|
|
|
|
Promote a more inclusive and respectful approach to the description of digital collections and the telling of stories and histories of minoritized communities.
Cultural heritage practitioners
Germany, Italy, The Netherlands, Greece, Cyprus, Belgium, France
Digital Europe Programme
STRENGTHS
Open-Access and Free Tools
All DE-BIAS tools, vocabulary lists, and training materials are openly available at no cost, removing financial barriers for small institutions.
User-Friendly Interface
The DE-BIAS tool is accessible via a simple web application and includes clear suggestions, contextual explanations, and multilingual support.
Low Technical Skill Requirement
Designed for non-technical cultural professionals; no coding or system integration required for using the standalone version.
Ready-to-Use Training Resources
Interactive courses, guides, and glossaries are tailored for cultural staff without prior DEI or digital training.
Modular Use
Institutions can use just the parts they need (e.g., vocabulary, auditing tool, or community engagement templates) making the project highly adaptable.
OPPORTUNITIES
Growing Institutional Focus on Diversity and Inclusion
Rising public and policy expectations for cultural organisations to address equity and representation strengthen the project's relevance.
Potential for Local Partnerships
The initiative’s methods are ideal for collaboration with local schools, community archives, and advocacy groups to enhance inclusive documentation.
WEAKNESSES
Initial Development Complexity
While the tools are open-access, the development of AI-based solutions and multilingual vocabularies required high technical expertise and coordination that may be difficult for small institutions to replicate independently.
Limited Local Customisation Support
Adapting the vocabulary or tool for highly specific regional or community contexts may require additional digital or linguistic capacity.
Dependency on Metadata Quality
The tool is most effective when metadata is already structured and machine-readable. Many small institutions may have inconsistent or unstandardised metadata, limiting the tool's impact without prior data cleaning.
Training Time Investment
Although resources are user-friendly, staff still need time to engage with the training and implement the changes, which may be challenging in understaffed environments.
THREATS
Low Digital Literacy Among Staff or Volunteers
In small cultural institutions, particularly in rural or under-resourced areas, there may be difficulty in using even simple digital tools.
Resistance to Change or Terminology Updates
Updating historically embedded language in catalogues may encounter institutional or stakeholder resistance due to heritage concerns or fear of ‘erasing’ past terminology.
Language Adaptation Barriers
Although the tool supports multiple major European languages, local dialects or minority languages common in smaller communities may not be covered.
The DE-BIAS project is built around a series of carefully integrated digital components that work together to help cultural heritage institutions identify and reduce biased or harmful language in their metadata.
Application Programming Interface (API)
The DE-BIAS tool offers an API to enable seamless integration into an institution’s existing digital infrastructure. Rather than manually uploading files for analysis, institutions with larger or more advanced systems can automate the process. For example, a museum’s database system can “talk” directly to the DE-BIAS tool via the API, allowing metadata entries to be reviewed in real time and flagged for biased language as they are created or updated. This streamlines the workflow and supports sustainable implementation, especially in larger collection environments.
Metadata
At the core of the DE-BIAS project is the review and improvement of metadata, the descriptive information associated with digital objects, such as titles, summaries, subjects, and keywords. Many cultural institutions, particularly those managing colonial-era or legacy collections, hold metadata that includes outdated or exclusionary terminology. The project uses the digital tools to scan these metadata fields and identify where problematic language exists, offering practical ways to revise the descriptions to reflect more respectful and inclusive narratives.
Natural Language Processing (NLP)
To analyse the vast amounts of text data in metadata records, the DE-BIAS tool applies Natural Language Processing (NLP), a type of artificial intelligence that enables computers to understand, interpret, and process human language. NLP allows the tool to go beyond basic keyword matching by understanding sentence structure and the context in which terms are used. This makes the tool more accurate, able to differentiate between a term that is part of a historical quote and one that appears in a museum’s own description.
Named Entity Recognition (NER)
A specific feature of the NLP engine is Named Entity Recognition (NER), which identifies proper nouns and terms related to people, places, groups, or time periods. In DE-BIAS, this helps isolate culturally sensitive terms - such as names of ethnic groups, geographic regions, or religious communities - that may be used inappropriately or without context. By recognizing these entities, the tool flags metadata entries where extra care or revision is needed, helping institutions avoid misrepresentation or offense.
Controlled Vocabulary
The project also developed a multilingual controlled vocabulary, essentially a reference list of about 700 potentially problematic or outdated terms, with explanations and suggested alternatives. This vocabulary is central to the detection process and serves as an educational resource in its own right. Cultural professionals can consult it to understand the origins of certain terms and how best to revise them. The vocabulary is available in five languages, making it accessible to a wide European audience and adaptable to different institutional needs.
Bias Detection Tool
The central solution of the DE-BIAS project is a web-based bias detection tool that allows users to upload metadata (typically in spreadsheet form), scan it for biased language, and receive detailed feedback. Each flagged item includes the term used, its context, an explanation of why it may be harmful, and a link to the controlled vocabulary with suggested alternatives. This gives cultural staff actionable insights and a clear starting point for improving their records.
The tool is released as open-source software, meaning its source code is freely available for reuse, modification, and sharing. This supports institutional collaboration, enables local adaptation (e.g., adding regional terms or adjusting the interface), and reduces costs for small organisations. Open access also ensures transparency and encourages ongoing community-driven development.
Designed for accessibility, the tool runs in any web browser without installation, requiring only an internet connection. Its straightforward interface includes step-by-step instructions for uploading, reviewing, and exporting data, making it usable even by non-technical staff. This low-barrier approach ensures small and resource-limited cultural institutions can effectively benefit from the tool.
On Organisations
• 4.5 million+ metadata records analysed for harmful language across five languages.
• Institutions improved metadata inclusivity and integrated the DE-BIAS tool into workflows.
• Staff trained through dedicated webinars and capacity-building sessions.
On Target Groups
• Marginalized communities were directly involved in co-creating the inclusive vocabulary.
• Developed an interactive online course to help professionals recognize and reduce bias.
On the Wider Community
• Promoted sector-wide awareness of inclusive metadata practices.
• Recognized in professional forums for shaping digital heritage policy and ethics.
• Influenced broader adoption of community-led and DEI-informed cataloguing practices.
Knowledge
• Understanding of metadata standards and cataloguing practices.
• Awareness of inclusive language, bias, and diversity in cultural heritage.
• Basic familiarity with how AI and text analysis tools work.
Digital Skills
• Ability to use web-based tools and interpret analysis results.
• Competence with spreadsheets for reviewing and editing metadata.
• Basic knowledge of integrating APIs (Application Programme Interface) into existing systems.
Project & Planning Skills
• Ability to plan metadata audits and allocate staff time.
• Skills in documenting changes and evaluating progress.
• Capacity to engage external communities or stakeholders for feedback.
Interpersonal Skills
• Clear, inclusive communication—especially when discussing sensitive terminology.
• Collaborative mindset to work across teams or with community partners.
• Openness to feedback and co-creation.
Attitudes
• Commitment to inclusion and ethical representation.
• Openness to changing legacy practices.
• Attention to detail and willingness to question existing standards.
It aims to identify, reduce, and raise awareness of bias in digital cultural heritage data, especially in metadata and AI-driven applications. Core aspects of the initiative include:
Key innovative principles for transfer in small, local institutions include:
The DE-BIAS Project uses a data-driven, ethical, and community-informed methodology to detect and reduce cultural bias in digital heritage metadata. Key elements include:
The methodology is modular and scalable, adaptable for both large institutions and small archives.
While the DE-BIAS tool itself is free and open source, implementation does involve some minimal to moderate resource allocation:
Estimated start-up costs:
€5,000+ if deeper integration or dedicated DEI training sessions are pursued with consultants.
Since the DE-BIAS platform already offers open-access, cost-efficient resources, it is the low-cost solution for inclusive metadata auditing. However, institutions with very limited resources could implement a simpler approach by:
These strategies allow institutions to apply DE-BIAS principles even without digital infrastructure or tech skills.
