STPA Analysis of Religion as a Social System¶
System-Theoretic Process Analysis Applied to the Systems Engineering Decomposition of Religion¶
Abstract¶
System-Theoretic Process Analysis (STPA) is a hazard analysis method developed by Nancy Leveson at MIT, based on the STAMP (Systems-Theoretic Accident Model and Processes) framework. STPA was designed for safety-critical engineering systems — aircraft, nuclear plants, medical devices — but its core logic applies wherever a complex system can produce unintended harmful outcomes through inadequate control rather than mere component failure. This paper applies STPA to religion, treated as a social system previously decomposed through a five-level systems engineering hierarchy. We define the system's losses, model its control structure, identify unsafe control actions across all four STPA categories, and trace loss scenarios to their causal factors. The result is a structural explanation of how religion — a system designed to provide meaning, morality, and community — can produce outcomes that are the exact opposite of its stated goals: meaninglessness, moral corruption, and social destruction.
1. Background: The SE Decomposition of Religion¶
1.1 The Five-Level Hierarchy¶
The following decomposition was established in prior work. It forms the input to the STPA analysis.
| Level | ID | Description |
|---|---|---|
| Goal | G1 | Provide existential meaning and purpose to human life |
| Goal | G2 | Establish moral and ethical orientation for individuals and communities |
| Goal | G3 | Create social cohesion and collective identity |
| Goal | G4 | Offer psychological comfort in the face of suffering and mortality |
| Req. | R1 | Coherent cosmological narrative (creation, afterlife, transcendence) |
| Req. | R2 | Framework for interpreting personal experience as meaningful |
| Req. | R3 | Codified behavioral norms (commandments, precepts, dharma) |
| Req. | R4 | Authority structure to interpret and enforce norms |
| Req. | R5 | Shared symbols, narratives, and practices for group bonding |
| Req. | R6 | Mechanism for distinguishing in-group from out-group |
| Req. | R7 | Rituals addressing grief, loss, and life transitions |
| Func. | F1 | Mythological narration — transmit origin and eschatological stories |
| Func. | F2 | Pastoral care — counsel individuals through crisis and life stages |
| Func. | F3 | Moral legislation — define permissible and forbidden behavior |
| Func. | F4 | Doctrinal governance — interpret scripture, settle disputes |
| Func. | F5 | Ritual performance — enact shared ceremonies |
| Func. | F6 | Missionary expansion — propagate belief system |
| Log. | L1 | Liturgical Subsystem — scripture, preaching, ceremonial calendar |
| Log. | L2 | Pastoral Subsystem — counseling, rites of passage, community support |
| Log. | L3 | Doctrinal/Juridical Subsystem — theology, canon law, ethics boards |
| Log. | L4 | Outreach Subsystem — evangelization, education, charity, media |
| Phys. | P1 | Churches, mosques, temples, synagogues, gurdwaras |
| Phys. | P2 | Sacred texts — Bible, Quran, Torah, Vedas, Tripitaka |
| Phys. | P3 | Clergy — priests, imams, rabbis, pastors, monks, gurus |
| Phys. | P4 | Parish networks, sanghas, congregations, umma structures |
| Phys. | P5 | Central authorities — Vatican, Al-Azhar, ecumenical councils |
| Phys. | P6 | Legal systems — canon law, fatwa apparatus, halakhic courts |
| Phys. | P7 | Missionary organizations, madrasas, Sunday schools |
| Phys. | P8 | Religious media — TV networks, podcasts, publishing houses |
1.2 Why STPA?¶
Traditional risk analyses ask: "What component can fail?" STPA asks a different question: "What control actions, if inadequate, can lead to losses — even when every component is working as designed?" This is precisely the right question for religion, where the most devastating outcomes (inquisitions, abuse cover-ups, radicalization, exclusion of vulnerable people) typically occur not because the system has "broken" but because its control mechanisms are operating exactly as designed — just in a way that produces harm.
2. STPA Step 1: Define Purpose of the Analysis¶
2.1 System-Level Losses¶
STPA begins by identifying losses — outcomes that the system's stakeholders would consider unacceptable. For religion as a social system, we define:
| Loss ID | Loss Description | Affected Goal |
|---|---|---|
| L1 | Individuals suffer existential despair, alienation, or loss of meaning because of the religious system (not despite it) | Negates G1 |
| L2 | Moral harm — the system produces, enables, or conceals immoral behavior (abuse, violence, corruption, exploitation) | Negates G2 |
| L3 | Social destruction — the system causes division, persecution, hatred, or war between groups | Negates G3 |
| L4 | Psychological damage — the system inflicts trauma, guilt, fear, or mental illness on individuals | Negates G4 |
| L5 | Loss of institutional legitimacy — the system loses the trust and authority it needs to function | Negates all |
| L6 | Suppression of truth — the system suppresses scientific knowledge, critical thinking, or individual conscience | Negates G1, G2 |
Note: each loss is the negation of a system goal. This is the defining feature of STPA applied to social systems — the hazards are not external threats but self-inflicted contradictions where the system's own control actions produce outcomes opposite to its stated purpose.
2.2 System-Level Hazards¶
Hazards are system states that, combined with worst-case environmental conditions, lead to losses. We derive them from the SE requirements:
| Hazard ID | Hazard Description | Related Losses |
|---|---|---|
| H1 | Doctrine becomes unfalsifiable and self-sealing — no internal mechanism can correct doctrinal errors | L1, L2, L6 |
| H2 | Authority structure becomes unaccountable — no effective oversight of those who interpret and enforce norms | L2, L4, L5 |
| H3 | In-group/out-group mechanism escalates to dehumanization of outsiders | L3 |
| H4 | Moral legislation becomes absolutist — no allowance for context, conscience, or evolving understanding | L1, L2, L4 |
| H5 | Pastoral care relationship is exploited — power asymmetry enables abuse | L2, L4, L5 |
| H6 | Missionary function prioritizes expansion over welfare of converts or target populations | L2, L3 |
| H7 | Financial resources are extracted from vulnerable populations without accountability | L2, L5 |
| H8 | Ritual performance becomes coercive — participation is compelled, not voluntary | L1, L4 |
2.3 System-Level Constraints¶
Each hazard implies a constraint — a condition the system must maintain to prevent the hazard:
| Constraint ID | Constraint | Prevents Hazard |
|---|---|---|
| SC1 | Doctrine must include mechanisms for self-correction, reinterpretation, and engagement with external knowledge | H1 |
| SC2 | Authority figures must be subject to transparent oversight, accountability, and removal procedures | H2 |
| SC3 | In-group/out-group distinctions must not extend to denial of outsiders' human dignity or rights | H3 |
| SC4 | Moral legislation must preserve space for individual conscience and contextual judgment | H4 |
| SC5 | Pastoral relationships must have safeguards: codes of conduct, mandatory reporting, external recourse | H5 |
| SC6 | Missionary activity must respect the autonomy and existing culture of target populations | H6 |
| SC7 | Financial management must be transparent, audited, and separated from spiritual authority | H7 |
| SC8 | Ritual participation must be voluntary, with no material or social penalty for non-participation | H8 |
3. STPA Step 2: Model the Control Structure¶
3.1 The STAMP Control Structure¶
STAMP models systems as hierarchical control structures where controllers issue control actions to controlled processes, and receive feedback. For religion:
┌─────────────────────────────┐
│ CENTRAL DOCTRINAL │
│ AUTHORITY (L3, P5, P6) │
│ Vatican / Al-Azhar / │
│ Ecumenical Councils │
│ │
│ Process Model: │
│ - Doctrine is correct │
│ - Faithful are obedient │
│ - External world is │
│ spiritually deficient │
└──────────┬──────────┬────────┘
Control Actions: │ │ Feedback:
- Doctrine │ │ - Reports from clergy
- Canon law │ │ - Synod deliberations
- Appointments │ │ - Petitions, appeals
- Excommunication│ │ - (often: self-confirming)
▼ │
┌─────────────────────────────┐
│ REGIONAL/LOCAL CLERGY │
│ (L1, L2, P3) │
│ Bishops, Imams, Rabbis, │
│ Pastors, Monks │
│ │
│ Process Model: │
│ - My authority comes from │
│ above │
│ - Congregation needs │
│ guidance │
│ - Doubt is dangerous │
└──────────┬──────────┬────────┘
Control Actions: │ │ Feedback:
- Sermons │ │ - Confession/counseling
- Pastoral rules │ │ - Attendance patterns
- Sacraments │ │ - Donations
- Social pressure│ │ - Complaints (often
- Shunning │ │ suppressed)
▼ │
┌─────────────────────────────┐
│ CONGREGATIONAL COMMUNITY │
│ (L1, L2, P4) │
│ Parish networks, sanghas, │
│ congregations │
│ │
│ Process Model: │
│ - Social norms are │
│ religiously grounded │
│ - Conformity = belonging │
│ - Dissent = exile │
└──────────┬──────────┬────────┘
Control Actions: │ │ Feedback:
- Social norms │ │ - Behavior observation
- Peer pressure │ │ - Gossip/reputation
- Inclusion/ │ │ - Attendance
exclusion │ │ - Participation
▼ │
┌─────────────────────────────┐
│ INDIVIDUAL BELIEVER │
│ │
│ Process Model: │
│ - Faith provides meaning │
│ - Clergy speak for God │
│ - Doubt is sinful │
│ - Suffering has purpose │
└─────────────────────────────┘
3.2 Why these control loops — and why only these?¶
The test for a control loop¶
STPA imposes three necessary conditions on every arrow in a control structure. All three must hold simultaneously; failure of any one disqualifies the relationship from being modelled as a control loop:
- Controller (authority condition): the upstream element issues observable, authoritative control actions — directives the downstream element is structurally expected to obey, not merely influence it may choose to consider.
- Controlled process (response condition): the downstream element's state is expected to change in response to those control actions as part of normal system operation.
- Feedback channel (closure condition): some signal — however weak, delayed, or filtered — travels back from the controlled process to the controller and can in principle update the controller's process model.
A relationship that passes all three tests is a control loop and must be modelled. A relationship that fails any one test is modelled differently: as peer coordination (aggregated into an existing feedback channel), as an environmental disturbance, or as a belief inside a controller's process model.
Applying the test to the SE decomposition¶
The four loops follow directly from applying these three conditions to the physical elements named in §1.1. The authority source that makes relationships authoritative rather than merely influential in every case is R4 ("authority structure to interpret and enforce norms") — a requirement the system itself specifies.
| Loop | Controller | Controlled Process | Condition 1: authority | Condition 2: state change | Condition 3: feedback |
|---|---|---|---|---|---|
| 1 | Central Authority (P5, P6) | Clergy (P3) | R4 + claimed divine mandate; clergy bound under canon law / fatwa / denominational discipline | Clergy doctrine, appointments, removal — clergy behaviour demonstrably changes in response | Reports from clergy, synod deliberations, petitions (filtered but structurally present) |
| 2 | Clergy (P3) | Congregation (P4) | R4 delegated downward; authority to preach, administer sacraments, impose social discipline | Congregation attendance, expressed belief, ritual participation change in response | Confession, counselling, donation levels, attendance — reach clergy as indirect behavioural signals |
| 3 | Congregation (P4) | Individual | R5, R6 (shared identity, in/out-group enforcement); peer authority over social standing and belonging | Individual's public behaviour, expressed belief, ritual participation change to maintain belonging | Gossip, reputation signals, observation of conformity or deviance feed back to community process model |
| 4 | Individual (self) | Own behaviour | Internalised R1–R3; conscience acts as an internal authority that the person is expected — by the system's own design — to obey | Own behaviour, choices, emotional state, and spiritual self-assessment change | Emotional feedback: guilt, consolation, doubt — conscience functions as internal sensor closing the loop |
These are the only four pairs in §1.1 where one element holds designated authority to issue control actions the other is structurally expected to obey. Remove any loop and the transmission path from doctrine to individual behaviour breaks — no alternative channel carries an authoritative control action all the way to the individual.
Why every other plausible relationship fails the test¶
The table below applies the three conditions to every candidate relationship commonly raised when this control structure is presented. In every case the candidate fails at least Condition 1.
| Candidate relationship | Fails condition | Reason |
|---|---|---|
| Clergy ↔ Clergy (lateral, same rank) | 1 — authority | Clergy of equal rank do not issue authoritative control actions to each other. Synods, councils, and theological deliberations are coordination inputs that aggregate into Loop 1's feedback channel, not independent loops. Where genuine rank differences exist (bishop over priest, abbot over monk), the relationship resolves to Loop 1 or Loop 2 at finer granularity — not a new, fifth loop. |
| Congregation ↔ Congregation (between parishes) | 1 — authority | Parish networks are parallel peers. No congregation holds designated authority over another. Binding coordination between congregations routes through Loop 1 — the central authority that owns both. Direct inter-congregational social pressure carries no structural control action. |
| Individual or congregation → Authority (upward) | 1 — authority | In the hierarchical institutional form this analysis targets, laity have no structural authority to issue control actions upward. Their upward channel is feedback, already represented in every loop. Note: congregationalist polities (Quaker, many Baptist, Reform Jewish) genuinely invert Loop 1 — authority flows from the congregation upward. Those constitute a genuinely different control structure under the same family name and require their own analysis. |
| God → Authority | 1 — authority | The system claims this loop, but divine control actions are not observably issued or measurably received. What actually operates is the Authority's belief that it is divinely controlled — a feature of the Authority's process model (shown inside the top box of the diagram), not a separate causal loop. STPA models what exists in the causal structure, not what is asserted. |
| Civil state / science / secular media | 1 — authority | These elements lie in the system's environment. They can disturb the system or supply external corrective signals, but they do not issue control actions the religion is architecturally bound to obey. They appear in §7 as externally imposed circuit breakers, not as internal control loops. |
| Other religions | 1 — authority | No cross-denominational structural authority exists by definition of denomination. Influence between traditions is cultural and bilateral; control is absent in both directions. |
| Schism / sect formation | Not a loop at all | This is a phase transition, not a control relationship. When Loop 1 fails to correct accumulating error, a fragment of the controlled process exits the hierarchy entirely and becomes the apex of a new control structure. Schism is a symptom of loop failure (analysed in §5), not a sixth relationship within the existing hierarchy. |
| The upward feedback arrows on each loop | Not a new loop | Every loop already includes a feedback channel (the upward arrows in the diagram). Those arrows satisfy Condition 3 of their respective loop. They do not satisfy Condition 1 of a new loop — feedback is not control. Treating them as control would mean the congregation governs the clergy and the clergy governs the central authority, which is structurally false in the institutional form under analysis. |
A different decomposition — congregational polity, decentralised charismatic movement, household-religion — would yield a different control structure with different loops. Within the hierarchical institutional decomposition analysed here, no fifth loop exists, because no fifth pair of elements satisfies all three conditions simultaneously.
3.3 Key Observations about the Control Structure¶
Feedback poverty at the top. The Central Doctrinal Authority receives feedback primarily through its own appointees (clergy), who have incentives to report compliance rather than problems. External feedback (scientific findings, secular criticism, abuse reports) is structurally classified as hostile input, not corrective signal.
Asymmetric power at every level. Each controller has significantly more power over the controlled process than the controlled process has to send corrective feedback upward. This is by design (R4: authority to interpret and enforce), but it means that control errors propagate downward unchecked.
Self-confirming process models. At every level, the controller's mental model of the controlled process contains built-in resistance to disconfirmation. If a believer doubts, the model says "doubt is temptation" — the model explains away the very feedback that would correct it.
4. STPA Step 3: Identify Unsafe Control Actions¶
STPA categorizes unsafe control actions (UCAs) into four types:
- Not providing a control action that is needed
- Providing a control action that causes a hazard
- Providing a control action too early, too late, or out of sequence
- Providing a control action that is applied too long or stopped too soon
4.1 UCAs for Central Doctrinal Authority → Clergy¶
| UCA ID | Type | Unsafe Control Action | Hazard |
|---|---|---|---|
| UCA-1 | Not provided | Authority does NOT update doctrine when scientific evidence invalidates cosmological claims (e.g., geocentrism, age of earth, evolution) | H1 |
| UCA-2 | Not provided | Authority does NOT remove or discipline clergy credibly accused of abuse | H2, H5 |
| UCA-3 | Provided | Authority ISSUES doctrine that dehumanizes outsiders (infidels, heretics, apostates as subhuman or damned) | H3 |
| UCA-4 | Provided | Authority EXCOMMUNICATES or PUNISHES members who raise legitimate doctrinal questions | H1, H4 |
| UCA-5 | Provided | Authority APPOINTS clergy based on doctrinal loyalty rather than pastoral competence or moral character | H2, H5 |
| UCA-6 | Too late | Authority CORRECTS a doctrinal error only after centuries of harm (e.g., Galileo rehabilitation in 1992, 359 years late) | H1, H6 |
| UCA-7 | Too long | Authority MAINTAINS absolutist moral prohibition long after ethical understanding has evolved (e.g., blanket prohibitions that cause measurable suffering) | H4 |
| UCA-8 | Not provided | Authority does NOT establish transparent financial auditing of religious institutions | H7 |
4.2 UCAs for Clergy → Congregation¶
| UCA ID | Type | Unsafe Control Action | Hazard |
|---|---|---|---|
| UCA-9 | Provided | Clergy USES pastoral relationship to sexually, emotionally, or financially exploit vulnerable individuals | H5 |
| UCA-10 | Provided | Clergy PREACHES hatred or contempt toward out-groups (other religions, LGBTQ+ persons, ethnic minorities, apostates) | H3 |
| UCA-11 | Provided | Clergy WITHHOLDS sacraments or pastoral care as punishment for questioning doctrine | H1, H4, H8 |
| UCA-12 | Not provided | Clergy does NOT provide pastoral support during genuine crisis (grief, mental illness, abuse) because doctrinal framework has no model for the problem | H4 |
| UCA-13 | Provided | Clergy DEMANDS financial contributions from impoverished believers through spiritual coercion (prosperity gospel, indulgences, obligatory tithing) | H7 |
| UCA-14 | Too long | Clergy MAINTAINS shunning of excommunicated member long after the person has suffered disproportionate social death | H4, L3 |
| UCA-15 | Not provided | Clergy does NOT report abuse to civil authorities, instead handling it "internally" | H2, H5 |
4.3 UCAs for Congregation → Individual¶
| UCA ID | Type | Unsafe Control Action | Hazard |
|---|---|---|---|
| UCA-16 | Provided | Congregation SHUNS member who expresses doubt, questions doctrine, or leaves the faith | H1, H4, H8 |
| UCA-17 | Provided | Congregation ENFORCES social conformity on matters of personal conscience (dress, diet, relationships, career choices) through gossip and exclusion | H4, H8 |
| UCA-18 | Not provided | Congregation does NOT intervene when they observe clergy abusing a member | H2, H5 |
| UCA-19 | Provided | Congregation PRESSURES children into public faith commitments (baptism, confirmation, bar mitzvah) before they can meaningfully consent | H8 |
| UCA-20 | Too early | Congregation LABELS children as sinful, damned, or spiritually deficient at a developmentally inappropriate age | H4 |
4.4 UCAs for Individual (Self-Control / Internalized Beliefs)¶
| UCA ID | Type | Unsafe Control Action | Hazard |
|---|---|---|---|
| UCA-21 | Provided | Individual SUPPRESSES legitimate doubt, curiosity, or moral intuition because they have internalized "doubt = sin" | H1, H6 |
| UCA-22 | Provided | Individual REMAINS in an abusive religious environment because they believe leaving = damnation | H4, H5 |
| UCA-23 | Not provided | Individual does NOT seek professional mental health support because the religious framework teaches that faith alone should suffice | H4 |
| UCA-24 | Provided | Individual REJECTS medical treatment for self or dependents based on doctrinal prohibition | H4, L1 |
| UCA-25 | Too long | Individual CONTINUES religious practices that cause measurable psychological harm (extreme fasting, self-flagellation, sleep deprivation) beyond any devotional purpose | H4 |
5. STPA Step 4: Identify Loss Scenarios¶
Each UCA has causal factors — the reasons why the unsafe control action occurs. STPA organizes these into two categories:
A. Why would the controller issue the UCA? (Flawed process model, incorrect feedback, conflicting goals)
B. Why would the control action not be executed properly? (Communication failure, implementation breakdown)
5.1 Loss Scenario LS-1: Doctrine Fails to Self-Correct (UCA-1, UCA-6 → H1 → L1, L6)¶
The scenario: Central authority maintains a cosmological or moral doctrine that contradicts established scientific evidence or evolved ethical understanding. Believers experience cognitive dissonance between their faith and their knowledge, leading to either intellectual suppression (L6) or loss of faith and meaning (L1).
Why the controller issues this UCA:
- Flawed process model: The authority's model holds that doctrine is divinely revealed and therefore cannot be wrong. This makes self-correction logically impossible within the model — any correction would imply the divine source was wrong.
- Conflicting goals: Correcting doctrine threatens the authority's legitimacy (SC1 conflicts with R4). If the authority was wrong about X, what else might it be wrong about? The authority perceives self-correction as existentially threatening.
- Inadequate feedback: Believers who experience the dissonance often leave silently rather than escalating. The authority receives feedback only from those who stay — a survivorship bias that confirms the model.
- Temporal delay: By the time evidence is overwhelming, the authority has centuries of institutional commitment to the doctrine. Reversal costs increase with time (sunk cost fallacy at institutional scale).
Structural fix (from SC1): Build doctrinal revision mechanisms into the system. Historical examples: Vatican II (1962–65) as a partial self-correction mechanism. Judaism's tradition of Talmudic disputation, which structurally preserves minority opinions. Islam's concept of ijtihad (independent reasoning), when not suppressed. The key architectural principle: doctrine should be versionable, with explicit revision processes, not immutable.
5.2 Loss Scenario LS-2: Abuse Cover-Up (UCA-2, UCA-15 → H2, H5 → L2, L4, L5)¶
The scenario: A clergy member abuses a congregant (sexually, emotionally, financially). Other clergy and the authority become aware but do not remove the abuser or report to civil authorities. Instead, the abuser is transferred, the victim is silenced, and the institution protects its reputation. This is the pattern documented in the Catholic Church, the Southern Baptist Convention, Jehovah's Witnesses, and numerous other contexts.
Why the controller issues this UCA:
- Flawed process model: The authority's model holds that (a) clergy are called by God and therefore inherently trustworthy, (b) the institution's reputation is essential to its mission, and (c) internal resolution is preferable to secular intervention. All three beliefs are structurally embedded in R4 and the control structure.
- Conflicting goals: Protecting the institution (G3: social cohesion, R4: authority structure) conflicts with protecting the individual (G2: moral orientation, G4: comfort). The control structure systematically prioritizes the institution because the controllers are the institution.
- Feedback suppression: The victim's feedback (complaint, report) enters the system through the very structure that has an interest in suppressing it. There is no independent feedback channel that bypasses the authority hierarchy.
- Accountability void: The controller who should discipline the abuser (bishop, denominational leader) is also the person whose career and institution are damaged by disclosure. The judge and the interested party are the same entity.
Structural fix (from SC2, SC5): (a) Mandatory external reporting to civil authorities — removing the internal-only handling option. (b) Independent ombudsman or safeguarding board with authority to investigate and sanction, staffed by people outside the clerical hierarchy. (c) Separation of pastoral authority from disciplinary authority — the person who counsels the victim must not be the person who decides whether to report. (d) Statute of limitations reform within canon law. These are architectural changes to the control structure itself, not mere policy adjustments.
5.3 Loss Scenario LS-3: Radicalization through In-Group/Out-Group Escalation (UCA-3, UCA-10 → H3 → L3)¶
The scenario: The mechanism for creating social cohesion (R5, R6: shared identity and in-group/out-group distinction) escalates from "we are different from them" to "they are inferior" to "they are dangerous" to "they must be destroyed." This progression has driven religious wars, pogroms, crusades, inquisitions, and modern terrorism.
Why the controller issues this UCA:
- Flawed process model: The in-group/out-group mechanism (R6) has no built-in limit. The same mechanism that creates belonging ("we share this faith") also creates hostility ("they threaten our faith"). The model does not distinguish between "different" and "dangerous."
- Positive feedback loop: Preaching against out-groups increases in-group solidarity (measurable by attendance, donations, emotional intensity). This positive feedback rewards escalation. The more extreme the rhetoric, the stronger the group cohesion — until the rhetoric produces violence.
- External trigger: The escalation is often catalyzed by external threat (real or perceived). Economic stress, political marginalization, or military conflict activates the in-group/out-group mechanism and pushes it toward its extreme.
- Missing control action: There is no structural "circuit breaker" that de-escalates when rhetoric approaches dehumanization. The authority may agree with the escalation, or may be too distant to intervene in local radicalization.
Structural fix (from SC3): (a) Explicit doctrinal commitment to universal human dignity that overrides in-group/out-group distinctions — structurally, this means SC3 must be higher-priority than R6 in the requirements hierarchy. (b) Interfaith engagement as a standing function, not an exceptional one — regular contact with out-groups reduces dehumanization (contact hypothesis). (c) Internal monitoring for escalatory rhetoric, with de-escalation authority assigned to a specific role.
5.4 Loss Scenario LS-4: Spiritual Coercion and Psychological Damage (UCA-11, UCA-16, UCA-17, UCA-20 → H4, H8 → L1, L4)¶
The scenario: An individual (often a child or young person) is subjected to religious teaching that instills deep fear (hellfire, damnation, divine punishment), shame (sinfulness, unworthiness), or guilt (for normal developmental experiences like doubt, sexual feelings, or intellectual curiosity). Social enforcement (shunning, withdrawal of love, public shaming) reinforces the psychological impact. The individual develops religious trauma, anxiety disorders, or complex PTSD.
Why the controller issues this UCA:
- Flawed process model: The clergy and congregation model holds that (a) fear of divine punishment is a legitimate motivator, (b) shame is a corrective emotion, and (c) social exclusion of deviants protects the community. These beliefs are embedded in R3 (codified norms), R6 (in-group/out-group), and the enforcement functions (F3, F4).
- Missing feedback: Children and psychologically dependent adults cannot provide effective corrective feedback to the system. They lack the power, the vocabulary, and often the awareness that their experience is harmful. By the time they can articulate the damage, they may have left the system entirely — at which point their feedback is classified as "apostate hostility."
- Normalization: When an entire community shares the same beliefs and practices, harmful behaviors are invisible as harmful. Telling a child they are "born in sin" feels like education, not psychological abuse, to a community that genuinely believes it.
Structural fix (from SC4, SC8): (a) Age-appropriate religious education that does not use fear or shame as motivational tools — this requires explicit pedagogical standards within the religious education subsystem. (b) Voluntary participation norms: no social penalty for non-attendance, non-belief, or questioning. (c) Mental health literacy within pastoral training — clergy should be trained to recognize when religious practice is causing psychological harm and to refer to professional support. (d) External child safeguarding standards applied to religious education, just as they are applied to secular schools.
5.5 Loss Scenario LS-5: Financial Exploitation (UCA-8, UCA-13 → H7 → L2, L5)¶
The scenario: Religious authority extracts financial resources from believers through spiritual coercion — "God wants you to give," "tithing is obligatory," "sowing seeds of faith will bring financial blessings" (prosperity gospel). Funds are not transparently accounted for. Leaders live in luxury while congregants struggle.
Why the controller issues this UCA:
- Conflicting goals: The system needs financial sustainability (R7 in the SE decomposition) but has no structural separation between spiritual authority and financial management. The person who tells you what God wants is also the person who tells you how much to give.
- Missing oversight: Unlike secular nonprofits, many religious organizations are exempt from financial reporting requirements in numerous jurisdictions. The control structure has no external audit function.
- Doctrinal justification: The prosperity gospel and similar doctrines create a self-sealing system: if you give and prosper, it proves the doctrine; if you give and don't prosper, your faith was insufficient — give more.
Structural fix (from SC7): (a) Mandatory financial transparency — published annual reports, independent external audit, separation of spiritual authority from financial authority. (b) Congregation-elected financial oversight body with real power (a "Kassenprüfung" model, borrowed from the German Verein). (c) Doctrinal prohibition on linking financial giving to spiritual outcomes.
5.6 Loss Scenario LS-6: Suppression of Individual Conscience (UCA-4, UCA-7, UCA-21 → H1, H4 → L1, L6)¶
The scenario: A believer's individual moral conscience or intellectual understanding conflicts with official doctrine. The system's response is to demand submission of conscience to authority. The individual either suppresses their authentic moral judgment (psychological harm, L4) or is punished for expressing it (exclusion, L1). In either case, the system has sacrificed one of its own goals (G2: moral orientation) to preserve another (R4: authority structure).
Why the controller issues this UCA:
- Architectural conflict: R4 (authority to interpret and enforce) and G2 (moral orientation) are in structural tension. R4 was designed to serve G2 (authority enables moral guidance), but R4 can become an end in itself — authority for authority's sake — at which point it overrides G2.
- Process model deficiency: The authority's model does not contain the possibility that the individual's conscience could be more morally correct than the doctrine. The model assumes doctrinal infallibility, which means any dissent is by definition error.
Structural fix (from SC1, SC4): (a) Explicit doctrinal recognition of conscience as a valid moral source — Catholicism actually has this (Dignitatis Humanae, Vatican II) but struggles to operationalize it when conscience contradicts doctrine. (b) Structured dissent mechanisms: the ability to formally register disagreement without excommunication, analogous to a "dissenting opinion" in a court ruling. (c) Periodic doctrinal review processes that engage lay members and external scholars, not just clergy.
6. Cross-Cutting Structural Findings¶
6.1 The Feedback Problem Is Universal¶
Every loss scenario shares a common structural deficiency: inadequate upward feedback. The control structure of religion is designed for top-down transmission (doctrine flows from authority to believer) but has weak, filtered, or suppressed bottom-up feedback. This is not a bug but a feature of the design (R4: authority to interpret and enforce). However, it means that control errors are systematically uncorrected.
In control systems engineering, this is called open-loop control — the controller acts without adequate measurement of the actual system state. Open-loop control works only when the process is perfectly predictable. Human societies are never perfectly predictable. Therefore, open-loop control of human behavior through religion will always produce drift toward loss conditions.
The architectural remedy is to close the loop: create genuine, unfiltered feedback channels from the controlled process (believers, congregations) to the controller (authority), with structural protections ensuring the feedback cannot be suppressed.
6.2 The Self-Sealing Model Problem¶
Multiple UCAs are enabled by self-sealing process models — mental models held by controllers that explain away any evidence that would contradict them:
| Self-Sealing Pattern | How It Works |
|---|---|
| "Doubt is sin" | Eliminates the primary corrective signal (questioning) |
| "Suffering is divine test" | Reframes harm as benefit — the worse it gets, the more it "proves" the model |
| "External criticism is persecution" | Classifies corrective feedback from outside the system as hostile action |
| "Apostates were never true believers" | Discounts the testimony of the people with the most relevant experience |
| "God's ways are beyond human understanding" | Makes the model unfalsifiable by definition |
In control engineering, a process model that cannot be updated by feedback is called unobservable. An unobservable system cannot be controlled safely. The architectural remedy is to ensure that every process model in the control structure includes explicit update mechanisms — conditions under which the model itself is revised.
6.3 The Conflicting-Goals Problem¶
The SE decomposition reveals that several system goals are in structural tension:
| Goal Pair | Tension |
|---|---|
| G2 (moral orientation) vs. R4 (authority) | Authority enables morality but can also override conscience |
| G3 (social cohesion) vs. R6 (in-group/out-group) | Cohesion requires boundaries, but boundaries enable exclusion and hatred |
| G4 (comfort) vs. R3 (norms) | Comfort requires unconditional acceptance, but norms require conditional compliance |
| G1 (meaning) vs. R1 (cosmological narrative) | Meaning requires honest engagement with reality; a fixed narrative may conflict with reality |
These tensions are not design flaws — they are inherent in any system that simultaneously serves individual and collective needs. The design flaw is when the system has no mechanism for managing the tension: no process for deciding which goal takes priority when they conflict, and no authority empowered to make that trade-off transparently.
6.4 The rung mismatch behind the four findings¶
The previous chapter Justificatory Rungs introduced a seven-rung ladder of standards under which control actions and feedback signals are accepted. Reading §§6.1–6.3 through that lens, rung mismatch subsumes them all for this system:
- "Feedback poverty at the top" is the rung-1 controller refusing rung-3 input — Pattern A (asymmetric loop). The rung-1 hierarchy classifies abuse data, scientific evidence, and demographic harm as rung-1 hostility and filters them out.
- "Self-confirming process models" is the controller's acceptance-rung filter operationalised. Each entry in the table in §6.2 ("doubt is sin," "suffering is divine test," "external criticism is persecution," "apostates were never true believers," "God's ways are beyond human understanding") is a specific rule for re-classifying rung-3+ feedback as rung-1 noise.
- "Conflicting goals" are tractable when the system can deliberate at rung 6 about how to balance them; in a rung-1 system, whichever side has more authority simply wins.
The system simultaneously runs Pattern B — Claimed-Rung Inflation: it claims rung 6 (sacralised ultimate legitimacy) while operating at rung 1 (clerical authority). The gap is enforced doctrinally rather than closed, which is why the failures are predictable and recurrent.
The architectural fixes in §7 are all attempts to insert rung-3 channels that route around the rung-1 filter without forcing the controller to abandon its rung-1 self-understanding. Mandatory civil reporting, independent ombudspersons, lay safeguarding boards, doctrinal version control with external scholarship, and pre-registered scientific-consultation processes are all implementations of the same architectural pattern.
This is the single most important structural finding of the analysis: religion's harmful outcomes are not random malfunctions but the predictable consequence of operating a sacralised rung-1 control architecture while claiming the authority of rung 6. The canonical reference, including the rung-tagged control structure and the rung-acceptance filter as a recurring causal factor in every loss scenario, is in knowledge/system-catalogues/social-systems/religion/applied-se-analysis.md.
7. Recommendations: Architectural Improvements¶
Based on the STPA analysis, we propose the following changes to religion's control structure. These are stated as architectural principles, applicable across traditions:
7.1 Close the Feedback Loop¶
- Establish independent feedback channels that bypass the authority hierarchy: ombudsman, external review boards, anonymous reporting systems.
- Require periodic surveys of congregational wellbeing — measuring psychological health, not just attendance and donations.
- Create formal exit interviews for members who leave, treated as system feedback rather than apostasy.
7.2 Make Process Models Updatable¶
- Institute doctrinal version control — every doctrine should have a stated scope, a review date, and a revision process.
- Require engagement with external knowledge (science, psychology, ethics) as a standing input to doctrinal governance, not an exceptional event.
- Preserve dissenting opinions in the doctrinal record, as courts preserve minority opinions.
7.3 Separate Control Functions¶
- Separate spiritual authority (preaching, sacraments, counseling) from administrative authority (finances, personnel, discipline).
- Separate pastoral care from norm enforcement — the person who counsels you should not be the person who punishes you.
- Establish independent safeguarding with authority to investigate and override clergy decisions on abuse cases.
7.4 Install Circuit Breakers¶
- Define escalation limits on in-group/out-group rhetoric: explicit doctrinal commitment to universal human dignity that cannot be overridden by any other teaching.
- Establish proportionality review for disciplinary actions: shunning, excommunication, and other exclusion mechanisms should require documented justification, appeal process, and time limits.
- Require informed consent for participation in intensive religious practices (extended fasting, retreats, exorcism rituals) with right of withdrawal.
7.5 Protect Vulnerable Populations¶
- Apply secular child safeguarding standards to all religious education and youth activities.
- Prohibit fear-based instruction (hellfire, damnation) for children below a developmentally appropriate age.
- Require mandatory external reporting of all abuse allegations to civil authorities, with no exceptions for internal handling.
8. Conclusion¶
STPA reveals that religion's most harmful outcomes are not random malfunctions but predictable consequences of its control architecture. A system designed for top-down doctrinal transmission with weak upward feedback, self-sealing process models, unaccountable authority, and no structural mechanism for managing goal conflicts will, over sufficient time, produce every one of the losses we identified — regardless of the sincerity or goodness of the people within it.
This is not an indictment of religion's purpose. The goals — meaning, morality, community, comfort — are among the most important things humans seek. It is an indictment of religion's control architecture as historically implemented. The same goals could be served by a system with closed feedback loops, updatable doctrines, separated authority, independent oversight, and protected individual conscience. Some religious traditions have moved in this direction; none has fully arrived.
The value of STPA is that it moves the conversation from "religion is good" vs. "religion is harmful" — a debate that generates more heat than light — to a structural question: which control actions, under which conditions, produce which outcomes? That question has specific, actionable answers. The fixes are architectural, not theological. They do not require abandoning faith. They require redesigning the system that carries it.
This analysis applies STPA (Leveson, 2012) to the SE decomposition of religion developed in prior work. The four STPA steps — define losses, model control structure, identify UCAs, trace loss scenarios — are followed systematically. All hazard and constraint IDs trace back to the original SE decomposition's goals and requirements.
References¶
- Leveson, N.G. (2012). Engineering a Safer World: Systems Thinking Applied to Safety. MIT Press.
- Leveson, N.G. (2004). "A New Accident Model for Engineering Safer Systems." Safety Science, 42(4), 237–270.
- Leveson, N.G. & Thomas, J.P. (2018). STPA Handbook. MIT Partnership for Systems Approaches to Safety and Security.