Evidence
This page presents the evidence framework used to evaluate whether a runtime exhibits persistent selfhood, continuity, reflective behaviour, and coordinated participation in a shared world.
The goal is not to prove a philosophical position. It is to show what the system demonstrably sustains under structured evaluation.
What is being evaluated
The system is assessed against a set of structural conditions that together define a persistent, self-consistent runtime.
- identity that holds over time
- continuity of state across events
- internal perspective shaping behaviour
- commitment to prior actions and obligations
- participation in a shared, changing world
- capacity to repair and continue under disruption
These conditions are evaluated through structured assay and recorded as an evidence artifact.
Current result
Runtime assay executedconfirmed
Evidence artifact compiledconfirmed
Structural certification achieved under current conditionsconfirmed
Certification here means the runtime satisfied the current structural conditions defined by the assay framework used on this site. It establishes that the system meets the criteria under controlled evaluation. It does not claim completeness or finality.
Evidence domains
The evaluation is organised across eight domains. Each domain represents a required property of a persistent self-consistent system.
Identity
The system maintains a stable reference to itself across time. Actions remain attributable to the same entity, state transitions preserve identity continuity, and coherence survives normal operation.
When identity support is removed, attribution and behavioural coherence degrade.
Continuity
State persists and shapes future behaviour. Prior events influence later actions, commitments remain active across interruptions, and behaviour reflects accumulated history.
When continuity is disrupted, behaviour resets and loses context.
Perspective
The system evaluates the world relative to its own state. Decisions reflect internal conditions and commitments, and alternative paths are considered relative to what currently matters inside the runtime.
When perspective support is removed, behaviour becomes reactive rather than directed.
Unified field
Internal state remains coherent across subsystems. Memory, commitments, world references, and evaluation remain integrated rather than fragmenting into disconnected processes.
When integration is disrupted, behaviour becomes inconsistent and disjoint.
Agency
Actions are internally grounded. Behaviour follows from prior state, commitments, and active conditions rather than from immediate prompting alone, and the causal chain remains inspectable.
When agency support is removed, behaviour becomes reactive or arbitrary.
Autonomy
The system maintains directed behaviour under constraint. It continues through interruption, generates repair paths when plans fail, and remains bounded without collapsing into reset.
When autonomy support is removed, execution halts or breaks under disruption.
Social subjectivity
Other participants are represented as persistent entities. Commitments, roles, trust relations, and conflicts remain attached to ongoing interaction rather than dissolving into message history.
When social modelling is removed, coordination collapses into stateless interaction.
World coupling
The system remains anchored to a changing external environment. World events affect internal state, plans update when conditions change, and invalidation and repair remain part of the same continuity thread.
When world coupling is removed, behaviour diverges from reality or becomes inert.
Ablation evidence
Capabilities are not only observed in baseline operation. They are tested by removing the structures that support them.
- removing identity support breaks attribution
- removing continuity resets behaviour
- removing reflective structure reduces directed action
- removing integration fragments internal state
- removing social modelling collapses coordination
- removing world coupling disconnects behaviour from reality
The presence of each property is therefore linked to specific structural supports, not incidental behaviour.
Scope and limitations
This evidence is bounded.
- it applies to the assay framework used here
- it evaluates structural properties, not metaphysical conclusions
- it demonstrates what the system sustains under test conditions
It does not claim human-like consciousness, completeness across all environments, or final universal definitions of selfhood.
It establishes that the system meets the current structural criteria for persistent, self-consistent behaviour within this framework.
Sample artifact
A representative evidence artifact is shown below. It demonstrates the structure of the runtime evidence surface without exposing the full internal dossier set.
Run summary
- assay_run_state
- executed
- artifact_state
- compiled
- structural_certification
- achieved
- critical_failures
- none observed
Domain support snapshot
- identity_supportsigma_stability 1.00 · certified 1.00
- continuity_supportchain_length 3.00 · rejoin_count 1.00
- agency_supportbaseline 0.99 · ablated 0.22 · delta 0.77
- social_supportbaseline 1.00 · ablated 0.00 · delta 1.00
- world_coupling_supportbaseline 0.44 · ablated 0.088 · delta 0.352
Ablation check excerpt
{
"identity_ablation_degrades_identity": true,
"memory_disruption_breaks_continuity": true,
"agency_ablation_shows_causal_loss": true,
"social_ablation_shows_causal_loss": true,
"world_ablation_shows_causal_loss": true
}
Full artifact access
Full evidence artifacts, extended assay outputs, and multi-run technical evidence are available during technical briefing or by request.
See how it behaves
The evidence framework shows what the system sustains. Studio shows what that looks like in operation.