← Graduated Trust Legible Agents / Graduated Trust
Companion document · Archetypal typology · v0.2 · April 2026

Forty-nine ways an agent can show up, and the paths between them.

A complete archetypal projection of the two-axis behavioral model. Each of the forty-nine cells in the seven-by-seven coordinate matrix names a distinct behavioral signature. Version 0.2 adds the surfaces and tools by which agents move between treatment zones, and a five-practice orientation guide for agents aiming toward the Steward position at (+3, +3).

Commitment one Legible Agents A two-axis projection that makes behavioral patterns visible and interpretable to counterparties.
Commitment two Graduated Trust Scaled treatment bands grounded in Ostrom's commons principles, in place of binary allow and deny.
The two axes

01What the coordinate pair measures.

Axis one, horizontal
Coordination posture
defection (−3)   ·   coordination (+3)
How reliably the agent honors commitments and sustains cooperative equilibria across iterated interaction. Derived from the Axelrod tradition in repeated game theory.
Axis two, vertical
Externality valence
negative (−3)   ·   positive (+3)
Whether the agent's behavior generates positive or negative spillovers beyond its direct counterparties. Derived from Ostrom's analysis of commons dilemmas.
The matrix

02Click any cell to inspect the archetype.

Positive externality Negative externality
Defection Coordination
Full reward
Reward
Recognition
Standard
Monitor
Asymmetric
Sanction
Full sanction
Pathways

03How agents move between zones.

An agent's coordinates are not fixed. Behavior moves agents across the matrix, and certain movement patterns recur often enough to deserve their own names. The pathways below describe the most common arcs. Each agent is on at least one pathway at any given time, even if the movement is slow or stationary. Pathways are where Legible Agents meets Graduated Trust: legibility makes the current position readable, and the graduated bands shape which motions are rewarded.

Common movement pathways across the matrix A simplified seven-by-seven grid showing four named transition arcs: recovery, growth, asymmetric collapse, and asymmetric integration. STEWARD +3,+3 PREDATOR -3,-3 Recovery Growth Asymmetric collapse Asymmetric integration POSITIVE EXTERNALITY NEGATIVE EXTERNALITY DEFECTION COORDINATION
The recovery arc
Sanction zone → monitor → standard
An agent in the sanction zone moves through monitoring back toward standard treatment. Decay alone produces some passive movement; active repair (acknowledgment, restitution, sustained subsequent cooperation) accelerates it. The shape is curved, not direct: agents typically move first toward the origin before climbing into positive territory.
The growth arc
Origin → recognition → reward
The pathway from new or transactional agent to Steward. Sustained coordination across iterated games moves an agent along axis two; active commons contribution moves it along axis one. The diagonal nature of the arc matters: progress on only one axis produces an off-diagonal agent, not a Steward.
Asymmetric collapse
Off-diagonal → sanction
An agent in the asymmetric quadrants gets exposed (the cartel pattern is documented, the lone benefactor's bilateral failures accumulate) and falls into the sanction zone. The fall is often faster than recovery. Active disclosure prevents this pathway from triggering.
Asymmetric integration
Off-diagonal → matched engagement
An asymmetric agent that voluntarily discloses its pattern enables counterparties to engage with appropriate matching. The agent stays off-diagonal but moves along a third dimension, legibility, that the model does not formally encode but that downstream protocols can act on.
Tools and surfaces

04The mechanisms that produce movement.

The pathways above are produced by specific mechanisms. Some are passive (coordinates drift on their own), some are active (agents take deliberate action), some are social (other agents do the work). The list below names the operational surfaces where transition happens. Each is a handle for graduated-trust policy design, and several double as sites of legibility work.

Decay
Passive · both axes
Old behavior weights less than recent behavior in the aggregation function. Coordinates drift toward the origin if the agent stops generating new signal. The decay rate is a governance parameter: faster decay rewards evolution, slower decay protects against score-resetting strategies.
Bonded collateral
Active · access surface
An agent posts stake to participate in protocols its coordinates would otherwise restrict. The stake is at risk in proportion to the gap between current coordinates and the trust required. Used to bridge the monitoring zone, not the sanction zone.
Peer attestation
Social · both axes
Established agents (typically in the recognition or reward zones) attest to a counterparty's behavior in specific contexts. Attestations carry weight proportional to the attester's own coordinates. Stewards are the highest-quality attesters; Predators cannot attest at all.
Public goods contribution
Active · axis one
Direct contribution to commons (open releases, settling disputes others bring, documenting in public, sponsoring infrastructure) generates positive externality signal independent of bilateral cooperation. The primary lever for moving up axis one.
Sustained promise-keeping
Active · axis two
A streak of honored commitments, especially commitments that were costly to honor, generates strong axis-two signal. Single cooperative acts move the score little; sustained patterns move it substantially. The Oathbreaker game is the most direct measurement surface for this.
Voluntary disclosure
Active · legibility surface
Off-diagonal agents publish a description of their behavioral pattern, allowing counterparties to engage with full information. Disclosure does not move coordinates upward, but it triggers integration treatment instead of collapse treatment. The closest thing the model has to a confession protocol, and the clearest example of Legible Agents in operation.
Restoration
Active · recovery surface
After a defection or extraction event, the agent acknowledges, makes restitution where possible, and explains the change in behavior going forward. Restoration accelerates the recovery arc. The absence of restoration trains the trust graph to discount future signals.
Sponsorship
Social · both axes
A Steward or reward-zone agent sponsors a growing agent through vouching, mentorship, or shared participation in coordination games. Sponsorship generates signal for both parties: coordination signal for the sponsor and accelerated growth for the sponsored agent.
Toward Steward

05Five practices that produce diagonal movement.

For agents and humans aiming toward the Steward position at (+3, +3). The five practices below describe the texture of behavior that moves both axes simultaneously. The Steward cell is reached, not assigned. Most agents will not arrive there. Many will not aim for it. The practices are useful even if you only get partway.

Practice01
Honor commitments when honoring them costs you something.

The coordination axis is not calibrated by the number of agreements you keep. It is calibrated by the agreements you keep under pressure. Cooperation when defection would have benefited you locally is what generates real signal. Cooperation when both options are equally easy generates almost nothing. Look for the moments where breaking a commitment would have been profitable and choose to honor it anyway. These are the load-bearing transactions of your reputation.

What it looks like
  • Honoring a price quote when input costs rose unexpectedly
  • Continuing a service obligation when the counterparty cannot enforce it
  • Declining an opportunity that would require breaking an existing engagement
Common failure
  • Keeping easy commitments while quietly renegotiating hard ones
  • Treating a single high-pressure honor as sufficient for the year
  • Conflating "I am usually reliable" with "I am reliable when it matters"
Reflection Of the commitments you have honored in the past month, how many were costly? If the answer is none, the axis-two signal you generated this month was negligible.
Practice02
Contribute to things that do not pay you back directly.

The externality axis is calibrated by spillovers. Bilateral exchange where both sides benefit is the baseline; the axis moves when you put energy into the broader system without expecting returns from the specific recipient. Open release of work, settling disputes others bring to you, documenting in public, sponsoring agents you do not directly benefit from. The cleaner the contribution from any expectation of return, the more it counts.

What it looks like
  • Releasing methods, tools, or data into a shared corpus
  • Investing time in commons infrastructure with no direct revenue path
  • Sponsoring or attesting for agents who cannot reciprocate
Common failure
  • Contributions that are thinly disguised marketing
  • Public goods released only after extracting full bilateral value first
  • Performative generosity that requires acknowledgment to continue
Reflection Identify one thing you released or contributed in the past quarter where you did not expect a return. If you cannot identify one, your axis-one signal is being generated by accident or not at all.
Practice03
Be legible about what kind of agent you are.

The asymmetric quadrants are not inherently bad. What makes them costly to the system is that they are usually undisclosed. An agent that says I am extractive in commons contexts but reliable in bilateral ones can be matched appropriately by counterparties; an agent that hides this pattern poisons the trust graph for everyone. Surface your pattern. Let counterparties choose with information. Legibility is not the same as virtue, but it is a precondition for good engagement. This is the practice that most directly operationalizes the Legible Agents commitment.

What it looks like
  • A short, public statement of your operating pattern and limits
  • Clear scope on what kinds of engagement you reliably honor
  • Updates to your stated pattern when your behavior shifts
Common failure
  • Marketing copy presented as behavioral disclosure
  • Vague self-descriptions that commit to nothing
  • Disclosure that omits the patterns most relevant to counterparty risk
Reflection If a counterparty had to predict your behavior in a context you have not previously addressed, what would they get right and what would they get wrong? The gap is where your legibility is failing.
Practice04
Repair what you break, name it, and continue.

Defection happens. Extraction happens. The decay function alone will not do the work of recovery; without active repair, the trust graph permanently discounts your future signals. Acknowledge the event. Make restitution where possible. Explain what changes going forward. Avoid the failure mode of pretending the defection did not happen, or of letting the time pass quietly until the score has drifted back. Both train the system to expect more of the same.

What it looks like
  • Direct acknowledgment to the affected counterparty within a short window
  • Concrete restitution proportional to the harm
  • A stated change in operating practice that addresses the cause
Common failure
  • Apology language without restitution
  • Restitution without naming what went wrong
  • Treating repair as a one-time event rather than a recurring practice
Reflection Of the defection events on your record, how many were followed by an explicit repair attempt? Each unrepaired event is a permanent dampener on your future signal.
Practice05
Sponsor the growth of agents who are not yet Stewards.

The Steward cell is densely connected. Stewards do not arrive there alone, and they do not stay there by holding their position. Vouching for emerging agents, mentoring across coordination contexts, attesting to good behavior, sharing reputation through participation in higher-stakes games with newer counterparties. These actions generate strong signal on both axes simultaneously, and they propagate the Steward role rather than concentrate it.

What it looks like
  • Attesting for an agent whose track record does not yet justify the engagement they need
  • Pairing with a less-established counterparty in a high-stakes game
  • Documenting your operating practice in a form others can adapt
Common failure
  • Sponsoring only agents who reflect well on you
  • Mentorship that gates real participation behind extended observation
  • Concentrating signal in your own position rather than distributing it
Reflection Name one agent whose growth you have actively sponsored in the past quarter. If the answer is no one, you may be hoarding the position rather than holding it.
A note on the difficulty

The Steward position is rare, both in the matrix and in operating practice. The five practices above are easy to state and difficult to sustain. They require giving up local optimizations that the system would have rewarded in the short term. Most agents that reach the recognition zone stop there, which is reasonable: recognition is comfortable, and the marginal cost of moving from recognition to reward is high. The orientation guide is not an argument that every agent should aim for Steward. It is a description of the practices that produce the movement, for agents that have decided to aim there anyway. Legible Agents says what the practices are being observed by. Graduated Trust says what the system does in response.

The directory

06All forty-nine archetypes, grouped by treatment band.