The contemporary moment seems to be one of “enchanted determinism”—a constructed belief that technology will inevitably find the right answers if fed enough data. Yet the familiar principle of “Garbage In, Garbage Out” remains as relevant as ever. The “garbage” in this equation increasingly takes the form of bias, manifesting in algorithms that discriminate against marginalized populations and (digital) systems that reproduce harmful content.
For digital humanities researchers, this challenge is compounded by multiple intersecting forms of bias they must navigate: archival biases in source selection, historical power structures in interpretation, representational biases in digitisation, and algorithmic biases in analysis. Despite growing attention to ‘bias mitigation’, the term carries different meanings across disciplines, complicating systematic approaches. This conceptual instability, if left unexamined, has the tendency to render bias both omnipresent and at a risk of becoming meaningless.
The Combatting Bias project focuses specifically on datasets as its primary unit of study, as a critical intersection point where creators, users, and digital infrastructures meet. It acknowledges that eliminating bias entirely is an impossible task. Instead, it examines the vocabulary of bias, how it transforms across research stages and contexts, using it as a category of analysis and tool for critical reflection. The project is developing a framework with three components:
- A Bias Thesaurus: A list of the concepts connected to bias (such as representation, offensive language, FAIR, CARE, silences, ETC)
- A Bias-Aware Data Lifecycle Model: Showing where and how bias manifests at different research stages
- Practical Guidelines: Includes reflective questions at each stage of the dataset lifecycle and illustrative examples, and “good-better-best” recommendations
The idea of this framework is to shift thinking about bias merely as a problem, to thinking about it as conditions of production and features of the dataset itself. By explicitly describing these conditions of production, researchers can enhance transparency, improve dataset documentation, and enable more informed reuse of their data. Mrinalini will present this ongoing work through examples and interactive exercises with participants.
Mrinalini Luthra is a Data Ethics Specialist on the Combatting Bias project. She engages with questions around responsible stewardship of (historical) data and how to express the subjectivities and relationality of data, technologies, and interfaces. Trained as a mathematician, philosopher, and logician, her interests lie at the intersection of technology, ethics, and design.
Combatting Bias is a NWO funded project, which focuses on the ethical creation of datasets for the social sciences and humanities. It is a collaborative initiative based at the Huygens Institute and International Institute of Social History in Amsterdam, Netherlands. It consists of partnerships with four projects working with digitisation of colonial archives: Slave Voyages, GLOBALISE, Exploring Slave Trade in Asia, and Historische Databases of Suriname and Curaçao. CB’s work is enriched by its advisors from different geographies and disciplines such as museum studies, critical archival studies, ethnomusicology, history, computer science, and economics.
Wednesday, 26 March 2025
14.00 - 15.00
C²DH Open Space, 4th floor Maison des Sciences humaines
and online
