ID: 34865
Libraries, archives, and mission-driven publishers have been key players in the global movement to increase open and equitable access to scholarship and to primary source materials. One key question that stewards of archival and general collections, and publishers of scholarly content, must wrestle with today is whether the principles behind making content open for individual readers and users can be applied to LLMs and generative AI tools.
Concerns over loss of provenance, control, and lack of attribution bump up against a conviction that the high quality content stewarded by research libraries, archives, and scholarly publishers would enhance the quality of output produced by AI tools. As AI systems seek access to scholarly content for training data, long-standing assumptions and values about openness, stewardship, control and provenance are being challenged and reexamined.
In this panel discussion, we bring together different perspectives on the core question of whether and how scholarly content should be open for AI use.
Panelists:
- Dave Hansen, Executive Director of Authors Alliance, argues that control and gatekeeping are the wrong approach for libraries and archives, and instead asserts that “building the infrastructure that supports open, accountable research of every kind.” will be the most values-aligned and productive role for the library community.
- Alison Muddit, Chief Executive Officer of the Public Library of Science (PLOS), asserts that AI and open access are not naturally in tension; but/and that a mission-driven publisher like PLOS must take seriously the fact that AI intensifies the need for rigor, transparency, and signals of trustworthiness. She emphasizes the responsibility to ensure that the scholarly record functions as trustworthy infrastructure for both human and machine reasoning.
- Chris Bourg, Director of Libraries at MIT, is a global leader in open scholarship and an advocate for the public mission of knowledge institutions. At MIT, she is co-chair of the MIT Working Group on Scholarly Content and Generative AI, and a member of the MIT Committee on AI in Teaching, Learning, and Research Training.
Panel facilitator:
- Mike Furlough, Executive Director of HathiTrust, works with dozens of member libraries to steward over 19 million digitized items from their collections, recognizing that memory institutions have a responsibility to make collections broadly accessible for all modes of research. However, emergent modes of research have brought new, more urgent demands for access to those collections, which in turn pose new questions regarding sustainable stewardship.
