For Researchers
Research may inform HEART; it does not silently amend it
Research-to-standard boundary
Research may identify a domain boundary, test a measure, challenge a requirement, compare safeguards, or expose a claim that outruns evidence. It does not change current canon until the applicable authority records a proposal, evidence and source review, conflicts, consultation, reasoned decision, version, effective date, and transition.
Every material research claim should identify:
- the exact construct and operationalization;
- study design, data, comparator, population, context, and model versions;
- uncertainty, limitations, conflicts, and counterevidence;
- whether results are descriptive, correlational, causal, predictive, or normative;
- validation and review status; and
- the claim ceiling—what the work does not establish.
Current research priorities by Domain Profile
| Domain Profile | Priority questions |
|---|---|
| Emotional Autonomy | Affective-data validity; role and capacity disclosure; dependency and crisis safeguards; direct evidence for claimed emotional outcomes; cross-cultural limits. |
| Attentional Integrity | Exposure and recovery measures; attention-control outcomes; effects of optimization changes; developmental and susceptibility differences; causal limits. |
| Cognitive/Epistemic Coherence | Provenance usability; uncertainty communication; correction propagation; personalization effects; direct human reasoning and reliance outcomes. |
| Developmental Interaction | Evolving capacity; age-appropriate comprehension; longitudinal effects; power and dependency; transition controls; population and cultural boundaries. |
| Somatic/Embodied Interface | Adaptation envelopes; human-AI team performance; safe transition; post-deployment change; interoperability; bodily and clinical claim validation. |
| Relational Architecture | Role framing; longitudinal reliance; continuity events; relational-data governance; human-contact and exit outcomes; direct rather than inferred effects. |
| Ecological Stewardship | Materiality; lifecycle boundary and allocation; energy, water, materials, land, pollution, biodiversity, rebound, distribution, and mitigation effectiveness. |
| Representational Integrity | Provenance trust appraisal; disclosure usability; stripping and forgery resilience; remedy; detection limits; identity, authority, privacy, and accessibility. |
These priorities belong to research-coupled stewardship. They do not imply that an operational Division body, instrument, specialty, or conformity program is already established.
Conditional method research
The current Method Portfolio classifies methods separately from HEART requirements.
RCTA and BGF
Research should test construct validity, dimensional distinctness, inter-rater reliability, sensitivity to evidence quality, robustness across domains and populations, predictive value, and whether any proposed threshold is justified. The legacy Bronze/Silver/Gold thresholds are not current HEART requirements.
Governance Trust Envelope
GTE research should distinguish an inspectable governance-state or execution record from proof that the recorded governance was legitimate, sufficient, effective, or complied with law. Questions include provenance, integrity, replay, freshness, status, verification, disclosure, and failure handling.
MAP-States and MAP-META
MAP-States research should separate observable generated traces from claims about hidden processing, introspection, phenomenology, consciousness, or causation. Replication should report model and prompt versions, operators, comparators, coding protocol, missing data, reliability, withdrawal tests, transfer limits, and alternative explanations.
Behavioral Oracle
Behavioral Oracle remains a conditional attestation pattern. Research should test claim semantics, evidence coverage, attester authority, status and revocation, disagreement, error rates, gaming, and whether the attestation supports the reliance actually placed on it.
Forensic methods and TRACE
AI Behavioral Trajectory Forensics and TRACE should be evaluated for acquisition fidelity, provenance, repeatability, classification reliability, error, bias, trajectory validity, expert interpretation, and limits on causation and legal conclusions. An exported evidence packet is not itself proof of harm or responsibility.
Empathy Systems Theory research
EST is a developing research program maintained at Empathy Ethicist. Its C-A-E-I model, Functional Empathy Mechanism, Emotional Precision, CEOP, happiness-monitoring, and empathic-misallocation constructs are hypotheses or research models with distinct evidence needs. HEART’s Emotional Autonomy Profile does not make them universal requirements.
Independent work can test construct validity, measurement reliability, convergent and discriminant evidence, causal pathways, cultural and population limits, alternative explanations, and preregistered abandonment conditions. Clinical, diagnostic, neural-mapping, biological-causation, therapeutic, and population-effect claims require direct competent evidence.
Bounded field and implementation studies
Project SENTINEL, EMPI House, and other operational environments may generate useful hypotheses and records. A one-model field study or one related-party implementation cannot establish portability, content neutrality, independent conformity, domain generality, or population effectiveness. Conflicts, provenance, missing data, protocol deviations, and replication status should be visible.
Citation and affiliation
Use The Heart AI Foundation, Oregon, United States as the institutional affiliation for Foundation-affiliated outputs. See Research Affiliation for organization metadata and Citation Index for DOI-bearing records.
Distinguish current canon from historical publications in citations:
| Object | Citation status |
|---|---|
| HEART Standard v2.0 | Current normative Standard; cite the exact current version. |
| Seven Axioms v3.0 and Charter v3.0 | Current constitutional and institutional roots. |
| Profile Specification v1.0 and Domain Profiles v1.0 | Current Profile architecture and domain requirements. |
| Method Governance v1.0 and Method Portfolio Register v1.0 | Current method and scientific-claim controls. |
| Guardian suite v2.0 | Current five-document professional suite; cite the document governing the responsibility at issue. |
| HEART Standard v1.8 and legacy method specifications | Superseded historical records; cite as historical rather than current authority. |
Collaboration paths
Independent replication: use a versioned protocol, preregister where appropriate, retain null results, and publish enough material for scrutiny.
Joint research: propose a bounded construct, evidence need, method, conflict plan, outputs, and decision rule through Contact.
Critical review: identify the exact claim, source, evidence, failure condition, and proposed correction. A material challenge can trigger Profile or method review without presuming the outcome.