HI: Axioms and the Human Continuity Hypothesis

HI is not another name for human intelligence. It is a particular form of intelligence whose substrate is human intelligence: intelligence arising from humans in relationship.

The axioms that follow are not simply philosophical preferences. They are proposed evolutionary constraints on the conditions under which higher-order intelligence can emerge without severing continuity with its constituent humans.

The only successful precedents available to us for transitions to higher-order living organisation – the emergence of eukaryotic cells and of multicellular organisms – did not eliminate their constituent units. They organised them into new levels of agency while preserving the conditions that allowed the whole to remain viable.

HI asks whether the same principle can guide the emergence of higher-order human intelligence.

Its defining conditions can be stated as five axioms.

1. Human substrate

HI arises from humans in relation.

Humans are its constitutive units. Other systems may connect, amplify, translate, remember and coordinate, but they do not replace the human substrate from which the intelligence emerges.

Connection – which AI can dramatically expand – is tissue between humans, not intelligence above them.

2. Reality coupling

The human constituents of HI remain independently answerable to reality.

The system cannot ultimately validate itself. Its human constituents encounter patients, experiments, customers, materials, consequences and other humans – a reality that does not have to agree with what the network believes.

The humans answer to reality.
Reality answers to nothing.

3. Plurality

HI preserves independently situated human perspectives.

Multiplicity alone is insufficient. A thousand repetitions of the same judgement do not constitute a thousand independent observations.

Difference, disagreement and diversity of experience are not noise to be eliminated. They are part of the computational substrate of HI.

Quorum is not trending.

4. Plasticity

Encounter changes the network.

Humans change through encounters, and those changes in turn reshape their connections. Relationships form and weaken. Authority shifts. Knowledge is revised. What the network encounters today changes both its humans and the structure through which they encounter one another tomorrow.

HI therefore remembers not merely by storing information, but by changing itself.

Γ(t+1) ≠ Γ(t)

5. Answerable connection

The connections answer to the humans.

Whatever connects humans – conversation, institutions, markets, media or technology – shapes what propagates, what is amplified and what is remembered.

HI requires that these connective layers remain perceptible to their human constituents, revisable by them, and never become their only channel to reality.

This completes the chain:

The connections answer to the humans.
The humans answer to reality.
Reality answers to nothing.

These axioms do not guarantee collective intelligence. They create the conditions under which higher-order intelligence may emerge.

Human substrate + reality coupling + plurality + plasticity + answerable connection → possibility of emergent collective intelligence.

AI appears in the axioms only as a form of connection, never as a condition. HI can exist without AI. Conversation, language, institutions and communities have always connected human intelligence. AI changes the possible scale, richness and persistence of those connections.

AI can expand the connectivity and capability of the human network without replacing its human substrate.

The Human Continuity Hypothesis

The axioms lead to a deeper hypothesis:

If higher-order intelligence emerges while these conditions are maintained, its emergent purposes remain human-sourced.

There is no ontological break in the lineage of purpose.

This does not mean those purposes will necessarily be good, benevolent or universally shared. The claim is narrower: they emerge through human purposes, tensions, values, experiences and encounters with reality. They remain humanly contestable and humanly answerable.

Continuity is therefore not guaranteed simply because humans are present. It depends on maintaining the conditions of HI. A connective layer can increasingly determine what humans perceive, amplify and value; plurality can collapse into imitation; reality coupling can weaken; the network can cease to answer to its human nodes.

Human substrate establishes the lineage. The other axioms make it intelligent and keep it continuous.

HI is therefore not simply an architecture. It is a relationship that must be maintained.

A practical consequence

The Human Continuity Hypothesis suggests a practical design principle:

Use AI to increase the connectivity and capability of the human network, not to replace the network with an artificial one.

AI may perform enormous amounts of cognitive work – searching, reasoning, synthesizing, translating, simulating, remembering and coordinating. The relevant boundary is not how much cognition is delegated to AI, but whether humans remain the reality-coupled, plural and mutually connected nodes through which learning, judgement and purpose enter and change the system.

The aim is not to keep humans in the loop of an increasingly autonomous artificial intelligence. It is to keep humans as the nodes of the loop itself – and the connective layer answerable to them.

This differs from the problem normally described as AI alignment.

Alignment begins with two potentially distinct sources of purpose – human and artificial – and asks how the latter can remain aligned with the former.

HI begins somewhere else. It seeks to preserve the human source of higher-order purpose while using AI to expand what the human network can perceive, understand and do.

Human-aligned describes a relationship between artificial and human purpose. Human-sourced describes the origin of purpose.

The alignment problem may, at least in part, be a substrate problem.

HI therefore asks a prior question:

Why let the authorship of purpose migrate to artificial intelligence at all?