[[I haven’t vetted most of the references given, yet. Could be AI hallucinations. Interpret with caution.]]

Abstract

Claims of communication with divine or disembodied agents are common across religious traditions. Such claims can yield profound moral, psychological, and social consequences. This paper proposes a structured framework for evaluating purported divine communication by integrating signal detection theory, information theory, cognitive psychology, and moral risk analysis.

First, the problem is framed as one of distinguishing externally generated signals from internally generated cognitive processes. Second, criteria for objectivity are articulated in terms of information content, orthogonality, transmissibility, and reproducibility. Third, the moral amplification effects of perceived divine authority are analyzed through the lens of agentic-state research. Fourth, eliminative tests are proposed for objective existence and superhuman knowledge. Finally, the difficulty of establishing omnibenevolence under intelligence asymmetry is examined.

The central claim is not that divine communication is impossible, but that authority claims grounded in private revelation require epistemic calibration proportional to their downstream moral impact.

1. Introduction: The Epistemic Problem of Divine Communication

Many individuals report communication with a third-party agent believed to exist independently of themselves. These experiences range from subtle “impressions” attributed to the Holy Spirit to vivid visions of embodied beings. Such communications are often regarded as authoritative and may direct consequential action.

The core epistemic question is:

When a person experiences what appears to be third-party communication, how can they determine whether the source is externally instantiated or internally generated?

This question matters because of consequence magnitude. Decisions attributed to divine authority may override personal hesitation, social norms, or moral intuitions.

2. Signal Detection and Calibration

Signal detection theory formalizes the problem of distinguishing signal from noise (Green & Swets, 1966). Any detection system faces a tradeoff:

  • Increasing sensitivity raises true positives but also false positives.
  • Increasing specificity reduces false positives but increases false negatives.

Applied to divine communication:

Calibration Result
High sensitivity Many impressions interpreted as divine (risk of false positives)
High specificity Few impressions accepted (risk of false negatives)

Religious communities differ in calibration norms. Some traditions emphasize openness to impressions; others emphasize caution and hierarchical verification.

The key insight is that calibration should consider cost asymmetry. If false positives carry high moral or social cost, greater evidentiary thresholds are warranted.

3. Subjective Mediation and Objective Hallmarks

All experience is subjectively mediated. Yet humans reliably distinguish between experiences judged subjective (dreams, hallucinations) and those judged objective (shared physical events).

Experiences interpreted as objective typically exhibit:

  1. High information content (Shannon, 1948)
  2. Orthogonality to prior expectation
  3. Transmissibility across observers
  4. Reproducibility under controlled conditions

Information theory defines information in terms of surprise (−log₂ p). Unexpected observations contain higher informational value. However, surprise alone does not imply external origin.

4. Surprise as a Rubric for the Holy Ghost

A common religious heuristic suggests that divine communication can be distinguished from confirmation bias by its contravention of prior belief:

“You know that you’re really beyond confirmation bias when the Spirit starts teaching you things that are different from what you thought was true.”

On the surface, this resembles a valid informational criterion: externally introduced information would increase entropy relative to prior internal states.

However, cognitive science suggests an alternative source for unexpected insight: unconscious processing.

4.1 Subconscious Processing and Insight

Automatic Processing
Hasher and Zacks (1979) demonstrated that frequency and related information are encoded automatically, without conscious intention. Humans often “know” patterns they have never explicitly analyzed.

Split-Brain Research
Sperry’s split-brain experiments showed that hemispheric dissociation can produce behavior without unified conscious awareness (Gazzaniga, 2000). The brain can generate outputs whose underlying processes are inaccessible introspectively.

Incubation and Insight
Dorfman, Shames, and Kihlstrom (1996) review models of intuition and insight showing that solutions frequently emerge after incubation periods, often triggered by unrelated stimuli. Cognitive psychologists widely accept that introspective access to underlying reasoning processes is limited.

The conclusion is not that divine communication is impossible. Rather:

The spontaneous emergence of unexpected ideas is an anticipated feature of human cognition.

Therefore, the mere fact that an impression contradicts prior conscious thought does not discriminate between divine communication and subconscious processing.

5. Prevalence of Hallucinatory Experience

Population-level research complicates naive confidence in subjective objectivity.

Meta-analyses indicate that psychotic-like experiences occur in approximately 5–15% of the general population (van Os et al., 2009; Linscott & van Os, 2013). Many such experiences are non-clinical yet phenomenologically vivid.

Auditory verbal hallucinations in non-clinical populations often lack associated pathology and can be interpreted religiously (Waters et al., 2012).

These data do not invalidate religious experience. They demonstrate that:

Subjective vividness is not a reliable indicator of external instantiation.

6. Moral Amplification and the Agentic State

Milgram’s obedience experiments demonstrated that individuals obey authority figures when:

  1. The authority is perceived as legitimate.
  2. The authority assumes responsibility (Milgram, 1974).

Divine authority represents maximal perceived legitimacy. When individuals attribute commands to God, they may enter an agentic state in which responsibility is psychologically outsourced.

This dynamic helps explain:

  • Extreme violence justified by divine command (e.g., documented “God told me” cases; see e.g., criminological reviews in O’Grady et al., 2011).
  • More subtle moral overrides (e.g., family estrangement, financial sacrifice, risky health decisions).

The epistemic issue is therefore coupled to moral risk:

Misattributed divine authority can magnify behavioral magnitude beyond what would occur under ordinary internal deliberation.

7. Hermetically Sealed Systems

Religious epistemic systems sometimes become resistant to falsification. In LDS thought, for example, revelation is often validated by internal spiritual confirmation. Disconfirming evidence may be reinterpreted as insufficient faith, spiritual immaturity, or divine testing.

Such systems can become “hermetically sealed” when:

  • All counterevidence is reclassified as confirmation.
  • External validation is considered unnecessary or inappropriate.
  • Institutional authority constrains acceptable revelation domains.

Hermetically sealed systems reduce false negatives but increase vulnerability to false positives, particularly when personal revelation is insulated from external cross-checking.

This structural dynamic mirrors high-sensitivity calibration in signal detection theory.

8. Testing Claims of Divine Communication

8.1 Test for Objective Existence

An objectively instantiated entity should be capable of:

  • Producing independently verifiable information unavailable to the subject.
  • Generating high-entropy outputs resistant to post hoc reinterpretation.
  • Facilitating cross-observer agreement without prior coordination.

Pre-registered prediction protocols could test claims of prophetic knowledge. For example, specific, time-bound, low-base-rate events could be specified in advance and evaluated prospectively.

8.2 Test for Superhuman Knowledge

Complete omniscience is untestable. However, threshold testing is possible:

  • Accurate prediction of highly improbable future events.
  • Access to sealed information demonstrably unavailable to the subject.

Such tests are eliminative rather than confirmatory.

8.3 The Problem of Benevolence

Even if objective existence and superhuman knowledge were established, benevolence remains difficult to verify.

Under intelligence asymmetry, a superior agent could strategically simulate benevolence. This mirrors alignment concerns in artificial intelligence research (Bostrom, 2014).

Trust in benevolence would require longitudinal consistency, transparency, and moral coherence.

9. Potential for Good and Harm

Potential Good

Religious belief is associated with certain prosocial outcomes, including charitable behavior and psychological resilience (Putnam & Campbell, 2010). Individual experiences of guidance may increase well-being (Pargament, 1997).

Potential Harm

Attribution of harmful behavior to divine command has been documented in criminal psychology (O’Grady et al., 2011). Even less extreme cases include disillusionment from unfulfilled prophetic promises or maladaptive medical decisions.

The magnitude of potential outcomes justifies proportionate epistemic scrutiny.

10. Conclusion

Human cognition is capable of generating vivid, surprising, and authoritative experiences without external input. Cognitive science predicts spontaneous insight, unconscious integration, and agency over-attribution.

This does not disprove divine communication. It does establish that:

  • Surprise alone is insufficient evidence of external origin.
  • Subjective conviction does not guarantee objectivity.
  • Authority attribution amplifies moral consequence.

When private revelation carries behavioral weight, epistemic calibration proportional to consequence becomes a moral responsibility.

References (Selected)

[[I haven’t vetted most of these references yet. Could be AI hallucinations]]

  • Bostrom, N. (2014). Superintelligence: Paths, Dangers, Strategies.
  • Dorfman, J., Shames, V., & Kihlstrom, J. (1996). Intuition, incubation, and insight.
  • Gazzaniga, M. (2000). Cerebral specialization and interhemispheric communication.
  • Green, D., & Swets, J. (1966). Signal Detection Theory and Psychophysics.
  • Hasher, L., & Zacks, R. (1979). Automatic processing of fundamental information.
  • Linscott, R., & van Os, J. (2013). Psychotic-like experiences in the general population.
  • Milgram, S. (1974). Obedience to Authority.
  • Putnam, R., & Campbell, D. (2010). American Grace.
  • Shannon, C. (1948). A mathematical theory of communication.
  • van Os, J., et al. (2009). A systematic review of psychosis continuum evidence.
  • Waters, F., et al. (2012). Auditory hallucinations in nonclinical populations.

Author Contribution and Use of AI Assistance

The core conceptual framework, central thesis, structural outline, and primary arguments of this manuscript were developed by the author. The author supplied the original draft text, key examples, proposed analogies (including signal detection and information-theoretic framing), and the integration of LDS-specific epistemological analysis.

ChatGPT (OpenAI GPT-5) was used as an editorial and developmental assistant. Its contributions were limited to:

  • Reorganizing and tightening prose for clarity and logical flow
  • Expanding arguments in directions explicitly suggested by the author
  • Recommending relevant academic literature and providing general summaries of established frameworks
  • Assisting in the drafting of example citations based on well-known, publicly available research
  • Suggesting structural refinements to improve coherence and accessibility

ChatGPT did not originate the central thesis, theoretical integration, or argumentative direction of the manuscript. All interpretive decisions, claims, framing choices, and final content determinations were made by the author. The author reviewed, modified, and approved all AI-assisted revisions and bears full responsibility for the content of the final manuscript.