Within the DE-BIAS project, a crowdsourcing campaign targeted to Italian Institutions was held the 25th September 2024. The objectives of the evaluation via the CrowdHeritage platform were to assess the performance of the DE-BIAS tool and the accuracy of the detection as well as to assess whether it is correct that a term is marked as bias in the context in which it is used.

During this event, organised by the Michael Culture Association, the European Fashion Heritage Association and ECCOM, the DE-BIAS project was presented and a live demonstration of the CrowdHeritage platform was performed followed by evaluation activities, where the participants were asked to leave feedback on the potential biassed terms identified through the CrowdHeritage platform and to add comments if they felt it necessary.

All participants found the experience highly engaging and gained insights that could be valuable for their work. During the final phase of the meeting, following the evaluation activity, certain issues were discussed. It was noted that the challenge often lies not in the individual words themselves, but in the context created by the words that come before or after them. In general, a problem of nounisation has emerged; individual subjectivity and sensitivity plays a key role in identifying bias, so empathy is needed.

This campaign was very important for the improvement of the quality and functionality of the DE-BIAS tool targeted to the cultural heritage sector.

___

About the DE-BIAS Project

Cultural heritage institutions in Europe and beyond have been undertaking work to describe and catalogue the historical objects in their collections for decades. However, once created, catalogue data are rarely updated to reflect changes in language and society. As a result, many object and collection descriptions that once fit into popular social narratives now convey outdated views that not only ignore and therefore alienate a wide range of people, but in some cases use language that is offensive, inappropriate or even harmful.

The DE-BIAS project aims to promote a more inclusive and respectful approach to the description of digital collections and the telling of stories and histories of minoritised communities. Over the course of two years, the project will develop an AI-powered tool to automatically detect problematic terms in cultural heritage metadata and provide information about their problematic background. It will use vocabularies that combine offensive language with contextual information and suggestions for appropriate terms.