie folgt zu zitieren:
Sofia Baroncini, Constanze Buyken, Fabian Cremer, Ian Marino, Sarah Oberbichler, Cindarella Petz, Elena Suárez Cronauer, Thorsten Wübbena (12. Dezember 2025). Data ethics in historical research: Recognizing responsibility. DH Lab. Abgerufen am 16. Mai 2026 von https://doi.org/10.58079/15bzu
2 Gedanken zu „Data ethics in historical research: Recognizing responsibility“