Exploring the Intersection of AI and Archives: Key Insights from Experts

Opening Keynote: AI and Archives – Applications, Implications, and Possibilities

Presented by Ray Pun, Helen Wong Smith, and Thomas Padilla at the AI and Libraries 2 Mini Conference

Introduction

In the opening keynote of the "AI and Libraries 2: More Applications, Implications, and Possibilities" mini conference, Ray Pun welcomed attendees to an engaging session focused on the intersection of artificial intelligence (AI) and archives. The event built upon the previous month's conference, which saw over 16,000 sign-ups, and celebrated Ray Pun's recent election as President-Elect of the American Library Association (ALA).

Joining Ray Pun were two distinguished professionals in the field of archives:

• Helen Wong Smith: President of the Society of American Archivists (SAA) and Archivist for University Records at the University of Hawaii at Mānoa.
• Thomas Padilla: Deputy Director of Archiving and Data Services at the Internet Archive.

The session aimed to explore the current landscape of AI in archives, discuss ethical considerations, and examine how AI can make born-digital collections more accessible and usable.

The Professional Landscape of AI in Archives

Helen Wong Smith's Perspective

Helen emphasized that AI integration into the archival profession offers promising opportunities for enhancing efficiency, accessibility, and deriving new insights from archival collections. However, she also highlighted several challenges that require careful management:

• Quality and Ethics: Ensuring that AI-generated metadata and content maintain the authenticity and trustworthiness of archival records.
• Privacy: Navigating data privacy concerns when implementing AI technologies.
• Professional Adaptation: The need for ongoing dialogue, research, and training to effectively integrate AI while preserving the nature of archival records.

Helen introduced the concept of "paradata," which involves capturing information about the procedures, tools, and individuals involved in creating and processing information resources. She stressed that paradata is essential for maintaining the authenticity, reliability, and usability of records in the context of AI-generated content.

Thomas Padilla's Perspective

Thomas noted the significant engagement of senior leadership in exploring the potential of AI within libraries and archives. He mentioned initiatives like the ARL (Association of Research Libraries) and CNI (Coalition for Networked Information) Task Force focused on scenario planning for AI in these fields.

However, Thomas expressed concerns about the for-profit capture of library and archival work, cautioning against over-reliance on specific products or proprietary technologies. He emphasized the importance of a more holistic and less product-centric approach to AI integration, suggesting that focusing on overarching frameworks and values would be more beneficial for the profession.

Ethical Considerations in AI and Archival Processing

Thomas Padilla's Insights

Thomas highlighted the ethical complexities that arise when AI perpetuates existing societal biases, referencing Safiya Noble's work on "Algorithms of Oppression." He argued that it's insufficient to accept these biases as inevitable and stressed the need for proactive responses to address inequities exacerbated by AI technologies.

He advocated for "bias management," suggesting that while subjectivity in archival description is unavoidable, it must be anchored in consistent values that prioritize human rights and historical understanding. Thomas also called for regulatory frameworks to provide clarity and consistency in ethical approaches to AI in archives.

Helen Wong Smith's Insights

Helen echoed the importance of addressing biases in AI-generated metadata and content. She raised concerns about AI's potential to perpetuate inaccuracies and misconceptions, particularly in generative AI that produces new content based on existing data.

She emphasized the necessity of codified record-keeping practices for creators using AI, referencing Jessica Bushey's work on AI-generated images as an emergent record format. Helen reiterated the importance of paradata in documenting not just the tools used but also the reasons, methods, and contexts in which they are applied.

AI and Born-Digital Archives

Enhancing Accessibility and Usability

Helen outlined several ways AI can improve access to born-digital collections:

• Automating metadata creation
• Content classification
• Natural language processing
• Image recognition
• Enhanced search capabilities

However, she also identified barriers to implementation, including:

• Lack of knowledge and competencies within the archival profession
• Reliable technologies and interoperability issues
• Economic constraints and personnel expertise
• Data privacy and security concerns
• Ethical considerations and cultural sensitivity

Addressing Backlogs with AI

Thomas discussed the potential of AI in addressing access issues for backlogged materials, particularly those in less commonly known languages. He highlighted the challenge of insufficient resources and the difficulty in hiring personnel with the necessary language skills to catalog these materials.

Thomas proposed leveraging AI advancements in world languages, possibly in collaboration with companies like Meta, to process and make these materials discoverable. He emphasized that minimal digitization combined with AI could help fulfill the access and preservation mission of archives.

Staying Informed and Managing Overwhelm

Thomas Padilla's Approach

Thomas acknowledged the feeling of overwhelm many professionals experience due to the rapid developments in AI. He recommended:

• Adopting a utilitarian approach to AI as a tool
• Grounding oneself in the history and values of the profession
• Practicing careful curation of information sources
• Utilizing platforms like LinkedIn for professional updates
• Setting up curated Google Alerts for topics like AI in libraries, archives, and regulation

Helen Wong Smith's Resources

Helen suggested leveraging collaborative initiatives like the InterPARES Trust AI project, an international and interdisciplinary effort aimed at designing and developing AI to support trustworthy public records. The project's goals include:

• Identifying AI technologies to address critical records and archival challenges
• Determining the benefits and risks of using AI on records and archives
• Ensuring archival concepts and principles inform the development of responsible AI
• Validating outcomes through case studies and demonstrations

Helen emphasized the importance of engaging with such resources to stay informed and contribute to the ongoing dialogue around AI in archives.

Conclusion

The opening keynote provided valuable insights into the intersection of AI and archives, highlighting both opportunities and challenges. Ray Pun thanked the speakers and attendees, encouraging continued dialogue and exploration of these critical topics.

As AI technologies continue to evolve, the archival profession must navigate ethical considerations, enhance competencies, and develop strategies to leverage AI responsibly. By fostering collaboration, staying informed, and grounding practices in core values, archivists can effectively integrate AI to enhance accessibility and preservation.

Note: This summary is based on the opening keynote delivered by Ray Pun, Helen Wong Smith, and Thomas Padilla at the AI and Libraries 2 Mini Conference.