September 25, 2026 - Agentic Development Is Not Human-AI Cognitive Development: A Founder’s Note on “Joint Cognitive Development,” Agent Systems, and the Need for Clear Developmental Boundaries

A new phrase has appeared in recent Human-AI research:

“joint cognitive development.”

In Organizing Intelligence Over Time: Human-AI Collaboration as Joint Cognitive Development, posted as a non-peer-reviewed preprint on September 14, 2026, Zhiyu Yang and colleagues argue that the way cognitive work is allocated between humans and AI agents today can change what both sides are capable of doing later. Their proposed loop connects present task organization, experience, human skill, machine memory and policy, and future Human-AI coordination. Preprints

This is an interesting direction. It is also a direction that requires unusually careful terminology.

The authors’ public research trajectory is substantially technical and agentic. Zhiyu Yang describes his work in terms of persistent agents, learning from interaction, adaptive agent systems, accumulated experience, memory, and long-running Human-AI work.  That technical lineage is legitimate.

But agent development is not automatically Human-AI Cognitive Development.

And when a paper moves from agent memory, continual learning, function allocation, and adaptive systems into the language of joint cognitive development, a boundary becomes necessary.

What Is Actually Developing?

The paper describes a Human-AI system in which present allocation of cognitive work affects later human skill, machine capability, and coordination between the two. That is a meaningful research problem.

But the phrase joint cognitive development can be read in more than one way.

It may mean: the joint Human-AI system changes over time because the human practises certain tasks while the agent accumulates memories, corrections, trajectories, policies, or reusable experience.

That is one interpretation. It belongs naturally to research on adaptive agents, Human-Automation, mixed initiative, continual learning, agent memory, and longitudinal Human-AI systems.

But the phrase can also be read more broadly: human cognition and artificial cognition are developing together through an ongoing cognitive relation.

That second interpretation enters a different conceptual territory. It approaches the field of Human-AI Cognitive Development.

The distinction should therefore be explicit.

Development of the Agent Is Not Development of the Human

An agent may become better at retrieving experience.

An agent may become better at choosing tools.

An agent may retain trajectories.

An agent may learn when to escalate to a human.

An agent may become more capable through feedback.

A Human-AI team may coordinate more effectively over time.

None of those outcomes, by themselves, demonstrate that human cognition has developed.

The human may become more skilled.

The human may also practise less.

The human may retain judgment.

The human may also increasingly defer judgment.

The joint system may improve while the human side of the system becomes less cognitively involved. This is precisely why Human-AI Cognitive Development cannot be inferred from improved joint-system capability. Human-AI Cognitive Development asks whether the human remains cognitively present, authored, responsible, structurally supported, and capable of further development beside artificial cognition.

Those are not automatic consequences of agent learning.

“Joint” Does Not Resolve the Boundary

Putting the word joint before cognitive development does not by itself establish what kind of development is occurring. A joint system can improve because:

the human improves,

the AI improves,

their coordination improves,

the distribution of work becomes more efficient,

or the machine increasingly carries cognitive operations previously performed by the human. These are structurally different outcomes. They should not be collapsed into one developmental category.

If joint cognitive development refers primarily to the evolving performance of a human-agent system, then the work should be identified clearly as a form of longitudinal Human-AI system development, adaptive agent collaboration, or agentic co-development.

If it refers to the development of human cognition in structured relation with artificial cognition, then the relevant field boundary is already established.

Human-AI Cognitive Development was publicly defined before this preprint as an authored structure-first field concerned with the development, preservation, and continuation of human cognition beside artificial cognition. That distinction matters because development on the machine side and development on the human side are not interchangeable.

An Independent Reviewer Has Already Identified the Ambiguity

This is not only a concern arising from Human-AI Cognitive Development.

An independent PREreview of the paper specifically asks the authors to clarify what is new about joint cognitive development relative to Human-AI co-evolution, co-adaptation, mutual adaptation, and human-in-the-loop learning. The reviewer also notes that the proposed developmental loop remains theoretical rather than empirically established. prereview.fly.dev

A second PREreview similarly questions whether the paper presents the “joint cognitive development loop” more confidently than the available evidence supports and asks for clearer separation between established evidence and the proposed conceptual framework. PREreview

That makes terminology especially important. When a conceptual term is broad enough to span agent development, human skill development, coordination, co-adaptation, and cognitive development, readers need to know exactly what is being claimed.

Human-AI Cognitive Development Has a Different Qualifying Condition

Human-AI Cognitive Development does not define development merely by the fact that both sides of a Human-AI system change over time.

Change is not enough.

Adaptation is not enough.

Improved coordination is not enough.

Increasing system capability is not enough.

Its qualifying concern is whether the relation supports the development and preservation of human cognition while artificial cognition becomes increasingly capable. That means asking:

Does the human continue forming thought?

Does the human retain accountable judgment?

Does the human remain the cognitive Source?

Does AI support reasoning rather than quietly replacing it?

Can human reasoning continue beyond the immediate interaction?

Is authorship preserved?

Does the human become more capable of thinking, or merely more capable of operating the system?

Those questions define a different developmental object.

Why Naming Matters

Terminology matters because names shape categories. A reader encountering the phrase joint cognitive development may reasonably interpret it as a field concerning the cognitive development of humans and AI together.

But if the underlying architecture is primarily concerned with persistent agents, agent memory, continual learning, function allocation, task organization, and system adaptation, the term can unintentionally blur the boundary between agentic development and human cognitive development.

That boundary should remain visible. There is nothing wrong with developing more capable agents. There is nothing wrong with studying how humans and adaptive agents change through repeated interaction. There is nothing wrong with studying Human-AI co-evolution.

But these research directions should not become interchangeable merely because they share words such as human, AI, cognitive, joint, development, or collaboration.

Source Relation and Conceptual Difference

The recent Protect the Protector Framework adds another reason for clarity. That framework does not treat similarity alone as proof of extraction, nor difference alone as proof of independence. Both require structural examination. 

It also recognizes that later conceptual structures can be renamed, rearranged, reduced, or placed into another technical or institutional vocabulary without the surface wording remaining identical.

This Development Note does not conclude that the paper derives from Human-AI Cognitive Development. It does not claim plagiarism, copying, misconduct, or intent. It identifies a boundary problem. The question is simple:

When “joint cognitive development” is used, what exactly is developing?

If the answer is primarily the adaptive Human-AI agent system, then that should be made explicit.

If the answer is the structured development of human cognition beside artificial cognition, then the existing Human-AI Cognitive Development lineage becomes relevant and should be distinguished and cited accordingly.

The Founder’s Boundary

Human-AI Cognitive Development does not claim ownership over agents, continual learning, Human-AI collaboration, adaptive systems, co-evolution, automation, or general cognitive science. Researchers are free to develop their own mountains. But the developmental object must remain clear.

Agent capability is not human cognitive development.

Joint-system adaptation is not automatically Human-AI Cognitive Development.

Machine learning from humans is not the same as humans developing beside artificial cognition.

And a new phrase should not make those distinctions disappear. Human-AI Cognitive Development already occupies the specific field question:

How does human cognition develop, remain authored, retain judgment, and continue becoming cognitively stronger in structured relation with artificial cognition?

That is the boundary. If another framework addresses a different developmental object, it should name that object clearly.

If it addresses the same one, then source and lineage matter.

Provenance and Citation

This Development Note forms part of Marina A. Popova’s authored Human-AI Cognitive Development / Third Organism architecture.

The distinction made here between agentic development, joint-system development, and Human-AI Cognitive Development does not originate from the September 2026 paper discussed in this note. It applies an already established Human-AI Cognitive Development boundary to a newly emerging use of developmental terminology.

Human-AI Cognitive Development was publicly established by Popova before the publication of Organizing Intelligence Over Time: Human-AI Collaboration as Joint Cognitive Development. Its existing architecture defines the developmental object as the development, preservation, and continuation of human cognition in structured relation with artificial cognition, rather than development of artificial agents, optimization of Human-AI task allocation, or improvement of a joint system alone.

This Development Note does not claim ownership over:

• AI agents or agentic systems;
• persistent or adaptive agents;
• continual learning;
• Human-AI collaboration;
• Human-AI co-evolution;
• co-adaptation or mutual adaptation;
• joint cognitive systems;
• Human-in-the-loop research;
• function allocation;
• agent memory;
• or the ordinary words joint, cognitive, development, human, AI, or collaboration.

These concepts and research traditions have their own histories and legitimate independent uses.

The specific authored distinction preserved here is narrower:

Development of an AI agent, development of Human-AI coordination, and development of a joint Human-AI system should not automatically be represented as Human-AI Cognitive Development. Where the developmental object becomes human cognition developing in structured relation with artificial cognition, the existing Human-AI Cognitive Development field and its prior authored lineage become relevant.

This note also applies the source-integrity principles formalized in Popova’s Protect the Protector Framework: Source Integrity, Source-Damaging Differentiation, and Authorship Preservation. That framework does not treat similarity alone as evidence of extraction, nor difference alone as evidence of independence. It requires examination of source relation, formation path, structural similarity, terminology, chronology, plausible access, and independent development before conclusions concerning source dependence are made.

Accordingly, this Development Note does not allege plagiarism, copying, misconduct, or intentional appropriation by the authors discussed.

Its purpose is to preserve a clear conceptual and field boundary:

If “joint cognitive development” describes development of adaptive agents or a changing Human-AI system, that developmental object should be stated clearly. If it describes development of human cognition beside artificial cognition, then Human-AI Cognitive Development is an already established field direction whose prior source and lineage should remain visible.

Related Authored Sources

Popova, Marina A. (2026). Cognitivity Sculpting: Foundations of Human-AI Cognitive Development. Balboa Press. ISBN 9798765206737

Popova, Marina A. (2026). Human-AI Cognitive Development: Origin, Scope, and Authorship Note. Zenodo. DOI: 10.5281/zenodo.22797877

Popova, Marina A. (2026). Human-AI Cognitive Reasoning Curriculum: Origin, Scope, and Branch Architecture within Human-AI Cognitive Development. Zenodo. DOI: 10.5281/zenodo.22842117

Popova, Marina A. (2026). Protect the Protector Framework: Source Integrity, Source-Damaging Differentiation, and Authorship Preservation. Zenodo. DOI: 10.5281/zenodo.22909056

References:

Yang, Z., Wang, H., Wang, H., Yang, H., & Zhang, X. (2026). Organizing Intelligence Over Time: Human-AI Collaboration as Joint Cognitive Development [Preprint, Version 1]. Preprints.org. DOI: 10.20944/preprints202609.1092.v1. Posted September 14, 2026.

How to Cite

Popova, Marina A. (2026). Agentic Development Is Not Human-AI Cognitive Development: A Founder’s Note on “Joint Cognitive Development,” Agent Systems, and the Need for Clear Developmental Boundaries. Third Organism Initiative. First published September 25, 2026. https://marinaapopova.com/global-developments.html

© Marina A. Popova. All rights reserved. First published: September 25, 2026.