Saturday 01 February 2025
As the world of artificial intelligence continues to evolve, concerns about data governance and ownership have become increasingly pressing. A recent experiment, dubbed the Choral Data Trust Experiment, aimed to address these issues by exploring novel approaches to collective data governance.
The project brought together 15 community choirs from across the UK, who recorded performances of a songbook composed by artists Holly Herndon and Mat Dryhurst. The recordings were then used to train state-of-the-art models for an exhibition at the Serpentine in London. But what sets this experiment apart is its focus on empowering data contributors through collective governance.
The researchers behind the project recognized that traditional approaches to data governance, such as opt-in and opt-out systems, may not be sufficient. Instead, they sought to create a framework that would allow data contributors to have a greater say in how their data was used and governed.
To achieve this, the team developed a Trusted Data Intermediary (TDI) – a legal entity responsible for managing the dataset and ensuring that it is used in accordance with the preferences of the data contributors. The TDI also served as an intermediary between the data contributors and the AI developers, facilitating communication and negotiation around the use of the dataset.
One of the key findings of the experiment was the importance of transparency in data governance. When asked about their comfort levels regarding the use of their data by users beyond the project, over 25% of participants responded that they were not comfortable with it. This highlights the need for clear and transparent communication around data practices and usage.
The survey also revealed that participants were more interested in group recognition than individual credit. This suggests that governance mechanisms should be able to interface with semantic groups rather than solely individuals.
To address these findings, the researchers developed a Data Rights Mandate that enables solo singers whose voices are captured in the dataset to mandate the TDI to exercise their GDPR UK data rights on behalf of the group.
The experiment also included a Licence Preferences Polis, which allowed participants to vote on different scenarios for potential dataset users and use cases. The results indicated that a significant majority of participants were interested in sharing the dataset with users who comply with licence terms and are willing to re-license under the same terms.
The Choral Data Trust Experiment has demonstrated the potential benefits of collective data governance, particularly when it comes to empowering data contributors.
Cite this article: “Harmonizing Data Governance: The Choral Data Trust Experiment”, The Science Archive, 2025.
Data Governance, Ai, Collective Ownership, Choral Data Trust, Trusted Data Intermediary, Transparency, Data Rights, Gdpr Uk, Licence Preferences Polis, Group Recognition, Individual Credit
