Research Document No. 09 Authenticity, Trust and Artificial Intelligence How Archival Science can contribute to trustworthy AI
PDS & Ged/A | Research Notes
Research Document No. 09
Authenticity, Trust and Artificial Intelligence
How Archival Science can contribute to trustworthy AI
10 May 2026
🌎 Open Science
The PDS & Ged/A Research Notes form part of an ongoing Open Science initiative dedicated to documenting the evolution of our research programme in Digital Archival Science. Rather than presenting only completed results, we openly share emerging questions, provisional findings, evolving hypotheses and research agendas, inviting the international archival community to participate in the development of knowledge.
📚 What have we learned so far?
During the past several months our research programme has gradually evolved.
We argued that digital preservation must be systemic.
We proposed the concept of Digital Archival Ecosystems.
We revisited the OAIS Reference Model from an archival perspective.
We expanded the meaning of the Chain of Archival Digital Custody.
We recognised an epistemological intersection between the European archival tradition, the iSchools movement and Computational Archival Science.
We observed that computational methods generate new archival questions.
Finally, we began to understand that intelligent agents are becoming active participants within Digital Archival Ecosystems.
Looking back, it now seems that all those discussions were gradually leading us toward a much broader question.
How can trust be preserved?
🔄 How our research question evolved
Our research questions have changed throughout this series.
Initially, we asked how digital records could be preserved.
Later, we investigated how documentary continuity could be maintained.
More recently, we explored how intelligent agents reshape Digital Archival Ecosystems.
Today our central question appears to be different.
How can trust be preserved when archival processes are increasingly shared between human professionals and intelligent agents?
Perhaps this evolution describes our research programme better than any formal definition.
Perhaps Archival Science has always preserved trust
For decades we have said that Archival Science preserves records.
That statement remains true.
Yet it may no longer be sufficient.
Institutions rely on records because they trust them.
Courts accept records because they recognise their authenticity.
Citizens exercise rights because documentary evidence is trusted.
Organisations preserve records because institutional accountability depends upon them.
In every one of these situations, something precedes the record itself.
Trust.
Perhaps records have never been the final objective of Archival Science.
Perhaps they have always been the means through which institutions preserve public trust.
📓 Laboratory Notes
May 2026
During one seminar meeting someone made a remarkably simple observation.
"Perhaps we have spent decades asking how to preserve records, when our real question should always have been how to preserve trust."
No one replied immediately.
The room became unusually quiet.
Not because the statement solved a problem.
But because it fundamentally changed the way we interpreted our own research.
💡 One of our research findings
Throughout our recent discussions we gradually realised that authenticity can no longer be understood solely as an intrinsic property of records.
Increasingly, it also depends upon the transparency, traceability and intelligibility of the human and computational processes that create, organise, describe, preserve and provide access to those records.
Perhaps authenticity itself is becoming a property of the entire archival ecosystem.
Artificial Intelligence does not simply generate records
It also generates decisions.
For many years we associated Artificial Intelligence primarily with automation.
Today that interpretation appears incomplete.
Intelligent agents classify records.
Generate descriptions.
Recommend decisions.
Organise information.
Prioritise search results.
Support institutional workflows.
Influence documentary processes.
In other words, they increasingly participate in archival decision-making itself.
That observation changes the focus of our research.
Preserving records alone may no longer be sufficient.
We may also need to preserve the processes that shaped them.
⚖️ An Archival Dilemma
Imagine two archival descriptions that are completely identical.
One was created by an archivist.
The other was generated by an intelligent agent.
Both contain exactly the same metadata.
The same classification.
The same descriptive elements.
The same documentary relationships.
Yet one important question remains.
If we cannot understand how the intelligent agent produced that description, should both descriptions inspire the same degree of archival trust?
At present we simply do not know.
But this question will continue guiding our research.
Perhaps transparency must also be preserved
Traditionally Archival Science has focused on preserving records.
Our recent discussions suggest that another responsibility is emerging.
We may also need to preserve the intelligibility of computational processes.
When algorithms reorganise metadata.
When intelligent agents recommend archival descriptions.
When Artificial Intelligence participates in appraisal.
When machine learning models influence documentary decisions.
Those computational processes also become part of the documentary history.
Perhaps preserving records alone is no longer sufficient.
Perhaps we must also preserve the pathways through which records acquired their present form.
🔬 An unexpected outcome of our seminar
While examining several Artificial Intelligence platforms we noticed an interesting pattern.
Almost all of them preserve outputs.
Very few adequately preserve how those outputs were produced.
Perhaps one of the greatest archival challenges of the coming decade will not be preserving AI-generated records.
It will be preserving evidence of algorithmic processes themselves.
Is authenticity changing its meaning?
For generations authenticity has been understood primarily as a property of records.
Our discussions now suggest another possibility.
Authenticity may increasingly become a property of processes as well.
Records remain authentic.
But trust must also extend to the computational systems that classify, describe, organise and interpret them.
Perhaps documentary authenticity and process authenticity will become inseparable.
This hypothesis still requires careful investigation.
🤔 A hypothesis we partially revised
At the beginning of our research programme we believed it would be sufficient to preserve records created or processed by Artificial Intelligence.
Today that assumption appears incomplete.
We may also need to preserve evidence documenting the computational processes that participated in archival decision-making.
That revision significantly expands the role of Archival Science.
Towards an Archival Science of Trust?
This question has repeatedly emerged throughout our seminar.
Perhaps the greatest contribution Archival Science can make during the coming decade will not simply be preserving digital records.
Perhaps it will be preserving trust.
Institutional trust.
Documentary trust.
Algorithmic trust.
Public trust.
We certainly do not yet know whether this hypothesis will prove correct.
Nevertheless, it has become one of the central organising ideas of our research programme.
🌱 A hypothesis under construction
Perhaps future authenticity will depend simultaneously upon preserving records and preserving the intelligibility of the computational processes that shape them.
This hypothesis will continue guiding our future investigations.
🧭 Research Agenda
At the conclusion of our seminar we agreed to preserve several questions that will shape the next phase of our research.
How should algorithmic decisions become archival evidence?
How can computational processes themselves be preserved?
How should transparency and explainability become archival requirements?
How can the Chain of Archival Digital Custody evolve within AI-assisted environments?
Are we witnessing the emergence of an Archival Science of Trust?
These questions do not conclude our research programme.
They define its next stage.
We continue learning
Perhaps the greatest contribution of this Research Note lies not in the answers it offers.
But in the question it ultimately asks.
For decades Archival Science has asked how records should be preserved.
Today we increasingly ask how trust itself can be preserved.
If that shift truly represents the future of our discipline, then Archival Science may have an essential contribution to make—not only to archives, but also to the broader international discussion on trustworthy Artificial Intelligence.
We do not yet have definitive answers.
That is precisely why we believe this conversation should remain open.
💬 Let's continue the conversation
Within your own professional or research environment,
will authenticity alone remain sufficient to sustain institutional trust in AI-assisted archival environments?
Or will Archival Science need to develop new concepts capable of addressing transparency, explainability, accountability and algorithmic trust?
We would genuinely welcome your perspective.
Your experience may help shape the next stage of this collective research programme.
📅 Next Research Note
Algorithmic Governance and Archival Science
How archival principles may contribute to transparency, explainability, accountability and trustworthy Artificial Intelligence.
25 May 2026
The research continues.
And we hope you will continue this journey with us.
Comentários
Postar um comentário
Comente aqui, por favor .....