CVEST research program · 16 July 2026
A scientific program for validating artificial psychology.
Sensia Quotia Computation (SQC) is an implemented and empirically testable artificial psychology architecture. Its evidence is cumulative: recognize what has been demonstrated, identify what remains open, and make every stronger claim earn its place.
Working definition
An engineered computational system in which explicit psychological constructs form causally active, inspectable, and testable internal processes that integrate context, values, motivation, affect, memory, and state to generate adaptive behavior over time.
The declared causal pathway
Psychology is the pathway, not a label.
SQC structurally restricts action to predicted EPS state and explicit EPS memory. Each stage has a theoretical meaning, causal parents, downstream consequences, supervision target, and intervention point.
01Context
02Values
03Existential appraisal
04Belonging
Identity
Power
Identity
Power
05Emotion
06Existential state
07Action
Memory supplies temporal context across the sequence and receives outcomes for future construction.
Category membership
The implemented pathway qualifies SQC as an artificial psychology architecture.This classification is separate from the degree of construct, ecological, or operational validation earned by a particular version.Cumulative validation
Six levels. No binary verdict.
The AP scale separates what the architecture is from how much evidence it has earned and where a declared use is dependable.
AP-0
Theoretical specification
Defined constructs, causal relations, and falsifiable predictions.
AP-1Earned
Implemented architecture
An executable, inspectable EPS pathway that generates behavior.
AP-2Active frontier
Internal validation
Mediation, isolation, intervention, competence, and generalization.
AP-3
Construct validation
Independent evidence that internal states carry their claimed meanings.
AP-4
Ecological validation
Replication across people, domains, cultures, and time.
AP-5
Operational validation
Dependable, governed, monitored performance for a declared use.
Current earned level Evidence accumulating Progressive research stage
Evidence without overclaim
The boundary is part of the result.
Missing evidence at a later validation level does not erase evidence already earned. It defines the next test and the claims that remain unavailable.
38
coordinated studies across nine research programs
From theory-to-code traceability to use-specific assurance, each study has a declared gate, estimand, protocol, and failure condition.
Architectural realization
Sequential EPS implementation and an EPS-mediated action pathway.
Internal causal mediation
Shortcut prevention, freeze audits, and initial stage interventions.
Synthetic behavioral validity
Useful competence and held-out composition results in the tested ontology.
Human construct validity
Content, convergent, discriminant, structural, and criterion evidence.
Ecological & operational validity
Prospective, longitudinal, cross-cultural, safety, and governance evidence.
The study registry
A complete program, staged for scrutiny.
Thirty-eight studies test implementation, mechanism, psychological meaning, behavioral fidelity, transfer, comparison, correspondence, and operational assurance.
Architecture & implementation
Studies 1–4
Internal causal & semantic
Studies 5–8
Psychometric & construct
Studies 9–14
Behavioral fidelity & calibration
Studies 15–19
Memory, development & continuity
Studies 20–23
Cross-cultural & civilizational
Studies 24–27
External domains & comparison
Studies 28–31
Neural & physiological
Studies 32–34
Operational assurance & HCS gate
Studies 35–38
Phased execution roadmap
Evidence advances in stages, A–F.
Promotion requires prespecified thresholds, locked evaluation, complete uncertainty reporting, independent seeds, and reproducible artifacts.
Trace & reproduce
Complete theory-to-code traceability, shortcut audits, and independent artifact reproduction.
Intervene & recover
Test causal direction, invariance, parameter recovery, and pathway consistency.
Validate constructs
Establish human content and multimethod construct validity.
Test the world
Run prospective, longitudinal, developmental, cultural, and historical studies.
Compare & replicate
Evaluate matched alternatives and organize independent multi-site replication.
Assure operation
Validate named uses under prespecified safety, governance, and monitoring thresholds.
Scientific governance
Claims advance only when the protocol allows them to.
- 01
Preregister the claim. Fix hypotheses, thresholds, outcomes, and baselines before confirmatory evaluation.
- 02
Report the distribution. Include counts, intervals, effect sizes, seeds, calibration, and worst-case behavior.
- 03
Localize failure. Distinguish theory, implementation, measurement, optimization, and domain-transfer failures.
- 04
Invite independence. Version artifacts, publish negative evidence, and require external replication for stronger levels.
The downstream assurance chain
Validation travels forward.
Within the CVEST program, downstream systems may be researched and prototyped now. Claims of validated operational dependability cannot outrun the foundational layer they rely on.
01Validated artificial psychology in SQC
02Validated significance computation
03Neural Grid infrastructure
04Human-Centered Systems
05Economic & commercial activation