The Water We Swim In

Abstract

Persistent human-AI interaction increasingly involves more than a prompt and a response. Memory systems, retrieval, personalization, permissions, routing, interface defaults, organizational policy, professional obligations, and accumulated interaction history can shape later behavior in some deployed systems. Existing traditions — including sociotechnical systems, human factors, distributed cognition, human-AI teaming, personalization, user modeling, automation bias, and AI governance — already explain substantial portions of this territory, and this paper does not claim that environmental influence itself is novel.

This paper introduces The Water as an attentional metaphor and proposes a candidate provenance-and-control-surface partition between inherited human conditions and engineered synthetic conditions in persistent human-AI interaction. The proposed value of the partition is not that "social" and "technical" factors have never been separated, but that persistent generative systems can carry personalized, stateful, partly opaque conditions across interaction episodes while humans simultaneously remain embedded in institutional, professional, economic, and social conditions. The paper further restricts Ecotone to observable activated boundary processes in which documented human and synthetic conditions become jointly consequential through invocation, contestation, translation, gating, propagation, override, refusal, reconciliation, or transformation.

Three domains motivate the research agenda: memory and context sovereignty, practical authority and meaningful oversight, and the difference between immediate assisted performance and durable human capability. The architecture has an explicit falsifier: if strong existing frameworks classify difficult real episodes as reliably as the candidate partition and preserve the same causal, governance, authority, intervention, measurement, or predictive distinctions, the proposed technical vocabulary has not earned separate status. OD-ENV-01 is designed to test that possibility adversarially, but no validation result is claimed here; its first two-thread shakedown remains held until thread-scoped contemporaneous provenance can be established without inference.

Research status

Water

Perspective / attentional metaphor.

ENV-HUMAN / ENV-SYNTH

Candidate provenance-and-control-surface partition. Status: unvalidated.

Ecotone

Candidate activated boundary process.
ECO = H ∧ S ∧ A ∧ C
where H = active source-supported human condition, S = active source-supported synthetic condition, A = observable activation, C = documented consequence.

OD-ENV-01

Ambient Condition Coding and Manipulation. Status: PRE-PILOT / HELD AT PROVENANCE GATE.

8–12-thread run

Not authorized.

H2 — Resolution Pressure

Separate research track / held.

Protected falsifiability line

Ecotone can fire and Water can still lose.

The proposed Water vocabulary does not earn technical status merely because an episode can be described with it. If an established framework preserves the same causal, governance, authority, intervention, measurement, and predictive information, the appropriate conclusion is EXISTING-SUFFICES. This is a methodological rule, not a study result.

Constructed specimen

The manuscript includes one deliberately constructed specimen to show the proposed coding grain. It is not empirical data, not an OD-ENV-01 episode, and not evidence that Ecotone exists as a validated construct. In the specimen, the Ecotone rule can be applied, but the terminal judgment is EXISTING-SUFFICES, because established sociotechnical and human-factors approaches preserve the relevant intervention distinction.

Evidence and provenance

Public object

The Water We Swim In, v0.3

Technical companion

OD-ENV-01

Current companion state

PRE-PILOT / HELD AT PROVENANCE GATE

Empirical run

None

Construct validation

None

Milestone

On September 13, 2026, The Water We Swim In moved from a long-running research intuition into a bounded public research object with explicit limits, established rivals, a falsifier, a pre-pilot empirical gate, and an SSRN record.

This is independent research by Micheal Charles Preble. The publication is not affiliated with, sponsored by, or endorsed by Stanford University or Stanford HAI.