A corpus is not found. It is made.
This ongoing research-creation project grows from my SSHRC-funded investigation of participatory AI knowledge environments and my collaboration with computer scientists, developers, and students who build and maintain HelpMe, a university-hosted retrieval-augmented AI system.
My work with HelpMe has become an experiment in interdisciplinary collaboration and the possibilities of computational design. Working with graduate students in Making with Machines, I use the system in ways its original design did not anticipate: as a collectively curated knowledge environment, an object of critical inquiry, and a site for speculative experimentation. Students contribute sources, question what becomes retrievable, and consider how context, provenance, and cultural authority might be preserved as knowledge is fragmented and recomposed.
These experiments also create productive tensions with the system's technical architecture. Through ongoing conversations with the computer science team, I push questions of meaning, interpretation, and stewardship into decisions about retrieval and system design. Their understanding of the system's computational possibilities and constraints, in turn, reshapes my own thinking. Our different disciplinary perspectives continually complicate and deepen what we understand HelpMe to be, what it does, and what else it might become.
Speculative design takes these inquiries further. Rather than accepting existing computational arrangements as given, I invite students and participants to imagine futures in which the relations between people, knowledge, and machines have developed differently. Through fictioning, prototyping, and coding, they construct artifacts from these possible futures: interfaces, computational experiments, and fragments of alternative technological worlds. These are not simply illustrations of imagined futures, but material propositions through which different relations, values, and forms of agency can be explored and tested.
The embedded coded artworks are speculative imaginings of flows of thinking: ideas gathering, branching, colliding, disappearing, and returning. They do not represent the actual operations of HelpMe, but explore how code might give imaginative form to processes of knowing that exceed any single computational representation.
I approach this evolving environment—its corpus, architecture, social relations, and speculative artifacts—as material for further research-creation. Through fictopoiesis, I intend to deepen these experiments into the design of other possible machines, asking how artistic inquiry and interdisciplinary collaboration might transform not only what computational systems know, but how they make meaning and participate in the construction of worlds.
The corpus began with texts selected for the course and has expanded through student contributions.
87 texts · 35 added by students · 8 flagged for careful stewardship
Each addition is a curatorial decision: what does this source add, what does it complicate, and what context should remain attached to it once it becomes retrievable through an AI system?
⚑ A stewardship flag means: slow down.
It marks context that should travel with a source, such as provenance, cultural or community authority, genre, methodological limits, jurisdiction, copyright, or conditions of use.
This raises a design question:
Can provenance and stewardship context travel with a retrieved chunk, rather than remaining only at the document level?
Retrieval changes knowledge.
When documents are chunked and returned through a RAG system, context can disappear. Sources with different forms of authority can begin to look equivalent.
We ask:
- What can the system answer now that it could not answer before?
- What disappears through chunking?
- Does a citation preserve enough provenance?
- Which perspectives remain absent or overrepresented?
- What might be inappropriate to make retrievable at all?
Retrieval is not the same thing as knowing. A citation is not automatically provenance.
A speculative design workshop invites participants to follow an AI practice into a future where it has become ordinary, make a fragment from that future, and then change one relation.
A rule.
An interface.
Who controls the corpus.
What travels with a source.
Where judgment remains human.
How refusal becomes possible.
Prototype to learn, not to predict.
Workshop slides, ETEC 565R Making with Machines, UBC MET, 2026. Participant work from these workshops is part of ongoing research and isn’t shown here. Workshop tool: AI Futures map.
The larger SSHRC project investigates how people can exercise judgment, participation, and responsible refusal within AI-mediated knowledge environments.