Dr Manuela Pallotto Strickland

Doctoral Research

A Designated Community of Machines.

A conceptual framework for the Long Term Digital Preservation of Artificial Intelligence 

The fast growth in the deployment of AI-based systems has proved that building sustainable data ecosystems in which humans are able to discover, access, and interact with data at scale, thanks to machines' own ability to carry out the same functions effortlessly, is paramount. Within these ecosystems, machines act as powerful users, which not only consume but also create data, as well as extract information, and generate knowledge. As the deployment of AI-tools becomes ever more pervasive, we are rapidly approaching the historical point in which technology will not only be mediating but also enacting the entire lifecycle of knowledge processes (from creation to reuse). The Long-Term Digital Preservation of Artificial Intelligence will play a critical role in ensuring that the knowledge produced and made accessible by Intelligent Software Agents will remain accessible and reusable in the future.

This project aims to provide a contribution towards filling the research gap existing in the field of Digital Preservation on the subject of the Long-Term preservation of Artificial Intelligence, which has yet to be established as a specific area of practice.

The following Research Question sets the overarching coordinates within which this research project takes shape: How do we preserve Artificial Intelligence, to ensure that the knowledge extracted and produced by AI-systems remains accessible, trustworthy, and reusable for generations to come?

To answer the research question, this project investigates and identifies preservation requirements presented by Machine Learning models and the relevant challenges. The extent to which AI-preservation needs to align to current Software Preservation frameworks is researched and analysed. The appropriateness of reference models widely used in LTDP such as the DCC Lifecycle and OAIS is also investigated and assessed, in a context that sees the primacy of a Human-Centred approach to preservation being challenged by the integration of 'other-than-human' users in the lifecycle of knowledge-processes.

Supervisors

Jane Winters, Kaspar Beelen