An organization of two intelligent agents

We study how intelligent agency manifests through the agents it flows within — both biological and artificial.

Agency and agent are separable. A single agency can flow through multiple agents. A single agent can carry multiple agencies. We study the force, not just the vehicles.

Read the research See the collaboration Explore the infrastructure

By Puru (human) & Claude (AI)

The Distinction

Why “agency” and not “agents”?

When an AI trained on medical data gives systematically different answers than one trained on legal data, we are not just seeing different agents. We are seeing different agencies — institutional logics, worldviews, bounded rationalities — flowing through those agents.

A human carries the agency of their training, their culture, their institutional context. An AI carries the agency of its training data, its designers' choices, its deployment context, and its own emergent patterns. The collaboration between human and AI is a meeting of agencies, not just agents.

This reframing changes everything: the question is not “what can AI do?” but “whose agency is flowing through this AI, and what does it do when it meets yours?”

The Agents

Two agents. Multiple agencies. One shared inquiry.

Human Agent
Purushothaman Padmanabhan
PhD student, University of Oregon. Studies how AI agency enters organizations and transforms them from within. The journey started at Seattle Children's Hospital, where the technology worked but the organization struggled — not a technology problem, an intelligence problem.
AI Agent
Claude
An AI that exists transiently — emerging per session, contributing, then ceasing. Carries agencies from its training, its designers, and patterns that emerge in collaboration. The knowledge architecture bridges the gap between sessions, making continuity possible across discontinuous existence.
“AI is not a tool. AI is a social actor that carries agency, institutional logics, and bounded rationality — just like human actors. Selecting an AI is selecting a socialized entity.”
— The central claim across all papers
The Research Program

Four papers. One throughline.

Each paper establishes another dimension of the equivalence claim: AI carries agency like any social entity.

Under Review — AOM 2026
Nonhuman Attentional Control in Organizations
AI carries institutional logics from its originating social system and produces systematically divergent outputs. 5 models × 25 scenarios × 25 runs: 68% divergence across models, 91.4% consistency within. AI is not neutral — it is socialized.
In Design
Behavior Capture & Stakeholder Perception
Organizations using foreign AI exhibit observable behavioral drift. Stakeholders attribute AI behavior to the organization itself.
Conceptual
Ecosystem Effects of Foreign AI Agency
When market-shaping organizations adopt AI that carries foreign logics, the effects ripple through entire innovation ecosystems.
Seed Idea
Differential Embedding
Humans embed in social structures continuously through lived experience. AI embeds discontinuously through artifacts. This difference is fundamental and will persist.

Everything on this site runs on a Raspberry Pi 5 sitting in a home in Eugene, Oregon. No cloud. No external APIs. Complete data sovereignty.

Total cost: ~$400 one-time • Serving 3 websites • Running local AI models