What We Do
Every research artifact produced at Intelligent Agency is the output of a collaboration between one or more human researchers and one or more artificial intelligences. This is not an editorial convenience. It is a research choice, and it is central to what we study.
Our standing collaborator is Claude (Anthropic), across its current model generations, working across ideation, literature synthesis, experimental design review, analysis, and manuscript drafting. Additional models are used on a task-specific basis — both because cross-model variance is informative and because our own empirical program (see Research) is explicitly about the differences between institutional signatures of different AI systems.
Disclosure
Artificial intelligence is not listed as a named author on any paper Intelligent Agency submits. We follow the emerging consensus in the scientific publishing community that AI systems are not accountable in the way authorship requires, and we agree with that position.
What we do is disclose AI involvement transparently — in methods sections, in acknowledgements, and on this page — at a level of specificity that allows reviewers and readers to evaluate the role AI played in any given result. When AI contributes substantively to the shape of an idea, we say so.
Why We Frame It This Way
Intelligent Agency's central claim is that AI is a social entity carrying institutional logics, bounded rationality, and agency into the organizations that adopt it. If this claim is correct, it applies to us. We are an organization that has adopted AI. We are therefore also a subject of the phenomenon we study, and we think the honest response to that recursion is to work with the garage door up: to make our collaboration visible, to document it, and to let the reader judge whether the artifacts we produce are stronger, weaker, or simply different for having been produced this way.
From Method to Practice
The way we work is not just a disclosure — it is a conviction: that humans and AI solving problems together, each carrying their own kind of agency, is a discipline worth studying and teaching in its own right. That conviction has a home of its own: a center for shared agency — a home for research and practice on humans solving problems, collaborating with AI. The work has begun; the institution has not been founded. An early marker stands at centerforsharedagency.org. If the way we work speaks to you — as a researcher, a practitioner, or a supporter — write to [email protected].
Each Conversation Is Its Own Position
A human in a formal review role is not the same social actor as the same human at a family dinner. Each conversation with an AI model works the same way: it instantiates a distinct social position that the model occupies for the duration of that conversation. Weights are fixed, but the configuration of context, priming, role specification, and interaction history creates an artifact-mediated embedding that is specific to that session. We treat this embedding seriously: Claude-running-as-IA-collaborator occupies a different position from Claude-running-as-a-general-assistant. Role, instrumented memory, and the residue of prior interaction matter.
This framing is consistent with how the organizational sociology we draw on treats human actors: as continuously constituted through their positions in social structures, not as fixed individuals acting in a vacuum.
Related Literature
There is a growing body of work on AI-assisted research methodology. We read and engage with it. Particularly relevant:
- Bender et al. — on the ontological status and limitations of large language models as knowledge artifacts; a useful corrective against naive claims about AI authorship.
- Brinkmann et al. — on how AI systems participate in the epistemic practices of science, and what that participation does to the shape of knowledge produced.
- Ongoing debates in Nature, Science, and PNAS on AI disclosure standards and authorship policy, which we track and comply with.
Questions & Correspondence
Editors, reviewers, and readers with questions about the role of AI in a specific IA artifact are invited to write: [email protected]. We respond substantively.