Author: Lustre Network

Speaker: Glen Layne-Worthey, University of Illinois Urbana-Champaign 

Abstract: In a bona fide classic in the Digital Humanities canon, John Unsworth proposed seven “scholarly primitives” for humanities research: shared methods in the “traditional” humanities that he suggested should be reflected in the digital tools we create. Although Unsworth articulated these thoughts more than a quarter-century ago, he meant them to reflect longstanding scholarly practice, and by the same token they should reflect future scholarly practice as well — digital, AI-enhanced, and whatever comes next. 

Consistent with the stated purpose of the GLOW-Ready project, Unsworth’s scholarly primitives begin with “Discovery” — but crucially, they go beyond that.  This talk will focus on a few more of Unsworth’s scholarly primitives that I would hope we keep in mind as we make government information “GenAI-Ready”: Annotation, Comparison, Reference, Sampling, Illustrating, and Representing. 

Bio: Glen Layne-Worthey is the Associate Director for Research Support Services in the HathiTrust Research Center, based in the University of Illinois at Urbana-Champaign School of Information Sciences.  Glen was Digital Humanities Librarian in the Stanford University Libraries from 1997 through 2019, and was the founding head of the Libraries’ Center for Interdisciplinary Digital Research (CIDR), and a founding member of the Stanford Literary Lab. 

Long active in the international Digital Humanities community, he hosted the international DH2011 conference at Stanford, and was co-chair of the Program Committee for DH2018 in Mexico City.  He recently served as Executive Board Chair in the Alliance of Digital Humanities Organizations (ADHO), and as co-convener of its “DH in Libraries” Special Interest Group.  He is co-editor (with Isabel Galina) of The Routledge Companion to Libraries, Archives, and the Digital Humanities, published in 2025. 

In 2021 through 2023, Glen served as US PI (with Lise Jaillant as UK PI) of the international AEOLIAN project, jointly funded by the US National Endowment for the Humanities and the  UK Arts & Humanities Research Council.  Lise, Glen, and other AEOLIAN colleagues co-edited the volume Navigating Artificial Intelligence for Cultural Heritage Organisations, also published in 2025.   

Speaker: Dr Frédéric Clavert, University of Luxembourg 

Abstract: Chatbots are based on language models trained in various ways using vast amounts of training data, which, from a historian’s perspective, can be seen as a form of archive. Chatbots users enter prompts to obtain responses in the form of text, images, videos, etc. These outputs, just like the prompts themselves, can also be regarded as primary sources for the social sciences, both today and in the future. Using prompts that refer to the historical past as an example, this presentation will attempt to explore how a historian might make use of some of these sources, and by what methods. 

Bio: Frédéric Clavert is an assistant professor at the C2DH (University of Luxembourg), where he heads the European History Research Group. Having studied political science and the history of international relations at the University of Strasbourg, his early research focused on the monetary policy of the Third Reich. He then gradually turned his attention to the study of online memory phenomena on the one hand, and historians’ relationships with their primary sources in the digital age on the other. He recently co-edited the Memory Studies Review special issue ‘Artificial Intelligence and Collective Memory’ with Sarah Gensburger (CNRS / Sciences Po) and, with Caroline Muller as lead author, Écrire l’Histoire. Gestes et expériences à l’ère numérique.   

Speaker: Prof Giovanni Colavizza, University of Copenhagen / University of Bologna 

Abstract: Researchers approaching large, heterogeneous archival collections face a familiar bottleneck: knowing what a collection holds and where the relevant material sits. General-purpose chatbots promise a shortcut but introduce a deeper problem — they generate plausible text rather than citing sources, producing fluent answers detached from precise grounding. Relay, a retrieval-augmented tool developed at the Centre for Digital and Computational Humanities (UCPH), takes the opposite stance: every answer is grounded in a curated corpus and linked back to the passages it draws from, so a response is a starting point for verification rather than an endpoint. This talk demonstrates Relay in practice, walking through how a researcher’s question is retrieved against the collection, augmented, and returned with traceable evidence. I argue that the value of such tools for archival discovery lies less in generation than in groundedness — turning a collection into something a scholar can accurately interrogate and trust, thanks to strict verification and actionable provenance.

Bio: Giovanni Colavizza is Professor and Head of the Centre for Digital and Computational Humanities at the University of Copenhagen, Denmark (https://cdch.ku.dk). He is also an Associate Professor of Computer Science at the University of Bologna, Italy, and the CTO and co-founder of Odoma, a Swiss-based company developing GraphChat (https://graphchat.ai). Colavizza has a background in Computer Science and History and is specialized inAI applications in the GLAM sector (Galleries, Libraries, Archives, Museums). 

Speaker: John Sheridan, The National Archives

Bio: John Sheridan As Chief Digital and Information Officer, John is responsible for digital services, enabling The National Archives to fulfil its ambitions to become a digital archive by instinct and design. His role is to provide strategic direction, transform the digital offer, and to shape and drive forward web-based services. Prior to this role, John was Head of Legislation Services at The National Archives where he led the team responsible for creating the legislation.gov.uk website, as well overseeing the operation of the official Gazette. A former co-chair of the W3C e-Government Interest Group, John has a strong interest in web and data standards. He serves on the UK Government’s Open Standards Board which sets data standards for use across government. John was an early pioneer of open data and remains active in that community. John’s academic background is in mathematics and information technology, with a degree in Mathematics and Computer Science from the University of Southampton and a Master’s Degree in Information Technology from the University of Liverpool. John recently led, as Principal Investigator, an Arts and Humanities Research Council funded project, ‘big data for law’, exploring the application of data analytics to the statute book, winning the Halsbury Legal Award for Innovation. 

Speaker: Michael D. Thomas, National Archives and Records Administration (NARA)

Abstract: The United States government generates and maintains vast amounts of information, put to use every day for the public good. Some of this information, gathered in the interest of our national security, demands special safeguarding. The Information Security Oversight Office is singularly positioned to provide insight into the management and health of this unique – and uniquely valuable – national resource, performing an essential role in ensuring that national security information is both assiduously protected and judiciously shared. Balancing such secrecy with transparency remains a foundational tension in American governance. Today, it is widely recognized that we too often privilege the former, with our government keeping far too many secrets for far too long, failing to meet modern expectations of accessibility to records of extraordinary public interest. Chronic overprotection of information, infrastructure fragmentation, and inadequate public access to records of significant civic interest have consistently eroded public trust and squandered institutional value, while paradoxically undermining the security such secrecy purports to serve. We are in a moment of epochal technological change and to meet this moment our current concepts for information management must evolve. Modernization — encompassing the thoughtful integration of emerging technologies including artificial intelligence — presents a path to a more efficient, accurate, and automated classification and declassification system for national security information. Such advances could meaningfully resolve systemic deficiencies that have persisted largely unaddressed for decades. The challenge is substantial, to shift the structures – and cultures – of a system that has operated largely unchanged decades. But it is more than balanced by the scale of the opportunity that awaits, when we unlock the full value of our information, making good on our foundational promise of accountable government, and in so doing, advancing our nation’s security, prosperity, and values. 

Bio: Michael D. Thomas is Director of the Information Security Oversight Office, responsible to the President for oversight of U.S. Government-wide information security programs. He previously served on the White House National Security Council and at the Office of the Director of National Intelligence. Thomas began his career at the National Archives, departing to work as an investigator for private clients as well as in the public interest. He also spent time with a series of start-ups, including two ranked among Inc. Magazine’s “500 Fastest Growing Companies in America.” Thomas is a graduate of George Washington University, studying Philosophy and International Affairs. 

Speaker: Dr Cassandra Kist, University of Strathclyde

Abstract: Archival institutions are widely perceived as guardians of public knowledge, memory, and historical truth. Yet in an era where AI is reshaping digital access and escalating concerns around bias, privacy, and misinformation, trust is no longer sustained by institutional authority alone. Instead, it must be actively designed into user experiences. This presentation explores how authenticity can be operationalised to sustain public trust during interactions with heritage-based conversational agents. 

Drawing on recent research examining case studies of conversational agents (CAs) in museum contexts, I discuss how the emotional, material, and factual dimensions of authentic experiences are negotiated during design decisions. This includes how systems feel (tone, immersion, personality), what materials and people they represent (responsible and accurate portrayals of people and heritage), and the credibility of the information they provide. I show how heritage professionals frame design decisions in ways that support ‘real’ and ‘genuine’ experiences and simultaneously avoid creating inauthentic experiences that could perpetuate visitor mistrust. 

I further expand on the potentials and risks of designing conversational agents for authentic heritage engagement based on findings from a survey of museum visitors in the UK. Incorporating a visitor perspective illuminate overlapping and diverging concerns regarding genuine engagement with heritage through conversational agents. By translating these insights to the more sensitive domain of government records, I outline how organisations can create authentic experiences that communicate responsibility, privacy, and accuracy – not only as (essential) technical features but also as elements of user experience. 

Bio: Dr Cassandra Kist is a Chancellor’s Fellow at the University of Strathclyde in Computer and Information Sciences (Glasgow) where she teaches about Human-Computer Interaction and ethical issues in computing. During her PhD research, she was a Marie Curie Fellow in the Horizon 2020 European Union International Training Network POEM (Participatory Memory Practices). Her research combines several disciplines including Anthropology, Digital Cultural Heritage Studies, and Human-Computer Interaction to investigate the overlaps and disconnections between cultural heritage practices, digital platforms, and processes of social inclusion. 

Speaker: Professor Christopher (Cal) Lee, University of North Carolina

Abstract: Digital collections rest on binary foundations. All digital objects and associated metadata are composed entirely of binary values (bits). Technologies to manage and use digital objects can only perform actions that are reducible to binary values. But archival work involves continuous grappling with dualities. According to Etienne Wenger, a duality is “a single conceptual unit that is formed by two inseparable and mutually constitutive elements whose inherent tensions and complementarity give the concept richness and dynamism.” 

Resources are limited, and digital curation professionals cannot pursue all objectives equally. However, rather than simply picking one side of a duality, digital curation professionals must often pursue them (in a parallel or serial form) to varying degrees while ensuring a certain threshold level of commitment to each, finding what Paul Evans and Yves Doz call the “zone of complementarity.” The proper balance depends on a variety of contextual factors that evolve over time. I will focus especially on digital curation dualities relevant to machine learning, including when defining and selecting training data, deciding whether to develop a new model or rely on an existing one, and preserving evidence of machine learning. The dualities of digital curation provide a powerful way to strategize complex, often messy human activities that must be enacted through entirely binary representations. 

Bio: Christopher (Cal) Lee is Professor at the University of North Carolina. He has served as Principal Investigator and Co-Principal Investigator of numerous digital curation research and education projects.  He is a Fellow of the Society of American Archivists and served as editor of American Archivist.   

Speakers: Dr James Lapping & Piers Walker, Department for Science, Innovation & Technology

Abstract: A digital heap of a large public sector organisation typically consists of billions of messages and hundreds of millions of documents. Every one of these messages and documents will be in a container of some sort – for example an email account, a site or a drive. There are orders of magnitude less email accounts than messages. There are orders of magnitudes less sites/drives than documents. 

AI can be used to support decisions made at the container level. Acting at the container level improves the scalability of a scarce resource in any digital heap programme – the expertise of recordkeeping professionals acting as humans-in-the-loop.  

This talk describes approaches and data science experiments that are being worked up in Government Digital Service in DSIT with a view to providing practical ways of implementing the principles set out in the AI Insights guide Using AI to manage the digital heap.   

 This work includes  

• the development of a set of steps for the use of AI in the review and processing of content created on shared drives 

• data science experiments to break email accounts into smaller sub-containers. Such a capability could help an organisation establish a pipeline for the appraisal and sensitivity review of an email account that it deems likely (in whole or part) to have historical value 

Bios:

Dr James Lappin  

James is senior policy lead on digital knowledge and information management in the Office of the Government Chief Data Officer in GDS within DSIT.  He coordinates the cross-government AI for GKIM group and helped draft the AI insights guide Using AI to manage the digital heap.  James has worked in archives and records management in the UK and Europe for over 30 years. He was awarded a PhD by Loughborough University in 2023 for his thesis The science of record keeping systems.  He is the author of the Thinking Records blog. 

Piers Walker  

Piers Walker is Data and Technology Lead in the Data Innovation and AI Readiness team at the Government Digital Service. A Chartered Engineer, his experience spans analytics, data science, and technology management. He has designed and tested AI systems, co‑authored the HMG AI Playbook to support responsible AI adoption across government, and currently leads work on AI readiness.