Home / Research / Zotero 5C Graph V5
Chapters (node color)
Ch 1 — Institutional Trust Formation
Ch 2 — Credibility Veneer Risk
Ch 3 — HITL Proxy Auditor · CSO / Capture–Replay
Deployment Scope Boundaries (embedded)
Cross-Chapter
Framework Node
Archived/Supporting-only
Tier Encoding
Tier 1 Highest (core ring)
Tier 1
Tier 2
Tier 3 / Supporting
Relations (edge color)
Conceptual basis
Empirical evidence
Design implication
Governance implication
Parallel case
Framework integration
Methodological inspiration
Counter-evidence
Active view is strict V5. Use Show archived/supporting to reveal nodes moved out from the former standalone Ch4 chapter (kept for traceability only).
Validation Chain (Ch1→Ch2→Ch3)
  1. Ellis → outcome layer (halo effect): Ch1 establishes that institutional framing lowers privacy concern — the halo effect makes users suspend critical evaluation when a health AI looks official.
  2. Wong → mechanism layer (endorsement cues): Wong is not a new outcome study; it unpacks how the Ellis halo is triggered. Qualitative data identifies specific cues — physician recommendation, institutional branding, clinical tone — that activate an institutional trust shortcut, temporarily bypassing critical thinking. Wong sits between Yu & Chen (quantitative social-influence confirmation) and Yeung (hypernudge theory), serving as the bridge from social influence to concrete interface signals.
  3. Yeung / Faraoni / Duane → exploitation layer (hypernudge): The endorsement cues Wong identifies are precisely the nudge handles that Yeung's hypernudge framework, Faraoni's cognitive-liberty analysis, and Duane's digital-nudge taxonomy describe as systematically exploitable. Ch2 shows how these cues are weaponized in dynamic, personalized interaction loops.

Ch2 conclusion → Ch3 start: veneer-driven trust can rise while real risk-check evidence remains absent. Therefore Ch3 introduces the HITL proxy-auditor requirement to force explicit, inspectable risk-check evidence.

R8 method chain (graph): Tanprasert (HelpCall) remains the Capture–Replay implementation anchor; Mackay & McGrenere (2025, CSO) supplies the structured qualitative comparison framework (paired with Lee & See as organizing lens in the Week12–13 proposal).

Ch2 Explanatory Frame (2×2)
  • Bottom-left (Legacy dark patterns): low hidden manipulation + weak human-centered governance. Baseline coercive UX patterns (Gray 2018; Brignull 2024 as contextual taxonomy).
  • Top-left (Agentic sludge / bypass): high hidden manipulation + weak governance. Ch2 risk zone where personalization and optimization outpace user verification (Yeung 2017; Duane 2025).
  • Top-right (Constrained hypernudge): high personalization under stronger governance constraints. Influence exists but must be transparent, reviewable, and opt-out capable (Faraoni 2023 as a core philosophical basis for UI-audit insufficiency under dynamic hypernudge).
  • Bottom-right (HCAI ideal): low hidden manipulation + strong human-centered governance. Target state where influence is visible and auditable — named explicitly as Shneiderman (2020) Reliable/Safe/Trustworthy and carried into Ch3 forcing functions + replay traces (R8).

Interpretation rule: Ch2 is not merely a dark-pattern inventory; it explains the transition from legacy static manipulation to dynamic hypernudge risk, and why this transition forces movement toward Ch3 control-and-audit design.