Skip to content

R9990/2026-03-20

Research R9990 — STAR interview neurodivergent impact
Mode Claim
Run date 2026-03-20
Claims 1
Prompt Unified Research Standard v1.0-draft
Model Claude Opus 4.6

Single-claim verification run examining whether the STAR behavioral interview format disadvantages neurodivergent individuals, with focus on ADHD and dyslexia.

Claims

C001 — STAR interview neurodivergent impact — Likely (55-80%)

Claim: The STAR interview format commonly used in the interview and hiring process is problematic for neurodivergent individuals such as dyslexics and people with ADHD.

Verdict: Partially correct with important nuance. STAR's cognitive demands (episodic recall, narrative structuring, executive working memory) directly conflict with documented deficits in ADHD, autism, and dyslexia. However, STAR's predictable structure also provides scaffolding benefits compared to unstructured interviews. No peer-reviewed study directly measures STAR-specific performance for neurodivergent candidates.

Hypothesis Status Probability
H1: STAR is substantially problematic Inconclusive
H2: STAR is partially problematic with nuance Supported Likely (55-80%)
H3: STAR is not problematic Eliminated

Confidence: Medium · Sources: 7 · Searches: 6

Full analysis


Collection Analysis

Cross-Cutting Patterns

Pattern Claims Affected Significance
Inference gap: no STAR-specific research exists C001 The claim requires inference from cognitive research + interview experience studies; no study directly tests STAR format against neurodivergent performance
Dual nature of structure C001 The same structured format can be both barrier (in real-time execution) and scaffold (in preparation), which makes simple "problematic"/"not problematic" framing inadequate
Condition-dependent effects C001 Impact varies across neurodivergent conditions (autism > non-autistic ND > neurotypical gradient in interview experiences)

Collection Statistics

Metric Value
Claims investigated 1
Fully confirmed (Almost certain) 0
Confirmed with nuance (Very likely) 0
Confirmed with caveats (Likely) 1 (C001)
Roughly even 0
Unlikely or below 0

Source Independence Assessment

The evidence base draws from three genuinely independent streams: (1) cognitive neuroscience research on working memory deficits in ADHD (academic laboratory studies), (2) social science survey research comparing interview experiences across neurotypes (UK-based mixed-methods study), and (3) practitioner and advocacy observations about interview challenges (career coaching, employment advocacy organizations). These streams were produced by different researchers, in different fields, using different methodologies, and reach convergent conclusions about the underlying cognitive challenges. One source (SRC01) provides an independent counterpoint from the neurodiversity employment services sector. The Zurich Insurance survey is cited by multiple advocacy sources, creating some derived agreement, but the survey itself is independent of the academic research.

Collection Gaps

Gap Impact Mitigation
No STAR-specific performance study Cannot isolate STAR format effect from general interview effects Inferenced from cognitive research + interview experience; assigned Medium confidence
Dyslexia evidence is thin Claim names dyslexia but direct dyslexia + interview research is sparse Used cognitive mechanism inference (verbal fluency, working memory in dyslexia)
Adult ADHD working memory Strongest cognitive study used children 8-13 Noted as limitation; adult ADHD WM deficits supported by broader literature
Paywall barriers Multiple peer-reviewed sources inaccessible Relied on open-access sources; may have missed contradictory evidence

Collection Self-Audit

Domain Rating Notes
Eligibility criteria Low risk Defined before searching; applied consistently
Search comprehensiveness Some concerns 6 searches, 70 results, but paywalls blocked several peer-reviewed sources; dyslexia research sparse
Evaluation consistency Low risk Same scorecard framework applied to all 7 sources
Synthesis fairness Low risk Actively searched for contradicting evidence (S06); included counterpoint (SRC01) prominently; chose nuanced H2 over simpler H1

Resources

Summary

Metric Value
Claims investigated 1
Files produced 102
Sources scored 7
Evidence extracts 8
Results dispositioned 7 selected + 63 rejected = 70 total
Duration (wall clock) 16m 28s
Tool uses (total) 99

Tool Breakdown

Tool Uses Purpose
WebSearch 6 Search queries across 6 angles
WebFetch 10 Page content retrieval (5 successful, 2 failed/403, 2 JS-only, 1 redirect)
Write 26 File creation
Read 12 File reading (methodology, output spec, placeholders)
Edit 0 File modification
Bash 7 Directory creation, batch file generation, file counting

Token Distribution

Category Tokens
Input (context) ~250,000
Output (generation) ~45,000
Total ~295,000