# AkashicOMNI Study Plan and Protocol v0.1

## Can a proposed fundamental-consciousness model make a distinctive, testable prediction?

**Protocol draft · 26 September 2026 · Not a registered or approved study**  
**Programme lead:** To be appointed; independent methods, statistical, ethics and cultural advisers required.  
**Framework:** AkashicOMNI v0.4.3 is the current public release. The v0.5.0 Claim Review Record is a candidate, not a released verification engine.  
**Status:** Design for review. No participants recruited, data collected, field detected or BQ001 status changed.

### Plain-language premise

Experience is directly available to the person having it. Other people inspect reports, behaviour and physical measurements. A proposed field or fundamental property becomes an empirical hypothesis only when it says what observation should differ from a specified alternative. Evidence for one difference cannot by itself establish that all of reality is conscious. The programme therefore keeps a **model test**, an **information-access test**, and a **metaphysical interpretation** in separate records.

This design borrows adversarial prediction-setting and theory-neutral analysis from the Cogitate protocol and results [R1, R2]. It uses the first-person/third-person measurement boundary illustrated by perturbational complexity research [R3]. Cosmopsychism is represented as a possible theoretical starting point, not a validated field claim [R4].

## A. Questions and outcome hierarchy

| Level | Question | What an answer would establish |
|---|---|---|
| A1, model specification | Can a named theory predict a measured quantity and a rival outcome before data collection? | Whether the model is presently testable. Failure to specify is **not** evidence that the ontology is false. |
| A2, instrument study | Does a calibrated physical readout depart from the rival prediction under blinded, prespecified conditions? | Support or challenge to the *specific physical prediction*, if artifacts are adequately ruled out. |
| B1, optional information task | Are one-off choices of future randomized targets above chance under locked conditions? | Evidence for an anomalous statistical result if independently replicated and audited. It would not identify a field or prove panpsychism. |
| C1, interpretation | Do alternative causal accounts still predict the findings? | A bounded explanation ranking, with gaps and dissent recorded. No automatic Big Question promotion. |

**Primary principle:** No single study here can prove universal or cosmic consciousness. If a fundamentalist and a non-fundamentalist model predict identical observations, the proposed ontological distinction is experimentally unresolved.

## B. Governance and preregistration

1. Form an adversarial steering group: at least one proponent of the specific fundamental model, one relevant critic, a theory-neutral experimentalist, statistician, measurement engineer and participant representative. An Indigenous or animist perspective is invited only on its own terms, with appropriate authority and compensation; it is not made a proxy for a universal physical hypothesis.
2. Publish a dated theory card *before* choosing instruments or seeing outcomes: exact claim, equations or operational rules, predicted effect size or minimum meaningful effect, units, sensor location, temporal window, necessary conditions, alternative predictions and observations that would weaken it.
3. Register protocol, code, analysis, exclusion rules, negative-control tests and interpretation matrix in a timestamped repository. Allow amendments only before the affected data are unblinded. Preserve both versions and reasons for changes.
4. Obtain research ethics approval for any participant recruitment. Use noninvasive tasks and ordinary consent language; no medication changes, drug administration, sleep deprivation or induced distress. Consent separates participation, data retention and any later quotation. Do not solicit detailed spiritual or clinical histories unless a specific approved question requires them.
5. Appoint an independent data custodian and audit group. Theory advocates can design predictions but do not control target generation, sensor calibration, exclusions or final analysis.

## C. Track A: physical field prediction

### C0. Mandatory theory-readiness gate

Proponents must complete the following before Track A can receive a sample size or registration as a confirmatory experiment:

| Field | Required entry |
|---|---|
| Proposed quantity | What the putative field is said to alter; physical units or an explicitly defined dimensionless readout. |
| Instrument | Existing calibrated sensor, sensitivity, noise spectrum, shielding and its independent validation. A newly invented sensor needs separate engineering validation before use. |
| Contrast | Two experimental conditions with a quantified difference predicted by the field model and a different quantified prediction from the strongest ordinary account. |
| Timing | Onset, duration, lag, location and permitted analysis window specified before recording. |
| Confounds | Temperature, vibration, electromagnetic coupling, shared clocks, respiration, movement, demand effects, experimenter access and analysis choices assessed for the selected instrument. |
| Failure threshold | A model-approved interval or outcome that would count against its proposed coupling, given adequate measurement sensitivity. |

If any row is missing, report **theory not yet ready for an empirical field test**. Do not substitute an EEG change or a meaningful subjective experience for the missing physical prediction: ordinary neurophysiology also predicts changes in EEG and experience.

### A1. Calibration and feasibility, if C0 passes

- Bench-test the sensor with blinded simulated positive and null signals. Verify that the prespecified effect can be detected at the proposed threshold. Time-stamp and independently archive raw signals.
- Run sham, active and sensor-disconnected controls in a randomized order. Record conventional physiology or environment only where needed to test the stated confounds.
- Lock a single primary readout and analysis window after calibration, **before** the confirmatory sample. Use pilot data only for engineering and power planning; exclude them from confirmatory inference.
- Compute sample size from measured variance, the model's minimum effect and the selected error threshold. An illustrative number is not a justified sample size for an unspecified signal.

### A2. Confirmatory test, if A1 passes

- Collect the locked sample at two independent laboratories, each with separate equipment, staff and analysis reproduction. Randomize conditions centrally; blind participants and recording staff to condition as far as the task permits; mask condition codes from analysts until code and exclusions are frozen.
- Primary analysis uses the exact model-predicted contrast. Report effect size, interval and raw distribution, including negative controls, deviations and sensor failures. A significant result found in an unregistered channel or time window is exploratory.
- Seek a new independent replication with the same physical prediction. A null result with insufficient sensitivity is inconclusive; a null result with verified sensitivity challenges the specific coupling model. A robust difference still requires independent checks for artifacts and alternative physical explanations before a field claim.

## D. Track B: fully specified optional information-access task

**Why separate:** Some models of fundamental consciousness predict no nonlocal information and no foresight. A result here is relevant **only if a named proponent has explicitly predicted it beforehand**. This task can nevertheless probe one concrete claim that information is acquired outside recognized routes. Generating the target *after* the response tests a future-target claim, not hidden contemporary target access; do not silently swap the two designs.

### B1. Primary question and preregistered hypotheses

- **H0:** each participant's single choice matches a future uniformly randomized one-of-four target with probability 0.25.
- **H1, conditional prediction:** a proponent commits that the probability is at least 0.30 under this task. The protocol tests the bounded statistical prediction, not a consciousness ontology.
- **Primary endpoint:** number of exact matches among 1,200 valid, first-attempt participants. One participant contributes exactly one binary result; no optional stopping or best-session selection.
- **Primary decision rule:** one-sided exact binomial test against 0.25 at alpha 0.005. With 1,200 trials, at least **340 exact matches** reaches that threshold (exact H0 tail approximately 0.00463). Power is approximately **90.2%** if the true match probability is 0.30. It has substantially less power for smaller effects. The threshold and power calculation are illustrative until statistician sign-off and independent code verification.

### B2. Participants and schedule

- Adults able to consent and complete a simple four-choice web task; no requirement for meditation, belief, psychedelics or special experiences. Offer accessible language and input alternatives.
- Recruit to *1,200 valid first attempts*, across at least two sites or independently supervised recruitment channels. Record all starts, withdrawals, technical exclusions and completion rates. A participant receives no real-time performance feedback before study closure.
- Conduct a separate **1,200-person replication** under new staff and infrastructure after the first study is frozen and reported. Success is not declared on the pooled sample if either independent stage misses its own prespecified criterion.

### B3. Single-trial procedure

1. Consent, anonymous study identifier and a lightweight eligibility/duplicate check. Store identifying verification separately from answers; do not publicly release identities.
2. Present four equally sized neutral symbols in a balanced randomized on-screen arrangement. Capture exactly one irrevocable selection and its server receipt time. Participants can withdraw before locking.
3. **Only after the selection is durably committed**, a separate audited service draws one of four equiprobable target labels with a cryptographically secure generator. It records a signed, append-only draw receipt and audit hash. The front-end cannot request a redraw; retries are logged and cannot replace failed trials silently.
4. A locked independent script joins each choice to its later draw by opaque trial ID. A second audit team reproduces the count from receipts without access to participant identities.
5. On completing the full study, disclose the scoring procedure and group result. Individual results may be released only where consented and without inviting clinical or spiritual conclusions about the participant.

**Pre-run validation:** simulate at least 100,000 dummy trials to confirm approximately uniform targets and a 25% match rate, rehearse interrupted requests and clock drift, audit duplicate controls and publish the generator/test code. Simulated data are excluded from the real result.

**Exclusions:** predefine lack of consent, no committed choice, invalid or absent signed target receipt and confirmed duplicate enrollment. Technical failures are reported by arm, time and site; no exclusion based on correctness, intuition ratings or whether the participant followed a preferred practice. All post-lock failures remain in an audit table with explicit treatment in the preregistration.

**Secondary analysis, explicitly exploratory:** site variation, participant belief or practice measured by voluntary minimal questions, elapsed response time, and whether simple symbol preferences affect choices. Correct for multiple comparisons if any are promoted to confirmatory hypotheses in a later independent study.

### B4. Threats and safeguards

| Threat | Safeguard | Remaining limit |
|---|---|---|
| Sensory leakage or knowledge of target | Target is generated only after the choice is locked; separate draw service and audited times. | This changes the claim to future-target access. |
| Predictable generator or biased symbols | Cryptographic draw, source audit, symbol-position balancing, independent statistical checks. | Software, hardware or audit compromise remain possible until independently reproduced. |
| Multiple tries and selective exclusions | One trial per person, append-only receipts, complete flow counts, frozen rules. | Duplicate detection must balance accuracy and privacy. |
| Publication or analytical selection | Registered primary endpoint and public report of null outcomes and protocol changes. | One surprising result may still be chance, error or an unknown mundane mechanism. |
| Cross-site dependence | Separate recruitment, code review, infrastructure and independent replication. | The second study must truly be independent of the first data and analysis tuning. |

### B5. Interpretation matrix

| Outcome | Justified wording | Wording excluded |
|---|---|---|
| Both stages near 25%, adequate quality | No support for an effect at this design's sensitivity; 30% prediction challenged. | “Consciousness is not fundamental.” |
| First stage passes, replication fails | Initial statistical anomaly unresolved; audit and follow-up required. | “Precognition confirmed.” |
| Both independent stages pass, audits clean | Replicated above-chance future-target matching in this task; investigate ordinary and novel mechanisms. | “Cosmic consciousness or a field has been detected.” |
| Technical imbalance or leakage identified | Results compromised; report and repair the design before any new test. | Selective post hoc rescue of favorable outcomes. |

## E. AkashicOMNI Claim Review Records

Make separate candidate records for: (1) the chosen ontology; (2) the field coupling, if specified; (3) the outcome of Track A; (4) the preregistered future-target prediction; (5) Track B results; and (6) any proposed relevance to BQ001. Each record retains the exact proposition, source access, source-dependence group, evidence and contrary evidence, rival models, participant/privacy restrictions, bounded outcome, second-review disagreement and revision condition. Never count an AkashicNET summary, N2N post and the same paper as three independent studies.

Outcome signals express editorial assessments, not numerical truth. A reported nonlocal-information result would require **further arguments** before it could bear on personal survival, universal consciousness or cultural interpretations. Existing BQ001 UNRESOLVED and canonical-edge, privacy, rights and publication gates remain unchanged.

## F. Deliverables, cost and stop points

**Before funding or recruitment:** theory cards, registered analysis plan, independent criticism, institutional ethics review, risk/consent materials, data-management plan, audit scripts, instrument specification (Track A), and exact test-power verification (Track B).

**Feasibility:** Track A stops at C0 if the proposed field has no discriminating physical readout. Track B requires engineering, participant recruitment and independent replication; 2,400 valid single trials are a substantial project. Do not treat the first 1,200 as exploratory and reuse them for a supposedly independent confirmatory analysis.

**Publication:** release full methods, aggregate results, code, deviations, null findings and an independent audit. Publish de-identified trial data only under ethics approval and consent, with access controls for any risk of re-identification. A public communication should state the exact tested claim, not a metaphysical verdict.

## G. Protocol synopsis and scope

| Item | Prespecified proposal |
|---|---|
| Design | Two conditional, separately registered research tracks; Track A is a blinded physical contrast only if its theory-readiness gate passes; Track B is a single-choice, future-target, four-alternative task in a discovery stage followed by an independent replication. |
| Population and setting | Track B: consenting adults recruited by at least two independently supervised channels per stage, with 600 valid first attempts per channel. Sites, languages, accessibility and recruitment venues require local ethics approval. Track A population follows from its eventual measurable contrast. |
| Exposure or intervention | Track B: no clinical intervention; one neutral-symbol choice followed by an independently generated target. Track A conditions and instrument must be defined in the theory card. |
| Primary outcome | Track B: total exact matches among 1,200 valid first attempts in each of two independent stages. Track A: one model-specific physical contrast, still to be specified. |
| Analysis and claims | Exact one-sided binomial test in each Track B stage; no pooled rescue. Track A analysis is locked only after calibration. Ontology is assessed in a separate interpretation review. |
| Registration status | Draft for investigator, statistician, engineer and ethics review. No preregistration, ethics approval, recruitment or study result is implied. |

This protocol is organized using relevant transparency items from [SPIRIT 2025](https://www.bmj.com/content/389/bmj-2024-081477) and [APA preregistration guidance](https://www.apa.org/pubs/journals/resources/preregistration). SPIRIT addresses randomized trial protocols; this proposed task is not a clinical randomized intervention trial and does **not** claim SPIRIT compliance.

### G1. What the study can identify

An unobserved entity can be inferred through distinctive, repeatable effects only after an explicit measurement model specifies how it acts on observables. The observation tests that *model*. Multiple mechanisms can explain the same effect. For Track A, the estimand is the mean or other preregistered functional of a calibrated sensor difference between named conditions, in named units and time window. For Track B, the estimand is the probability that a committed choice equals an independently drawn future target under this procedure. Neither estimand measures universal consciousness directly.

## H. Track B: operational study protocol

### H1. Roles, independence and sites

- **Sponsor and principal investigator:** named institution or legal entity, funding and conflicts disclosed before recruitment; obtains ethics approval and appoints a data controller.
- **Methods committee:** statistician, security engineer and at least one model critic review the protocol and implementation before registration. Proponents sign their predicted 0.30 match-rate claim or explicitly decline to link their model to this task.
- **Site leads:** oversee consent and recruitment; cannot see cumulative match counts. One channel should recruit beyond the core AkashicNET/N2N community to limit narrow self-selection. Convenience recruitment limits population generalization even when the statistical test is valid.
- **Target-service custodian:** controls the audited generator and signed receipts, independent of recruitment and the choice application. **Masked analyst:** receives site codes and validity flags, with targets withheld until analysis scripts and the dataset are frozen. **Independent auditor:** inspects chronological receipts and reproduces results.
- **Replication team:** distinct site leads, target-service deployment, audit keys and analyst; it may reuse the publicly locked protocol and code, but cannot tune exclusions or analysis using stage-one results.

### H2. Eligibility, consent and recruitment

Inclusion: age at least 18 (or a higher local age of consent), capacity to provide informed consent, ability to understand the task in an approved language and access to an approved interface. Exclusion before choice: inability to consent, previous enrollment in that stage, or failed eligibility verification. No screen for beliefs, perceived psychic ability, meditation, mental-health diagnosis or performance. Accessibility accommodations preserve the one-choice rule and are logged in advance; a usability pilot may refine the interface but contributes no outcome data.

Recruit through dated advertisements with neutral wording such as “one-choice study of prediction and random targets.” Avoid claims of diagnosing ability or establishing cosmic consciousness. Compensation, if any, is fixed and independent of accuracy. Keep a recruitment log by site: approached where countable, opened information sheet, consented, choice locked, target drawn, validity determined, withdrawn, duplicate or failed receipt. Use pseudonymous IDs; minimum identity checks must be locally approved and separated from outcome data. Do not expose personal identity to the target service.

Participants may leave before their choice is locked. Consent materials must explain what can be withdrawn afterward, given append-only audit records and applicable law; investigators must never promise deletion they cannot technically or legally perform. Provide contact details for questions and complaints, and an accessible debrief after study closure.

### H3. Trial sequence and provenance

1. Record the participant's consent version, stage, site and opaque ID. The interface shows four neutral symbols with fixed target labels and a uniformly shuffled display order. Run usability checks to ensure equivalent appearance and input ease.
2. Accept one choice. Write a durable choice receipt containing opaque trial ID, canonical selected label, server-side monotonic event order, UTC time, site and software version. The client receives an acknowledgement; duplicate submissions return the existing receipt rather than create a new trial.
3. The choice service sends only the locked trial ID to a separately operated target service. Before making a draw, that service verifies the signed choice receipt and that no target already exists for the ID.
4. Generate exactly two unbiased cryptographic random bits, map 00/01/10/11 to the four labels, and write a signed target receipt with the same trial ID and a later event order. Use a security-reviewed operating-system CSPRNG backed by documented entropy and health checks; document its implementation against the relevant [NIST SP 800-90A Rev. 1](https://nvlpubs.nist.gov/nistpubs/specialpublications/nist.sp.800-90ar1.pdf) guidance where applicable. A cryptographic pseudorandom generator is adequate for this protocol; do not call it a quantum generator.
5. Append both receipts to a tamper-evident audit log with chained hashes and independent daily timestamped checkpoints. Preserve error, retry and outage events. Neither site staff nor the front-end can preview, replace or redraw targets.
6. The blinded custodian resolves receipt integrity and eligibility using frozen rules. Only then does the analyst unmask targets, join by trial ID and score equality. Archive signed input snapshots, code hashes and verification reports.

Clock time alone cannot prove event order if services drift; signed receipt linkage and server-side causal order are required. An audit tests service isolation, permissions, software supply chain and whether the random sequence depends on user or site data. The target distribution, symbol-position balance and arrival-time patterns are reported even if the primary result is null.

### H4. Enrollment target, validity and missingness

Each stage ends when **exactly 600 valid first attempts per site** have accumulated across two sites, for 1,200 total. More sites are possible only if quotas and the analysis are set before that stage begins. Recruitment staff and the stopping service must see validity counts only, never match counts. No interim efficacy testing or stopping based on outcomes. If an unexpected outage interrupts a stage, log it and follow the registered hold/restart rule without changing already committed trials.

A valid attempt has approved consent and eligibility, a unique participant, a durable first choice, and exactly one cryptographically verifiable later target receipt. Frozen integrity checks run without match information. Every started and locked attempt remains in the public flow diagram. A missing target has no observable success status; exclude it from the primary denominator only under the prespecified rule, report its count and cause, and perform extreme-case sensitivity bounds treating every such locked attempt as a miss and then as a hit. A target receipt must never be invalidated based on its label or whether it matches. Participants with a locked choice and valid target remain in the primary analysis even if they later close the page or decline feedback, subject to approved withdrawal requirements.

Operational quality gate: pause recruitment if invalid locked attempts exceed 5% in either site, if receipt order cannot be verified, if target entropy checks fail, or if evidence of leakage or unauthorized access appears. The independent auditor records the reason, incident window and corrective action before any restart. A pre-unmasking amendment can authorize a new cohort, but compromised trials are reported and cannot be quietly replaced. The 5% threshold is an engineering safeguard, not a statistical exclusion optimized on outcomes.

### H5. Masking and quality assurance

Prior to launch, run at least 100,000 simulated trials, including forced retries, network interruptions, clock changes, duplicate requests and partial database failures; check generator mapping and the entire receipt-to-score pipeline. An independent engineer reviews deployment configuration and performs an attempted target-leakage test. Preserve test logs and separate all simulated IDs from real data. Randomness tests can reveal a broken implementation but cannot prove unpredictability or metaphysical independence.

Consent and task administrators are unaware of targets at choice time. The target service knows target labels after drawing but cannot alter prior choices. The outcome analyst is blinded until code, eligibility list and receipt-integrity report are locked. Log all access to raw target records. Masking is assessed through permissions and audit logs; simply calling a process “blind” is insufficient.

### H6. Statistical analysis plan

Let (X_i=1) when valid participant (i)'s committed label equals the later target, otherwise (0); (K=\sum_{i=1}^{1200}X_i). If target labels are uniform and independent of every committed choice, (K\sim\mathrm{Binomial}(1200,0.25)) under the operational null, even when participant symbol preferences are uneven. The primary test computes (P_{0.25}(K\ge k_{\mathrm{observed}})) exactly. The registered rejection rule is **(K\ge340)** in stage one and separately **(K\ge340)** in the independent replication. The exact one-sided tail at 340 is about 0.00463; if (p=0.30), the chance of reaching 340 in a single stage is about 0.902. These values require independent implementation checks before registration.

Report stage-specific (k/n), excess over 0.25 in percentage points, an exact binomial confidence interval (95% two-sided, labelled descriptive), exact (p), attrition, target-label counts and audit findings. The two-stage declaration requires both stages independently meet their thresholds **and** pass audit. Do not treat the second stage as an extension that makes the first significant, combine stages to rescue a miss, change (n), or add a second primary outcome after seeing data. Since success is an intersection of two prespecified tests, the individual 0.005 thresholds are conservative for the whole claim under the common null; this does not correct bias, leakage or selective reporting.

Site-level counts and confidence intervals, symbol-choice frequencies, response times and prior-belief associations are descriptive or exploratory. If future protocols make them confirmatory, predefine hypotheses and multiplicity adjustments separately. For locked attempts without valid targets, present missingness and the extreme-case sensitivity bounds above; do not impute success from other participants. Check serial dependence, label balance and outcomes around deployment incidents, but any post hoc filtered result is exploratory. A confidence interval covering small effects is not proof of the exact null; a result below 340 means the proposed primary criterion was not met.

### H7. Decision and interpretation after each stage

Stage-one outcomes are fully disclosed and frozen before replication registration. Replication uses a new cohort, independent infrastructure and the unchanged threshold. If both pass but an integrity or dependence audit fails, classify the finding as **uninterpretable pending a clean rerun**. If both pass with clean audits, report a replicated statistical anomaly *in the defined future-target task*. Then commission separate studies to distinguish software error, subtle selection, ordinary information routes and novel explanations. Repetition alone does not identify a field, prove that the universe experiences anything, or support survival after death. A miss in either stage defeats the protocol's strict two-stage success criterion; report both stage results without concealing a potentially interesting difference.

## I. Track A: specification template and conditional methods

No actual field experiment is approved by this document. Before any Track A recruitment, the steering group must fill in and publicly lock the following theory card; “unobservable” is not a substitute for a coupling function.

| Required registration field | Exact entry to supply |
|---|---|
| Model identity and prediction | Named version, equations or algorithm; which observable changes, its units and predicted direction/magnitude. |
| Rival and null models | At least one concrete ordinary causal account with a different predicted value or temporal/spatial signature. |
| Independent variable | Reproducible conditions, manipulation checks, randomization unit, allocation method and sham definition. |
| Measurement chain | Sensor, calibration references, limits of detection, sampling rate, filtering, time synchronization and raw-data retention. |
| Primary estimand and analysis | Exactly one contrast, window, transformation, statistical test, interval, error rate, sample-size calculation and smallest effect of interest. |
| Confound isolation | Shielding, movement, thermal and electromagnetic controls, staff access, blinding and physical artifact checks. |
| Falsification and sensitivity | Measurable range that would challenge the specified coupling model *given independently verified instrument sensitivity*. |
| Replication | Second site's equipment independence, code reproduction, acceptance criteria and plan for disagreement. |

Feasibility follows a staged order: (i) theory card and rival predictions; (ii) bench validation and blinded positive controls; (iii) pilot to estimate sensor variance without testing the final claim; (iv) locked confirmatory sample-size and analysis plan; (v) two-site collection; (vi) new replication and independent artifact review. If a theory predicts the same observation as its rival or declines a magnitude/measurement link, classify Track A as **not presently discriminating**. A very precise null may challenge a particular coupling, but it cannot disprove all panpsychist or cosmopsychist theories. Cross-cultural or animist concepts require specific, consented collaboration and should not be recoded as if they were interchangeable laboratory field hypotheses.

## J. Ethics, participant welfare and data stewardship

Seek approval or formal exemption determination from the competent institutional ethics body before recruiting. Submit exact advertisements, consent and debrief wording, privacy impact assessment, compensation, accessibility plan and security architecture. This is not a diagnostic test, therapy or measure of a person's spiritual worth. Participants should not be told that performance reveals a personal power. Provide a route for complaints and withdrawal questions; monitor for distress caused by study framing, withdraw misleading recruitment text promptly, and record and report adverse events according to the approved plan. No deception about the fact that the target is generated after the choice.

Collect only necessary variables. The research dataset contains opaque ID, stage/site, consent version, event and receipt references, selected and target labels, validity status/reason, software version and integrity hashes. Identity/duplicate-check data stay encrypted with a separate restricted custodian and a fixed, ethics-approved deletion schedule; specify actual retention durations and lawful basis in the site plan before launch. Restrict target keys and personal data by role, use encryption in transit and at rest, rotate credentials after incidents and document breach notification routes. Publish a de-identified dataset only if risk assessment permits; otherwise provide controlled access to auditors and a public aggregate plus executable analysis code. Hashes of sensitive rows should not be mistaken for anonymization.

The governing group must record funder influence, conflicts of interest, protocol ownership, data access, authorship, preprint/publication plans and a commitment to publish null and adverse audit outcomes. Respect community interpretations and cultural provenance in the interpretation review without treating an individual community as a source of universal consent or empirical endorsement.

## K. Monitoring, deviations and reporting

Monitor enrollment, receipt validity, software outages, consent errors and privacy incidents at least weekly with accuracy outcomes masked. The independent auditor can halt either stage for safety, security or invalid receipt order. Resume only after documented root-cause analysis, ethics notification where required and a dated amendment that leaves all prior records visible. If a serious protocol violation makes primary inference untenable, report the stage as compromised and run a newly registered study; do not repair significance by dropping unfavorable trials.

Maintain an amendment ledger with date, author, rationale, affected stages, whether anyone had seen outcomes, ethics decision, code/data hashes and public link. Publish a completed flow diagram, protocol, registration timestamps, signed theory cards, aggregate primary and negative-control results, all deviations, analysis scripts, independent audit summary and a plain-language limits statement. Archive raw auditable receipts securely. No BQ001 or AkashicOMNI status changes are automated by a study result.

### K1. Illustrative delivery milestones (not committed dates)

| Gate | Exit criterion | Key dependency |
|---|---|---|
| 1. Theory and governance | Signed Track B prediction or explicit independence from theory; Track A card assessed; named accountable team. | Proponent, critic, engineer and statistician participation. |
| 2. Registration and ethics | Frozen code, consent/data plan, institution approval and public timestamp. | Site selection and privacy assessment. |
| 3. Engineering pilot | Full-pipeline dummy audit, failure drills and independent security review. | Implemented isolated services. |
| 4. Stage one | 1,200 valid first attempts, clean receipts, public methods and result. | Recruitment and audit. |
| 5. Independent replication | New 1,200 valid first attempts, independent audit, joint interpretation. | New staff/infrastructure and resources. |
| 6. Model review | Bounded Claim Review Records and open publication, including nulls. | Two-stage report or documented stop. |

Cost cannot be responsibly priced without sites, jurisdiction, technical architecture, compensation and Track A instrument. Budget lines include personnel, accessibility and translation, participant compensation, ethics/privacy support, security audit, hosting, independent replication and long-term repository maintenance. Confirm resources for both stages before launching stage one.

## L. Appendices for implementation

### L1. Minimum participant information text for ethics review

“You will choose one of four symbols once. After your choice is recorded, a separate computer service will randomly select one of the four symbols. We will compare the two choices across many participants. The study cannot diagnose an ability or determine whether a philosophical theory about consciousness is true. You may stop before submitting your choice. The approved consent form explains data use and withdrawal after submission, compensation, privacy and how to contact the research team.” This is draft copy; local language, compensation and withdrawal rules require ethics approval.

### L2. Data dictionary and reproducibility package

| Field | Meaning | Public release default |
|---|---|---|
| `stage`, `site_code`, `trial_id` | Independent cohort/channel and opaque receipt join key. | Stage/site aggregated; trial IDs only after re-identification review. |
| `choice_label`, `target_label`, `match` | Canonical labels and computed equality. | De-identified rows if approved. |
| `choice_receipt`, `target_receipt`, `event_order` | Signed proof of committed choice before a unique target. | Redacted verification material or controlled audit. |
| `validity_flag`, `invalid_reason`, `software_version` | Frozen eligibility/integrity assessment and implementation. | Aggregate reasons and version; trial rows if approved. |
| `consent_version`, `withdrawal_status` | Approved participant rights and permitted analysis status. | Aggregate only unless approved. |

Publish code for receipt verification, deterministic score recomputation from the audited export and exact binomial test; a dummy dataset with expected output; software dependency versions; the frozen registry record; the analysis report; and the auditor's deviations and integrity statement. Commit a cryptographic digest of code and data snapshot before unmasking. Keep source identities and secret generator state out of public artifacts.

### L3. Before-launch sign-off checklist

- [ ] Named theory proponent signs the exact prediction it entails, or Track B is explicitly independent of the ontology.
- [ ] Track A theory-readiness gate passes or Track A is deferred without treating this as a negative result.
- [ ] Institutional ethics decision, consent/debrief, recruitment copy and retention schedule are recorded.
- [ ] Independent statistician reproduces the 340/1,200 rule and power by separate code.
- [ ] Security reviewer validates one-choice locking, post-choice target generation, isolation and recovery paths.
- [ ] Missing/invalid receipt and pause rules are simulated; auditors can access immutable logs.
- [ ] Both recruitment channels, quotas and replication independence are funded and named.
- [ ] Registration is publicly timestamped before real data, with all amendments traceable.

**Revision note, 26 September 2026:** Expanded the original v0.1 study plan into an operational draft protocol. The study retains draft status pending named investigators, site-specific ethics review, security implementation and registration. None of the proposed experiments has been run.

## Sources and method inspiration

- [R1] Melloni et al., [adversarial protocol for IIT and GNWT](https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0268577), *PLOS ONE* 18:e0268577 (2023). Protocol design, preregistration and competing predictions.
- [R2] Cogitate Consortium et al., [adversarial test and results](https://pmc.ncbi.nlm.nih.gov/articles/PMC12137136/), *Nature* 642:133–142 (2025). An example of mixed results and limits of theory comparison; not evidence for the field model proposed here.
- [R3] Casali et al., [perturbational complexity index](https://pubmed.ncbi.nlm.nih.gov/23946194/), *Science Translational Medicine* 5:198ra105 (2013). Human state discrimination from a brain measure; no ontology established.
- [R4] Keppler and Shani, [cosmopsychism and consciousness research](https://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2020.00371/full), *Frontiers in Psychology* 11:371 (2020). Perspective/theoretical proposal, not a validated measurement of cosmic experience.
- [R5] [SPIRIT 2025 statement](https://www.bmj.com/content/389/bmj-2024-081477), *BMJ* (2025). Transparency checklist for randomized trial protocols, adapted selectively here.
- [R6] [APA preregistration resources](https://www.apa.org/pubs/journals/resources/preregistration). Guidance for specifying hypotheses and analysis before observing results.
- [R7] [NIST Special Publication 800-90A Revision 1](https://nvlpubs.nist.gov/nistpubs/specialpublications/nist.sp.800-90ar1.pdf). Deterministic random-bit-generation guidance; deployment still requires independent security review.

**Authorship and limits:** Human-directed research question and AkashicNET governance context; AI-assisted methodological synthesis and approximate power calculation. No independent scientist, ethics board or engineer has yet approved this protocol. The field track deliberately has no sample size until a specific testable coupling is supplied. The information task is a test of a distinct hypothesis and must never be represented as a direct test of panpsychism or cosmopsychism in general.