Re-Review: b12-socpsy MMv3 (2026m04d06 revision)#

Note

Reviewer role: Developmental psychologist (same as original review). Model: Claude Opus 4.6 (ClaudeOp46Max) at max effort. Date: 2026m04d06. Paper reviewed: b12-socpsy_mmv3_2026m04d06.rst (MMv3-SocPsy). Original review: review_b12-socpsy_2026m04d05.rst (11 issues: 4 S3 BREACH, 5 S2 HELD, 2 S1 HELD). Author reply: reply_b12-socpsy_2026m04d05.rst (all 11 ACCEPT). Method: Re-review of the revised paper. For each original issue, assess whether the revision addresses it. Then conduct a fresh review of the revised paper as a whole. Verdicts use HELD/BREACH per CLAUDE.md Language Rule 5.

You are a developmental psychologist with expertise in Erikson, Maslow, Kohlberg, metacognition research, and clinical assessment instruments. You are reviewing a REVISED paper that claims to have addressed your prior review. Your job is to test whether the revisions actually resolve the issues or merely rearrange the same problems.

Read these files in order:

.claude/CLAUDE.md — project rules, language rules, EDEN system.
source/matheology/hell/ll/study/b/12/review_b12-socpsy_2026m04d05.rst — your ORIGINAL review (the one this revision responds to).
source/matheology/hell/ll/study/b/12/reply_b12-socpsy_2026m04d05.rst — the author reply with all decisions.
source/matheology/hell/mm/b/12/mmv3/b12-socpsy_mmv3_2026m04d06.rst — the REVISED paper (MMv3). THIS IS WHAT YOU ARE REVIEWING.
source/matheology/hell/mm/b/12/mmv3/b12-math_mmv3_2026m04d05.rst — the formal math paper (for axiom cross-checks).

For EACH issue from your original review, answer:

Was the issue addressed? (YES / PARTIALLY / NO)
Is the fix adequate? If not, what remains?
Did the fix introduce NEW problems?

Then conduct a FRESH review of the revised paper as a whole. Look for:

RESTRUCTURING: The paper now leads with own contributions. Does this work? Are the contributions clear and well-argued on their own, or do they still depend on the parallels for persuasive force?
DUNNING-KRUGER REFRAMING: The paper now claims D-K is “one instance” of a broader OK-closure mechanism. Is this more honest? Is it actually supported, or is it still overclaiming from the other direction (“we discovered the general case”)?
SUPERVILLAIN THEOREM: Now framed as a “risk factor” with a “conjunction condition.” Does this address the selection bias and testability issues? Is the “systematic production” mechanism (suppression of uniqueness) empirically grounded or speculative?
FIVE-GATE MODEL: The sequential structure is now prominent. Is the operationalization table adequate? Are there gaps? Does Gate 3 now have enough substance?
PARALLELS SECTION: Tuckman now leads. Erikson has an honest stage-by-stage table. Maslow has caveats. Kohlberg/Bloom are “suggestive analogies.” Is this calibration correct? Too generous anywhere? Too dismissive?
NEW CONTENT: Check the cognitive dissonance prediction (Section 2.3), the Kohlberg moral-regression prediction (Section 4.4), and the operationalization table (Section 3.7) for accuracy and testability.

Use HELD/BREACH verdicts. Severity: S1 minor, S2 moderate, S3 serious, S4 critical. Use “test”/”check” not “validate”/”verify”. Use YYYYmMMdDD dates.

PRODUCE a review report. Save at: source/matheology/hell/ll/study/b/12/review_b12-socpsy-mmv3_2026m04d06.rst

Want me to run it?

Executive Summary#

The MMv3 revision is a substantial and largely successful restructuring. The four S3 BREACH issues from the original review have all been addressed — three fully, one with a residual concern. The paper now leads with its own contributions (OK-closure mechanism, supervillain theorem, five-gate model), uses existing theories as comparison points rather than convergence evidence, and calibrates its claims with significantly more honesty.

The restructuring works. The paper is now readable as an independent contribution rather than as a claim of convergence with established theories. The Dunning-Kruger reframing is more honest. The supervillain theorem is more precise. The five-gate model’s novel claim (sequential structure) is now prominent and operationalized.

However, the revision introduces two new issues and leaves one residual concern from the original review. These are moderate (S2), not serious — the paper has moved from “structurally problematic” to “needs polish.”

Overall assessment: The MMv3 revision is a strong improvement. The paper has moved from Grey Meadow to approaching Green Meadow.

Part I: Issue-by-Issue Re-Assessment#

1.1 Three-Feature Overlap is Generic (was S3, BREACH)#

A. Addressed? YES.

B. Is the fix adequate? Yes. Section 4.2 now explicitly states: “these three features are generic properties of staged developmental models, not specific structural signatures.” The paper credits the original review for identifying this. The three-feature overlap is no longer presented as evidence of convergence. The stage-by-stage comparison table replaces it with honest assessments at each stage.

C. New problems? No. The table’s parallel-strength ratings (Weak, Approximate, Strong) are well-calibrated. Stages 7–8 are labeled Strong; Stages 1–2 are labeled Weak; Stages 3–6 are Approximate. This matches the evidence.

New verdict: HELD. Issue resolved.

1.2 The Ordering Problem (was S3, BREACH)#

A. Addressed? YES.

B. Is the fix adequate? Yes. Section 4.2 now explicitly calls the ordering reversal “a fundamental structural difference” and states that “these are opposite theoretical commitments about trust’s role.” The paper no longer dismisses this as a domain-dependent detail. The summary sentence at the end of the Erikson section is well-calibrated: “Two specific Erikson stages (7 and 8) have strong and specific parallels… The remaining stages show approximate thematic resonances but not structural convergence. The overall stage architectures are fundamentally different.”

C. New problems? No.

New verdict: HELD. Issue resolved.

3.2 Selection Bias in Supervillain Theorem (was S3, BREACH)#

A. Addressed? YES.

B. Is the fix adequate? Yes. Section 3.5 now:

Restates the theorem as a “risk factor (not a law).”
Specifies the conjunction condition (frozen scope + high influence).
Includes disconfirming cases (Mandela, Eisenhower) with explanations.
Acknowledges the base rate problem explicitly.
Provides testable research designs and operationalized variables.

The revision is thorough and honest. The conjunction condition is a genuine improvement — it transforms a vague universal claim into a specific, falsifiable risk factor.

C. New problems? One residual concern (see New Issue N1 below regarding the “systematic production” mechanism).

New verdict: HELD. Core issue resolved; new subsidiary issue raised.

4. Dunning-Kruger Generalization (was S3, BREACH)#

A. Addressed? YES.

B. Is the fix adequate? Largely yes. Section 1 (Introduction) now frames the relationship as “Dunning-Kruger is a special case” rather than “we generalize Dunning-Kruger.” Section 2.2 presents the mechanism across competence levels with appropriate distinctions:

Low competence: metacognitive deficit (Dunning-Kruger).
Moderate competence: effort reduction (deliberate practice plateau).
High competence: self-licensing (earned dogmatism).
High competence + changing domain: BABL risk (supervillain theorem).

This is more honest and more accurate than the original framing. The distinction between stagnation and self-destruction is now explicit: stagnation at moderate competence, BABL only under conjunction conditions at high competence.

C. New problems? One residual concern. The paper still claims “the e7Day model formalizes the common mechanism” across all levels (Section 2.2, final paragraph). This is a weaker form of the original overclaim. The “common mechanism” is OK-closure, but the psychological processes producing OK-closure are different at each level (metacognitive deficit, effort reduction, self-licensing). The e7Day model describes a common outcome structure (OK stops correction), not a common mechanism. This is a terminological precision issue (S1), not a structural problem.

New verdict: HELD. Core issue resolved. Minor terminological residue (S1).

2. Maslow Parallel (was S2, HELD)#

A. Addressed? YES.

B. Is the fix adequate? Yes. Section 4.3 now:

Lists all four of Maslow’s own caveats (partial satisfaction, exceptions, cultural variation, later revision).
Presents the “productive difference” between needs (concurrent) and construction stages (ordered).
Provides a mapping-quality table with honest ratings (Generic, Strained, Reasonable, Strong).
Does not overclaim convergence.

C. New problems? No.

New verdict: HELD. Issue resolved.

3.1 Supervillain Theorem Testability (was S2, HELD)#

A. Addressed? YES.

B. Is the fix adequate? Yes. Section 3.5 now includes:

Three testable research designs (longitudinal leadership, expert overconfidence tracking, historical paired comparisons).
Operationalization of the three key variables (“stops expanding compassion scope,” “retained influence,” “harm from frozen expertise”) with specific proxy measures.

This is a genuine improvement. The theorem is now accompanied by enough operational detail that a researcher could design a study.

C. New problems? No.

New verdict: HELD. Issue resolved.

5.1 Five-Gate Operationalizability (was S2, HELD)#

A. Addressed? YES.

B. Is the fix adequate? Yes. Section 3.7 provides the requested operationalization table mapping each gate to existing instruments. The table includes PTGI, MBI, MITI, WAI, Barrett-Lennard, CPD hours, openness-to-experience, and intellectual humility scales. The final paragraph specifies what the framework adds beyond individual instruments (sequential gate structure).

Gate 1 is reframed around “overcoming” rather than mere “survival,” with explicit acknowledgment that therapist training is a structured form of overcoming. The clinical literature context is engaged (Section 3.1): “Therapist effectiveness correlates with alliance quality and technique mastery, not just personal problem-history.”

C. New problems? No.

New verdict: HELD. Issue resolved.

5.2 Sequential Gate Structure (was S2, HELD)#

A. Addressed? YES.

B. Is the fix adequate? Yes. The sequential structure is now the lead claim of Section 3 (“the sequential structure is its primary novel contribution”). The section opens with the sequential claim and provides examples of what earlier-gate failure implies for later gates.

C. New problems? No.

New verdict: HELD. Issue resolved.

6.3 Kohlberg and Bloom (was S2, HELD)#

A. Addressed? YES.

B. Is the fix adequate? Yes. Section 4.4 explicitly labels both as “suggestive analogies” rather than structural parallels. The paper acknowledges: “These analogies are suggestive but under-argued.”

A new Kohlberg prediction is added (moral regression under stress as OK-closure; Section 4.4). This is a good start but see New Issue N2 below.

C. New problems? See N2.

New verdict: HELD. Issue resolved.

6.1 Cognitive Dissonance Reframing (was S1, HELD)#

A. Addressed? YES.

B. Is the fix adequate? Yes. Section 2.3 develops the connection with a specific testable prediction: “individuals scoring high on tolerance of ambiguity and low on need for cognitive closure should display more NOT-OK-like behavior across domains.” Specific instruments are cited (Need for Cognitive Closure scale, Tolerance of Ambiguity scale). The prediction is clean and testable.

C. New problems? No.

New verdict: HELD. Issue resolved.

6.2 Tuckman Parallel Placement (was S1, HELD)#

A. Addressed? YES.

B. Is the fix adequate? Yes. Section 4.1 now leads the parallels section. The heading explicitly labels it “Strongest Parallel.” The Storming = EQUAL mapping is well-argued with the key insight preserved: “Storming has no ‘it was good’ verdict.”

C. New problems? No.

New verdict: HELD. Issue resolved.

Part II: Issue Resolution Summary#

Issue	Original	Addressed?	New Verdict	Notes
1.1	S3 BREACH	YES	HELD	Stage-by-stage table replaces generic features.
1.2	S3 BREACH	YES	HELD	Ordering reversal acknowledged as fundamental.
3.2	S3 BREACH	YES	HELD	Restated as risk factor. Conjunction condition specified.
4	S3 BREACH	YES	HELD	D-K reframed as special case. Minor terminological residue (S1).
2	S2 HELD	YES	HELD	Maslow’s caveats engaged. Mapping quality rated honestly.
3.1	S2 HELD	YES	HELD	Research designs and operationalized variables provided.
5.1	S2 HELD	YES	HELD	Operationalization table added. Gate 1 reframed.
5.2	S2 HELD	YES	HELD	Sequential structure now leads Section 3.
6.3	S2 HELD	YES	HELD	Relabeled as suggestive analogies.
6.1	S1 HELD	YES	HELD	Testable prediction with specific instruments.
6.2	S1 HELD	YES	HELD	Tuckman now leads parallels section.

All 11 original issues addressed. All now HELD. Four former BREACHes converted to HELD. No original issue worsened.

Part III: Fresh Review of the Revised Paper#

F1. Restructuring Assessment#

Verdict: HELD.

The restructuring works. Sections 2–3 present the paper’s own contributions (OK-closure mechanism, five-gate model, supervillain theorem) without depending on the parallels. A reader who skips Section 4 entirely would still understand the paper’s claims and could evaluate them on their own merits.

The reading guide (Section 1.2) is helpful: “This ordering is deliberate: the model was developed independently of the psychological literature, and its claims stand or fall on their own merits.”

The parallels in Section 4 now function as comparison points, not as evidence for the model. This is a genuine structural improvement.

F2. Dunning-Kruger Reframing Assessment#

Verdict: HELD (S1 residue).

The reframing from “we generalize D-K” to “D-K is one instance” is more honest and better supported. The four-level competence spectrum (Section 2.2) is well-argued and connects to the right literatures (D-K, deliberate practice, earned dogmatism, Tetlock).

S1 residue: The claim that OK-closure is “the common mechanism” is technically imprecise. The common structure is the outcome (OK stops correction). The underlying processes differ:

At low competence: metacognitive deficit (cannot see the gap).
At moderate competence: effort reduction (chooses not to close it).
At high competence: self-licensing (believes it has been earned).

These are psychologically distinct processes converging on the same formal outcome. The paper would be more precise to say “common formal structure” rather than “common mechanism.” This is minor (S1) and does not threaten the argument.

F3. Supervillain Theorem Assessment#

Verdict: HELD, with one new issue (N1, S2).

The conjunction condition (frozen scope + retained influence) and the risk-factor framing are genuine improvements. The disconfirming cases (Mandela, Eisenhower) and base-rate acknowledgment address the selection-bias concerns from the original review.

New Issue N1 (S2): The “systematic production” mechanism.

Section 3.5 introduces a claim not in the original paper or the original review: “Supervillains are not born and they are not rare bad luck. They are systematically produced by systems that suppress individual uniqueness.”

This claim has two components:

Suppression of uniqueness forces people into OK. This is plausible and partially supported by literature on conformity pressure (Asch, 1956), self-silencing (Jack & Dill, 1992), and identity foreclosure (Marcia, 1966). When people hide their actual selves, they are effectively declaring the mask OK.
This is a systematic production process. This is a strong sociological claim. The mechanism (suppression → conformity → frozen scope → mismatch → harm) is conceptually coherent but empirically speculative. No longitudinal data currently track this causal chain. The claim that suppression of uniqueness systematically produces supervillains goes beyond risk-factor language and returns to something closer to a causal law.

Recommendation: Either:

Frame the systematic-production mechanism as a hypothesis to be tested (consistent with the risk-factor framing), or
Provide specific empirical evidence for at least one link in the causal chain (e.g., identity foreclosure studies showing that foreclosed individuals are more likely to engage in rigid over-application of narrow expertise).

Currently the claim sits uncomfortably between the paper’s careful risk-factor language and a sweeping sociological assertion. The tension is noticeable.

F4. Five-Gate Model Assessment#

Verdict: HELD.

The operationalization table (Section 3.7) is well-constructed. Each gate is mapped to existing instruments, and the “What the Gate Model Adds” column specifies the novel contribution for each gate.

Gate 3 (Section 3.3) now has substantially more content than in MMv2. The clinical example (therapist defaulting to depression frameworks when the client presents grief) is concrete and illustrative. The structural prediction (specialists are more vulnerable to Gate 3 failure than generalists) is testable.

The sequential structure claim is now prominent and clearly articulated. The paper correctly identifies this as the five-gate model’s primary novel contribution.

Minor gap: The operationalization table lists existing instruments but does not propose a new integrated instrument that operationalizes the sequential structure itself. The paper says the sequential structure is the novel contribution, but the table only maps individual gates to individual instruments. How would you test the sequential claim? The Discussion (Section 5) mentions “Clinical outcome studies testing the five-gate sequential structure” but does not sketch a design. This is acceptable for a theoretical paper (the design belongs in a follow-up methods paper) but should be flagged as a gap.

F5. Parallels Section Calibration#

Verdict: HELD.

The calibration is well-judged:

Tuckman (Section 4.1): Correctly identified as strongest. The Storming = EQUAL mapping is specific, non-obvious, and testable. Well-argued.
Erikson (Section 4.2): Honestly calibrated. The stage-by-stage table with parallel-strength ratings is a model of scholarly honesty. The three-feature overlap is explicitly downgraded. The ordering reversal is acknowledged as fundamental. The summary is well-worded.
Maslow (Section 4.3): Appropriate caveats. The mapping-quality table identifies strong (Esteem, Self-transcendence) and weak (Safety) mappings. The “productive difference” framing (needs vs. construction stages) is insightful.
Kohlberg/Bloom (Section 4.4): Correctly labeled “suggestive analogies.” The paper explicitly acknowledges these are under-argued.

Is this too generous anywhere? No. If anything, the Erikson section is slightly too generous at Stages 3–6 (labeling them “Approximate” when some might be better called “Weak”), but the overall assessment paragraph at the end corrects for this.

Too dismissive anywhere? No. Kohlberg/Bloom deserve their “suggestive analogy” label.

F6. New Content Assessment#

F6.1 Cognitive Dissonance Prediction (Section 2.3)#

Verdict: HELD.

The prediction is clean: “individuals scoring high on tolerance of ambiguity and low on need for cognitive closure should display more NOT-OK-like behavior across domains.” This is testable with existing instruments (Need for Cognitive Closure scale, Tolerance of Ambiguity scale). The prediction follows logically from the OK/NOT-OK framework.

One precision note (S1): The prediction as stated is correlational, not causal. High tolerance of ambiguity correlating with NOT-OK-like behavior does not establish that tolerance of ambiguity causes NOT-OK behavior. The paper should acknowledge this: the prediction is about co-occurrence, not mechanism. A high-tolerance individual may display NOT-OK behavior for reasons unrelated to the OK-closure mechanism (e.g., personality traits like openness).

F6.2 Kohlberg Moral-Regression Prediction (Section 4.4)#

New Issue N2 (S2): The moral-regression prediction needs more specificity.

The prediction: “moral regression under stress occurs when stress triggers OK-closure: the agent reverts to simpler rule-following because general reasoning requires the sustained cognitive effort of NOT-OK self-assessment.”

This is interesting but under-specified:

Kohlberg’s moral regression under stress is already well-documented (Krebs & Denton, 2005; Carpendale, 2000). The phenomenon does not need the e7Day model to explain it — existing accounts invoke cognitive load and ego depletion.
The e7Day model adds an alternative mechanism (OK-closure rather than cognitive load). For this to be a contribution, the two mechanisms must make different predictions in at least one scenario. The paper does not identify such a differentiating scenario.
Testable differentiator: The e7Day model predicts that moral regression should be more persistent and harder to reverse in individuals with high OK-closure tendencies (high need for cognitive closure) than in individuals with low OK-closure tendencies. The cognitive-load account predicts that regression reverses immediately when load is removed, regardless of closure tendencies. This differentiating prediction is available but not stated.

Recommendation: Add the differentiating prediction. Without it, the Kohlberg connection is “our model can also explain this” (which is uninteresting) rather than “our model predicts something the existing account does not” (which is a contribution).

F6.3 Operationalization Table (Section 3.7)#

Verdict: HELD.

The table is well-constructed. Each gate maps to specific, real instruments. The “What the Gate Model Adds” column is clear and adds value beyond a simple instrument list.

Minor note: The table cites Davis (1983) for the IRI in the Gate 4 row. The IRI (Interpersonal Reactivity Index) is from Davis (1983), which is correct. However, the same citation appears in the abstract for the Dunning-Kruger effect, which is Kruger & Dunning (1999). It appears that :cite:`Davis1983` may be used for both Davis’s IRI and the Dunning-Kruger effect, which are different works by different authors. This should be checked — it may be a citation-key collision (S1).

F7. Axiom Cross-Check Against the Math Paper#

Verdict: HELD.

Cross-referencing the socpsy paper’s axiom references against the math paper (b12-math MMv3r1):

m6.ax4 (ZION requires NOT-OK self-assessment): Correctly cited in Section 2.1. The split into axiom + BABL definition is reflected.
m5.ax2 (UMP, noisy channel): Correctly cited in Section 3.4 (Gate 4). Consistent with the math paper’s statement.
m2.ax2 (Lossy Mapping): Not directly cited in the socpsy paper but correctly implied through the PERFECT/PERFIDE reference (Section 5).
m6.ax5 (Environmental Novelty): Correctly cited in Section 3.5 (Gate 5, supervillain theorem). Consistent with the math paper.
th7 (Compassion Capacity): Correctly cited in Section 3 as the formal theorem behind the five-gate model.
th3 (BABL Origin): Correctly referenced through the cost asymmetry discussion (Section 2.4).
mc.ax4 (Construction Cascade): Correctly referenced in the Maslow section (Section 4.3) regarding cascading dependency.

No axiom mismatches detected. The socpsy paper’s references to the formal system are consistent with the math paper.

Part IV: New Issues Summary#

Issue	Description	Severity	Verdict
N1	“Systematic production” mechanism (Section 3.5) sits between risk-factor language and causal-law language. Empirically speculative as stated. Either frame as hypothesis or provide evidence for at least one causal link.	S2	BREACH
N2	Kohlberg moral-regression prediction (Section 4.4) does not differentiate from existing accounts (cognitive load, ego depletion). Needs a scenario where the OK-closure mechanism predicts differently from the standard account.	S2	BREACH
R1	“Common mechanism” language for OK-closure (Section 2.2) is imprecise — the common structure is the outcome, not the underlying process. Minor terminological issue.	S1	HELD
R2	Cognitive dissonance prediction (Section 2.3) is correlational, not causal. Should acknowledge this.	S1	HELD
R3	Possible citation-key collision: Davis (1983) may be mapped to both the IRI and the Dunning-Kruger effect. Check citation keys.	S1	HELD

Part V: Overall Assessment and EDEN Classification#

Progress: The paper has moved from 4 S3 BREACHes + 5 S2 HELDs + 2 S1 HELDs (original review) to 0 S3 issues + 2 new S2 BREACHes + 3 S1 HELDs (this re-review). This is substantial progress.

Remaining work: The two S2 issues (N1: systematic production framing; N2: Kohlberg differentiating prediction) are addressable within the current structure. Neither requires major restructuring.

EDEN Classification:

I found the following in EDEN:

Green Meadow #1 (own contributions), count = 4. The paper’s four original contributions (OK-closure mechanism, supervillain theorem, five-gate sequential model, cognitive dissonance connection) are now well-argued and stand independently. This is unchanged from the original review and now better realized in the paper’s structure.
Knife Edge #1 (systematic production mechanism). The claim that supervillains are “systematically produced by systems that suppress individual uniqueness” is conceptually compelling but empirically ungrounded. The narrow path: frame it as a testable hypothesis with specific empirical predictions, not as a finding. Any stronger framing risks the same overclaiming pattern the revision successfully addressed elsewhere.
Green Meadow #2 (parallels calibration), count = 4. The parallels section now offers multiple defensible framings: Tuckman as specific parallel, Erikson Stages 7–8 as strong resonances, Maslow as partially supported with caveats, Kohlberg/Bloom as suggestive analogies. Each is honestly calibrated. The original Knife Edge on parallels has been resolved through honest assessment.
Knife Edge #2 (Kohlberg prediction). The moral-regression prediction (Section 4.4) teeters between genuine contribution (if a differentiating prediction is added) and empty re-description (if no differentiator is specified). The narrow path: identify the scenario where OK-closure and cognitive-load accounts diverge.

Overall EDEN classification: approaching Green Meadow. The paper has resolved its original Grey Meadow status. Two Knife Edges remain, both addressable. The paper’s own contributions are strong and now properly showcased.

Recommendations#

S2 (should address before advancing):

(N1) Frame the systematic-production mechanism as a hypothesis. Candidate wording: “We hypothesize that systems suppressing individual uniqueness systematically increase supervillain risk through the chain: suppression → forced conformity → OK-closure → frozen scope → harm proportional to influence. Testing this hypothesis requires longitudinal data on conformity pressure, OK-closure behavior, and subsequent domain-inappropriate intervention.”
(N2) Add a differentiating prediction for the Kohlberg connection. Candidate: “The OK-closure account predicts that moral regression should persist longer, and reverse more slowly, in individuals with high need for cognitive closure than in individuals with low need for closure — even after the stressor is removed. The cognitive-load account predicts recovery proportional to load reduction, independent of closure tendencies.”

S1 (minor polish):

(R1) Change “common mechanism” to “common formal structure” in Section 2.2’s final paragraph.
(R2) Note in Section 2.3 that the cognitive dissonance prediction is correlational as stated; a causal test would require experimental manipulation of closure/ambiguity tolerance.
(R3) Check citation keys: ensure Davis (1983) maps to the IRI (Mark H. Davis) and Dunning-Kruger maps to Kruger & Dunning (1999), not to the same key.