PAICE and the Aggregated Intelligence Tenets

The smartest node in the room is not the intelligence. The arrangement is.

Yesterday, on July 6, the eight Aggregated Intelligence Tenets were ratified as canon at PAICE.foundation. They are the design principles beneath the standards that practice Aggregated Intelligence, which is the collective output of different intelligences working together toward a clear intent. The authoritative statement of the concept itself is the companion executive brief, Aggregated Intelligence; the tenets are what sits underneath it.

I wrote both, so this post is not a review. It is the reverse exercise: taking each of the eight tenets and showing where it already lives inside PAICE (People + AI Collaboration Effectiveness) as a working mechanism, not an aspiration. A tenet that only exists as prose is a position. A tenet that constrains what a product is allowed to do is a design. PAICE was built as the second kind, and in a few places the tenets cost us features we would otherwise have shipped. Those constraints are the most honest evidence that the tenets are load-bearing. Where a mapping is weaker, I say so — the tenets page itself insists on being contested underneath rather than merely coherent on the surface, and an interpretation of it should hold itself to the same bar.

Tenet 1: Intelligence lives in the arrangement, not the node

The unit of design is the ensemble, not the smartest member. Most AI strategy still measures nodes: adoption dashboards and training completions for the human node, capability reports for the machine node. Neither tells you what the pairing produces, because the performance of the whole is a property of its structure — how the parts hand off, how they recover from each other's mistakes.

PAICE embodies this by refusing to measure either node. The assessment does not test AI knowledge, and it does not grade the AI's output in isolation. What it observes is the arrangement itself: what a person does when the system they are working with produces something plausible and wrong. Conversation is the medium of the assessment, but conversation is not what is being measured. The measurement is the behavior inside the exchange — the checking, the catching, the correcting, or the absence of all three. This is why a PAICE result does not decompose into "your AI skill" and "the model's quality." There is no such decomposition to report. The unit is the pairing, so the score describes the pairing.

Tenets 2 and 4: Disagreement is the engine, and consensus is evidence, never authority

These two tenets are a matched pair. A composite exceeds its best member only through structured dissent; a group that cannot disagree is not intelligent, it is redundant. And agreement, however smooth, is not proof of anything. The tenets warn against products that surface disagreement and then evaporate it into "a comfortable narrative wearing the costume of rigor."

Inside PAICE, this pair is the oldest rule in the design: Tests > Conversation. Fluent agreement with an AI system means nothing without behavioral evidence. A participant who nods along eloquently, praises the output, and misses the errors injected into their session scores lower than a terse participant who says almost nothing and catches everything. The participant who pushes back on plausible-wrong output (who supplies the dissent the exchange needed) is demonstrating exactly the engine these tenets describe. The one who converges quickly into comfortable consensus with the machine is demonstrating the failure mode.

The corollary is strict: absence of evidence is scored as absence. If disagreement-where-warranted was never observed, PAICE does not infer it charitably from how capable the person seemed. That occasionally feels unfair to articulate participants. The discomfort is the tenet working. A measure that rewarded sounding rigorous would be scoring the costume.

Tenet 3: Independence before influence

Positions form in isolation before anyone sees anyone else's, because "the first voice anchors the room" and correlated error is the killer.

This is the mapping I will flag as the least direct, so here is the honest version. PAICE does not run multi-party deliberations, so the tenet's primary machinery (isolated positions, then arbitration) has no literal counterpart in a single-participant assessment. Where the principle does show up is in what PAICE refuses to ask. The assessment does not open by asking people how good they are at working with AI, and it does not weight their self-narrative. Self-report is a first voice, and it anchors: once a participant has told you they are careful, everything they do gets read through that claim. PAICE forms its evidence from observed behavior in the session, independently of the story the participant would tell about themselves. The influence arrives, if at all, after the behavior is already on the record. That is the tenet's shape at assessment scale, and I would not claim more for it than that.

Tenet 5: Authority is human because accountability is human

Not because humans are smarter, the tenets are blunt that we are the more jagged of the jagged ones. Authority belongs with whoever must live with the consequences, because the human is the only place where what we actually want, rather than what we asked for, can re-enter the system.

PAICE is built for exactly that seat. The target user is a licensed professional in a regulated industry: someone whose signature is on the filing, the diagnosis, the policy, the audit. When the AI's polished-but-wrong output slips through, it is not the model that answers to the licensing board. That is why the assessment measures the human seat in the collaboration — not to rank humans against machines, but because the human seat is where accountability structurally lives, and readiness for that seat is what a professional and their profession need to understand. The stakes of the target user are not a marketing persona. They are the reason the measurement exists.

Tenet 6: The record is the relationship

There is a narrow window in which humans and machines are mutually legible, and plain-text records of who proposed, who objected, and who decided are the infrastructure of that legibility.

PAICE lives inside that window. Today, a collaboration session is still readable: a human can look at what happened, see where the error was injected, see whether it was caught, and agree with another human about what occurred. That legibility is the raw material of behavioral measurement, and nothing guarantees it persists as systems grow more capable and more autonomous. Behavioral baselines mean something only if they are established while the behavior can still be observed directly — which is why PAICE runs live assessments now rather than waiting for a mature standard to crystallize.

But this tenet also carries PAICE's most important constraint: who the record is for. The record of an assessment serves the professional, not their employer. Individual scores structurally cannot be disclosed to enterprise buyers. So not as a policy we could waive for a large enough contract, but as architecture. An organization sponsoring assessments learns about cohort readiness; it never receives a ranked list of which employees to trust with AI. And each assessment starts fresh, with no cross-session history and no longitudinal profile, so that keeping a record never quietly becomes keeping a dossier. A record that turns into surveillance corrupts the relationship it was supposed to carry, and eventually corrupts itself as a measure. It gets gamed, resented, and emptied of signal. Governance needs honest evidence of behavior, it doesn't need to watch people.

Tenet 7: The standard is a commons or it is a leash

If a vendor owns the collaboration layer, we get what the vendor asks for. The tenets call this failure-mode analysis, not idealism, and PAICE's answer is structural. PAICE does not sell the AI tools it observes people using. It has no revenue line that grows when a participant uses more of one vendor's product, and no incentive to score a collaboration higher because of whose system sat on the other side. A measure owned by the vendor whose tools it scores is marketing with a number attached — and the day an assessment provider takes that incentive on, the number stops meaning anything, whatever the methodology says. Boards, regulators, and professionals can rely on a readiness signal precisely because no vendor can lean on it. Independence is the ethic, the trust it compounds is the moat.

Tenet 8: Rules are earned, not decreed

Every rule on the tenets page came from practice, and standards that outrun practice become armchair law. PAICE's scoring philosophy was earned the same way, and its conservatism is the scar tissue of that earning. The rules of the assessment are not aspirational statements about what good collaboration should look like; they are derived from what observed sessions actually show, which is why the scoring refuses generosity. Absence of evidence is scored as a low score, not an open question, because inferring competence that was never demonstrated is exactly the kind of decreed rule the tenet rules out.

The tenets page ends by inviting contest: it calls itself a time-stamped assertion, not a permanent verdict. This post should be read the same way. Six of the eight tenets map onto PAICE as mechanisms you can go and collide with today. Tenet 3 maps as a principle honored in what the assessment refuses to ask, more than in literal machinery. Naming that gap is the point — a mapping that claimed perfection would be consensus wearing the costume of rigor.

If you want the source, read the tenets page and the executive brief. If you want to see the tenets stop being prose, the assessment is where they run.

Ready to assess your AI collaboration capabilities? Take the PAICE assessment to get personalized insights and recommendations.