AI technologies are increasingly being used across society, and museums are increasingly providing a crucial space for critical technology discourse and civic engagement with these technologies. The Museums + AI Network created an ethically robust framework to help museums better understand the possibilities of AI technologies within the Cultural Heritage sector. This project was led by Goldsmiths, University of London and Pratt Institute, New York. It was funded by the Arts and Humanities Research Council (UK).
Innovation Overview
‘AI: A Museum Planning Toolkit’ provides a benchmark of technical possibilities for senior managers, policy makers, and a practical framework for practitioners. When we started this work, museums and cultural institutions were increasingly using AI technologies, however current sector codes of practice and regulatory frameworks did not support this emerging technology. In response, the world’s first Artificial Intelligence network for museums was founded. The Museums + AI Network brought together senior museum professionals, funders, and policy makers from the UK and USA to develop policy and practice around AI, ethics, and museums. One key outcome of this groups work is an edited open access toolkit.
Museums are already using Computer Vision technologies to make their collections discoverable; Natural Language Processing to analyse visitor feedback; Predictive Analytics to forecast visitor numbers, spend and exhibition naming; Collecting and Curating AI technologies; and using public programs to educate visitors about the impact of these technologies on individuals and societies. The landscape is broad and varied, but data and the processing of that data has always been at the core of museum practice. Indeed, much of the Web we know today is premised on archive and museum practices.
Museums are by their very nature data centric institutions, they are collectors and creators of a diverse range of data – be that the bone density of a dinosaur, the market value of an artwork, the most viewed collection item on their website, or how long visitors spend in a particular gallery. These datasets and the ethical and legal frameworks that govern them are complex, a complexity that is part drawn out of the differing motivations and rationales for the collection and creation of these data sets.
Given the swift evolution of the AI field, the toolkit does not offer a definitive ‘how to’ guide by design. Instead, it offers case studies and worksheets that enable museums too critically reflect on the capabilities and ethics of using AI. The toolkit includes case studies, glossary and three facilitated workshop guides. These focus on capabilities (data, tools, resources, skills, organization, stakeholders); Ethics workflow (data input, collection, clean up, training data, testing, application, data output, evaluation); Stakeholder management (beneficiaries, data ownership, conflict of interest, data privacy).
The toolkit was developed through executive workshops in London, New York, San Diego, the network facilitated in-depth discussions around the parameters, methods, and paradigms of AI. As project leads Dr Oonagh Murphy and Dr Elena Villaespesa designed an action research methodology to collaboratively develop new models of practice. Input was drawn from 50 world experts, 200 members of the public, participants included: Arts Council England, The Department for Digital Culture Media and Sport, Bloomberg Philanthropy, Arts and Humanities Research Council, National Endowment for the Humanities, The Getty, Smithsonian Institute, NESTA.
The toolkit has been used by network members, and members of the museums, and cultural policy sector more widely.
This is an innovative Tool(kit) because:
1) It has been iteratively developed through an action research project.
2) The tool has been developed with support from stakeholders drawn from: civic society, practitioners, government bodies, and members of the public.
3) The tool provides a framework rather than set of rules.
4) It supports critical thinking with big challenges.
The Tool has been used by a broad range of museums, policy makers, funders, educators and researchers internationally. The team who developed the tool continue to work with colleagues to support the development of policy initiatives, funding calls, and the development of AI projects in a cultural context. To date this tool has been shared with professional groups in: UK, USA, Germany, Spain, Austria, Brazil, South Korea, Malta. Museums that have engaged in this work are as broad the MoMA (New York), V&A (London), ZKM (Karlsruhe) and Museo Nacional del Prado (Madrid). In 2022 it was translated with additional case studies into Spanish and German.
Collaborations & Partnerships
The network was a collaboration led by Dr Oonagh Murphy (Goldsmiths, University of London) and Dr Elena Villaespesa (Pratt Institute), with founding institutional members being the American Museum of Natural History, The Metropolitan Museum of Art, The National Gallery (UK). The action research project brought together: 50 experts from 15 museums, 6 universities, multiple government departments and funders from the UK and USA. It also engaged with 200 members of the public.
Users, Stakeholders & Beneficiaries
Citizens: Through events at the Barbican Centre in London and Cooper Hewitt, Smithsonian Design Museum in New York, 200 citizens were provided with the opportunity to influence museum policy on data, AI and ethics.
Government officials / Civil Society Organisations: The government officials, civil servants, and policy makers from funding organisations that helped us to develop the toolkit were able to embed their learning into policy development and funding priorities for AI work in this area.
Results, Outcomes & Impacts
The English language version of the toolkit has been downloaded 3,000+ times, with print copies distributed to policy makers in the UK and USA. The Library of Congress stated that they sought to ‘mirror’ the strategic model outlined in the toolkit, in their Machine Learning strategy. The Toolkit was highlighted to the European Parliament’s Committee on Culture and Education who are seeking to develop policy around AI within the Creative and Cultural Industries. Evidence drawn from this project was submitted to the Inquiry on the Future of the Creative Industries, which is being held by the House of Lords Communications and Digital Committee (UK). Project leads Dr Murphy and Dr Villaespesa were commissioned to help the region of Baden-Württemberg develop an AI working group.
Challenges and Failures
The project did not examine the impact of these technologies on the climate emergency. Going forward we would advocate that climate impact be embedded into this research.
This project was developed in 2019 and published in 2020. At this time there was a real energy around these these technologies their possibilities and the role of museums as civic institutions. Since then Covid has shifted the focus from innovation to delivery in a museum context. It is important that museums now refocus equally on delivery and innovation to ensure they maintain their important civic role in society.
Conditions for Success
This Tool(kit) is designed to provide multiple entry points to the concept of AI, data and ethics. The case studies and glossary support senior leaders and policy makers to understand the possibilities and use cases of these technologies. While the facilitated workshop formats support those working with these technologies in a more applied setting. This universal ‘tool’ is unique in that it speaks to multiple internal and external stakeholders, with different degrees of technology knowledge, strategic thinking and leadership responsibilities. It provides a shared starting point for the development of technology applications within the realm of social purpose institutions, concerned with issues of national identity, race, representation, politics, and historical interpretation.
Replication
Since the ‘AI: A Museum Planning Toolkit’ was first published in 2020, we have worked with partners in Germany and Spain to translate and update the toolkit so it reflects the national context in those countries.
As a result, across the three languages we now have use cases from the UK, USA, Germany and Spain. Case studies are draw from The Metropolitan Museum, American Museum of Natural History; The National Gallery, The Science Museum; ZKM, Ludwig Forum Aachen, Das Badische Landesmuseum; Museo Nacional del Prado.
The combination of local case studies and facilitated workshop tools support the adoption of this tool across international contexts.
Lessons Learned
This was a fast paced, interdisciplinary, internationally focussed action research project that created a series of outputs. The focus of this case study is on the Toolkit that was produced. The toolkit was iteratively developed by participants in our network, and was published open access, and available in print and online. Accessibility is key. Academic work often exists behind paywalls, in academic journals. Whilst we have published academics articles from this project, we deem the open access toolkit to be the key outcome in terms of industry impact, influencing change in terms of policy and practice. It is important to think about how a project will be received, who you seek to influence and the change you seek to create when developing work aimed at public sector innovation.
The Museums + AI Network 