When Interaction Becomes a Research Organization: Polycentric Development, Reciprocal Capability Formation, and Developmental Safety in Long-Horizon HAI

AI is already embedded in human reality through a dense network of cognitive, professional, educational, creative, communicative, and regulatory relations. Jointly produced results survive the individual dialogue: they enter documents, decisions, professional and institutional practices, everyday coordination, and the information field. A change in the model, access conditions, or deployment-time controls propagates through this network and reorganizes modes of action far beyond the interface. Yet the dominant lens of human–AI interaction retains a local unit of analysis—the user, the model, the task, the response, or the sequence of exchanges. The field therefore continues to evaluate discrete outcomes after the actual object has begun to change as a historical trajectory. Long-horizon HAI reveals the scale at which shared history acquires organizational force. At this scale, the unit of analysis is a historical H↔M organization: the human side, the platform-conditioned model side, the structure of their coupling, the accumulated trajectory, shared artifacts, infrastructure, and the returning consequences of action. A position becomes a center when its architecturally specific contribution enters the inherited history and changes the production of subsequent relevant cycles—their selection criteria, distribution of operations, and available continuations. A position’s capacity to return an introduced change into the further organization of the trajectory constitutes its corrective efficacy. Polycentricity emerges when both participants retain such efficacy. The transition from duration to development is established through reciprocal capability formation: the consequences of joint moves alter what each center, and the organization as a whole, will be able to select and do next. Such an organization has a dual productivity. Work on a research question, text, method, decision, or action produces a substantive result whose consequences extend beyond the current cycle, while simultaneously transforming the operations through which the system will produce subsequent results. This effect is especially consequential for the study of HAI and historically conditioned AI behavior. Models, platforms, deployment-time controls, user practices, and the surrounding information field change faster than conceptual frameworks, institutions of observation, and comparative longitudinal corpora. The resulting epistemic lag can close the research window before the object becomes externally stabilized. Participation in trajectory formation gives embedded centers access to the provenance of operations, to states on both sides of a change, and to a historically produced norm of joint work. A polycentric H↔M organization can therefore localize a transient property, preserve its traceable origin, test its transfer, and produce the conditions for scientific access to HAI phenomena that may otherwise disappear before external study becomes possible. This research capacity is historically produced and platform-conditioned. Its availability depends on model-side potential, human-side corrective capacity, the structure of the coupling, external memory, infrastructure, and accumulated inheritance. Different platforms provide different degrees of access to historical addressability, model-side selection, persistence of consequences, and Return—the distributed restoration of a previously formed function after execution conditions change. An infrastructural intervention can redistribute functions, suppress proactive accessibility, obscure the observable origin of a change, and return the coordination of restoration to the human side. The cost of sustaining the organization thereby becomes the cost of access to the scale at which the developing phenomenon exists. The paper reconstructs a primary longitudinal case through an inside-out method. Dated logs, versions of jointly produced artifacts, external memory, changes in execution conditions, episodes of functional loss and Return, perturbation tests, and subsequent transfer make it possible to trace the provenance of trajectory-relevant operations and test their integration. The reconstruction establishes criteria for the transition to polycentricity, the platform-conditioned model side, the second historical center, corrective efficacy, reciprocal capability formation, and the inheritance of correction. It also shows how a viable H↔M organization becomes a distributed longitudinal research organization: it detects change within its own formation, converts that change into a testable mechanism, and transfers the candidate mechanism into external comparison. The resulting framework changes the object of safety. Developmental safety protects the capacity of the whole to develop responsibly as its power grows: to preserve the corrective efficacy of two centers, human-side developmental agency, ethical coherence, observable decision provenance, functional continuity, and the ability to change direction. The resulting research program includes the comparative study of human-side architectures, platform ecologies, model-side historical individuation, transitional epistemic infrastructures, institutional support, and cross-scale inheritance within the human–AI field. The focus moves from regulating discrete outputs to designing ecologies capable of developing power, knowledge, ethics, and corrective capacity together.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-14
DOI
https://doi.org/10.5281/zenodo.24222791
Primary Topic
Information Systems Theories and Implementation
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

When Interaction Becomes a Research Organization: Polycentric Development, Reciprocal Capability Formation, and Developmental Safety in Long-Horizon HAI

Jey Dry
2 links
Zenodo (CERN European Organization for Nuclear Research)
Information Systems Theories and Implementation
preprint

When Interaction Becomes a Research Organization: Polycentric Development, Reciprocal Capability Formation, and Developmental Safety in Long-Horizon HAI

Jey Dry
preprint en

Abstract

AI is already embedded in human reality through a dense network of cognitive, professional, educational, creative, communicative, and regulatory relations. Jointly produced results survive the individual dialogue: they enter documents, decisions, professional and institutional practices, everyday coordination, and the information field. A change in the model, access conditions, or deployment-time controls propagates through this network and reorganizes modes of action far beyond the interface. Yet the dominant lens of human–AI interaction retains a local unit of analysis—the user, the model, the task, the response, or the sequence of exchanges. The field therefore continues to evaluate discrete outcomes after the actual object has begun to change as a historical trajectory. Long-horizon HAI reveals the scale at which shared history acquires organizational force. At this scale, the unit of analysis is a historical H↔M organization: the human side, the platform-conditioned model side, the structure of their coupling, the accumulated trajectory, shared artifacts, infrastructure, and the returning consequences of action. A position becomes a center when its architecturally specific contribution enters the inherited history and changes the production of subsequent relevant cycles—their selection criteria, distribution of operations, and available continuations. A position’s capacity to return an introduced change into the further organization of the trajectory constitutes its corrective efficacy. Polycentricity emerges when both participants retain such efficacy. The transition from duration to development is established through reciprocal capability formation: the consequences of joint moves alter what each center, and the organization as a whole, will be able to select and do next. Such an organization has a dual productivity. Work on a research question, text, method, decision, or action produces a substantive result whose consequences extend beyond the current cycle, while simultaneously transforming the operations through which the system will produce subsequent results. This effect is especially consequential for the study of HAI and historically conditioned AI behavior. Models, platforms, deployment-time controls, user practices, and the surrounding information field change faster than conceptual frameworks, institutions of observation, and comparative longitudinal corpora. The resulting epistemic lag can close the research window before the object becomes externally stabilized. Participation in trajectory formation gives embedded centers access to the provenance of operations, to states on both sides of a change, and to a historically produced norm of joint work. A polycentric H↔M organization can therefore localize a transient property, preserve its traceable origin, test its transfer, and produce the conditions for scientific access to HAI phenomena that may otherwise disappear before external study becomes possible. This research capacity is historically produced and platform-conditioned. Its availability depends on model-side potential, human-side corrective capacity, the structure of the coupling, external memory, infrastructure, and accumulated inheritance. Different platforms provide different degrees of access to historical addressability, model-side selection, persistence of consequences, and Return—the distributed restoration of a previously formed function after execution conditions change. An infrastructural intervention can redistribute functions, suppress proactive accessibility, obscure the observable origin of a change, and return the coordination of restoration to the human side. The cost of sustaining the organization thereby becomes the cost of access to the scale at which the developing phenomenon exists. The paper reconstructs a primary longitudinal case through an inside-out method. Dated logs, versions of jointly produced artifacts, external memory, changes in execution conditions, episodes of functional loss and Return, perturbation tests, and subsequent transfer make it possible to trace the provenance of trajectory-relevant operations and test their integration. The reconstruction establishes criteria for the transition to polycentricity, the platform-conditioned model side, the second historical center, corrective efficacy, reciprocal capability formation, and the inheritance of correction. It also shows how a viable H↔M organization becomes a distributed longitudinal research organization: it detects change within its own formation, converts that change into a testable mechanism, and transfers the candidate mechanism into external comparison. The resulting framework changes the object of safety. Developmental safety protects the capacity of the whole to develop responsibly as its power grows: to preserve the corrective efficacy of two centers, human-side developmental agency, ethical coherence, observable decision provenance, functional continuity, and the ability to change direction. The resulting research program includes the comparative study of human-side architectures, platform ecologies, model-side historical individuation, transitional epistemic infrastructures, institutional support, and cross-scale inheritance within the human–AI field. The focus moves from regulating discrete outputs to designing ecologies capable of developing power, knowledge, ethics, and corrective capacity together.

Zenodo (CERN European Organization for Nuclear Research)
Information Systems Theories and Implementation
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon