The Inquiry
The Inquiry: Does the scientific literature on decision cognition, accountability, psychological safety, and organizational voice converge on design requirements for governance infrastructure that improves rather than degrades decision quality? If so, what are those requirements, and how do they differ from conventional governance system design?
Sub-questions: - RQ-1: Does cognitive load theory (Sweller 1988) formally constrain how governance infrastructure should present information to decision-makers? - RQ-2: Does the process vs. outcome accountability distinction (Lerner & Tetlock 1999) prescribe a specific accountability architecture for governance systems? - RQ-3: Does the psychological safety literature (Edmondson 1999) require structural voice mechanisms beyond leadership behavior? - RQ-4: Has anyone applied these decision science findings to AI governance system design? Has anyone applied Lerner & Tetlock to algorithmic accountability? - RQ-5: Can governance infrastructure be designed to make truth-indifference (organizational bullshit) structurally costly?
Falsifiable formulation: Governance infrastructure implementing process accountability, cognitive load reduction, and structural voice mechanisms will achieve higher adoption and better decision quality than equivalent infrastructure implementing outcome accountability and compliance monitoring. If outcome-accountable governance systems demonstrably outperform process-accountable ones on decision quality metrics, this thesis is falsified.
Executive Summary
The cognitive burden argument — governance that adds load produces worse decisions. Simon establishes bounded rationality as baseline. Sweller quantifies how extraneous load degrades performance. Kahneman shows System 2 disengages under load. Iyengar shows choice overload paralyzes. The synthesis: conventional governance systems (compliance documentation, exhaustive review requirements, unstructured information presentation) add extraneous cognitive load, push decision-makers from System 2 to System 1, and produce the "click and hope" behavior they were designed to prevent. The infrastructure failure is the governance system itself. CLT has not been applied to governance — this is the theoretical bridge.
The accountability reframe — process over outcome. Lerner & Tetlock's distinction is the design principle: capture HOW decisions were made, not retrospectively judge WHAT happened. Bernstein's transparency paradox shows surveillance degrades performance through impression management. Colquitt's procedural justice predicts compliance. The synthesis: governance infrastructure should prove process, not judge outcomes. This is not a framing choice — it is an empirically grounded design decision. This framework has not been applied to AI governance — another bridge.
Structural voice — from psychological safety to governed channels. Edmondson establishes that psychological safety enables learning. Morrison & Milliken show silence is structural. Detert & Edmondson show implicit voice theories suppress voice even in safe environments. Hirschman shows Voice is an organizational resource. The synthesis: psychological safety is necessary but fragile (leader-dependent, culture-dependent). Structural voice mechanisms — governed channels that are system-protected, not leader-dependent — provide the infrastructure layer that makes voice a first-class organizational operation.
The truth-indifference chain — from Frankfurt to botshit. Frankfurt defines bullshit as truth-indifference. Spicer shows organizations produce it systematically. Ferreira measures it on three dimensions. Hannigan et al. (2024) add "botshit" — AI-generated truth-indifferent content. The chain: organizations produce truth-indifferent communication (OBPS) → AI amplifies it (botshit) → humans cannot filter it (cognitive limits) → training breaks the invariances needed to detect it (S11/Chlon). Governance infrastructure designed to require engagement with evidence, authority traceability, and constraint acknowledgment makes truth-indifference structurally costly. This is the first formalization of anti-bullshit architecture — connecting Frankfurt's philosophy to Ferreira's measurement to governance design.
The triple System 1/2 convergence. Kahneman (cognitive science), LeCun (AI, via S8), and Beer (cybernetics, via S9) independently discover the same dual-process architecture. This convergence validates governance infrastructure designed with a principled mechanism for transitioning between reactive (routine) and deliberative (complex) modes. The governance parallel: routine decisions handled with minimal deliberation (system assists); complex decisions surfacing full context for deliberative reasoning (system structures). The transition is evidence-based, not arbitrary.
Findings18
F-RA-010-01 · theoretical-grounding · established
Bounded rationality — satisficing under cognitive constraints — is the scientifically established baseline for organizational decision-making, not an aberration to be corrected. Decision-makers use aspiration levels rather than global optimization; when an alternative meets the aspiration level, search stops. "Click and hope" is satisficing under extreme cognitive load.
F-RA-010-02 · gap-identification · lab-originated
Cognitive load theory formally constrains how information should be presented to decision-makers, but has NOT been applied to governance or compliance system design — a confirmed gap. Three load types: intrinsic (task complexity), extraneous (presentation burden — should be eliminated), germane (productive schema construction — should be maximized). When intrinsic + extraneous exceeds working memory capacity, germane processing collapses.
F-RA-010-03 · convergent-validation · lab-originated
The dual-process architecture (System 1 / System 2) is independently discovered by cognitive science (Kahneman), AI research (LeCun, via S8), and cybernetics (Beer, via S9) — a triple convergence on the same reactive/deliberative split with a load-driven transition.
F-RA-010-04 · root-cause-diagnosis · lab-originated
The "decision fatigue" concept is empirically grounded (the Israeli parole-board pattern: 65% favorable at session start → near 0% by session end → reset after break, is real) but its mechanistic explanation (ego depletion) is substantially contested. The general claim that decision quality may degrade with sequential decisions under load remains plausible but the specific mechanism is unresolved.
F-RA-010-05 · theoretical-grounding · established
Choice overload degrades decision quality — more options without structure produces worse outcomes (jam study: 24 options → 3% purchased; 6 options → 30% purchased). As choice sets grow, both satisfaction and confidence decline.
F-RA-010-06 · gap-identification · lab-originated
Process accountability improves decision quality; outcome accountability degrades it. This is empirical, replicated, and has NOT been applied to AI governance — a confirmed gap. Process accountability (accountable for HOW you decided) produces systematic thinking, reduced bias, consideration of alternatives, openness to disconfirming evidence; outcome accountability (accountable for WHAT happened) produces defensive reasoning, premature closure, confirmation bias, impression management.
F-RA-010-07 · gap-identification · lab-originated
The transparency paradox — complete observability can reduce performance by triggering impression management rather than genuine work. In a Chinese mobile-phone factory, complete observability REDUCED performance; workers behind curtains performed better because privacy enabled experimentation and productive deviance.
F-RA-010-08 · theoretical-grounding · established
Procedural justice predicts rule compliance; informational justice predicts engagement. Four independent justice dimensions: distributive (outcome fairness → satisfaction), procedural (process fairness → rule compliance and institutional trust), interpersonal (respect/dignity → supervisor attitudes), informational (transparency/explanation → voluntary helping behavior). People comply with rules they perceive as procedurally fair.
F-RA-010-09 · theoretical-grounding · established
Psychological safety is the primary predictor of team learning behavior — one of the most established findings in organizational behavior (~10,000+ citations). PS — a shared belief that the team is safe for interpersonal risk-taking — predicts asking questions, admitting mistakes, reporting errors, seeking feedback, and is distinct from trust and cohesiveness.
F-RA-010-10 · gap-identification · lab-originated
Organizational silence is a structural phenomenon — collective, driven by organizational structures and norms rather than individual shyness — and implicit voice theories ("managers don't want bad news," "raising concerns signals disloyalty") suppress voice EVEN in psychologically safe environments with pro-voice leadership.
F-RA-010-11 · design-requirement-derivation · lab-originated
The absence of internal voice channels predicts external whistleblowing — governed internal voice is both a governance benefit and a risk-mitigation mechanism. Whistleblowing occurs when internal voice channels are absent or ineffective.
F-RA-010-12 · design-requirement-derivation · lab-originated
Expert decision-makers use pattern recognition and mental simulation, not exhaustive comparison — decision support should enhance this, not replace it. The Recognition-Primed Decision (RPD) model: experts recognize patterns from experience and mentally simulate the first viable option, adapting if simulation reveals problems.
F-RA-010-13 · architectural-framing · lab-originated
Cognition is distributed across people, artifacts, and systems — governance infrastructure is a distributed cognitive artifact. Intelligence emerges from coordination across individuals, technologies, and procedures; the cognitive unit is the system (people + tools + procedures), not any individual.
F-RA-010-14 · gap-identification · lab-originated
Choice architecture affects decisions without restricting options — governance infrastructure can be designed as organizational choice architecture. Defaults matter enormously (organ donation opt-in vs. opt-out); information structuring and feedback timing matter. No published work applies choice architecture specifically to internal organizational governance design — confirmed gap.
F-RA-010-15 · design-requirement-derivation · lab-originated
Organizational bullshit is measurable on three dimensions (Regard for Truth, The Boss, Bullshit Language), and AI-generated "botshit" amplifies it — creating a truth-indifference chain. Frankfurt: bullshit is truth-indifference (worse than lying); Spicer: organizations are truth-indifference machines; Hannigan et al. (2024): AI cannot care whether its output is true, so uncritical human use makes humans bullshit amplifiers.
F-RA-010-16 · root-cause-diagnosis · lab-originated
The box-ticker risk — process accountability can become its own accountability theater if documentation is filed without use. Graeber's "box-tickers" create the appearance of governance without substance.
F-RA-010-20 · convergent-validation · lab-originated
Synthesis-level finding (continues the integer sequence past the sprint's highest F-number, F18): The five decision-science design requirements for governance infrastructure converge from independent literatures — (1) cognitive load management, (2) process accountability, (3) structural voice, (4) distributed cognition / decision support, (5) anti-bullshit architecture — answering the primary RQ in the affirmative.
F-RA-010-21 · contribution-synthesis · lab-originated
Synthesis-level finding: Four confirmed literature gaps mark original-contribution space — (1) CLT not applied to governance/compliance; (2) Lerner & Tetlock process/outcome not applied to AI governance; (3) choice architecture not applied to organizational governance; (4) no governance architecture designed against truth-indifference. (L0 statement of the gap set, drawn from the L0 surfaces.)
Open Questions5
OQ-035Can CLT be formally applied to governance system design?
OQ-036Can Lerner & Tetlock process/outcome framework be applied to AI governance?
OQ-037What mechanism explains decision quality degradation under sequential decisions?
OQ-038Can choice architecture be formalized for internal governance design?
OQ-039Does the truth-indifference chain constitute a formal argument for governance infrastructure as AI precondition?
Bibliography43
Simon, Herbert A. (1947) · Administrative Behavior: A Study of Decision-Making Processes in Administrative Organization
Simon, Herbert A. (1955) · A Behavioral Model of Rational Choice
Kahneman, Daniel (2011) · Thinking, Fast and Slow
Kahneman, Daniel and Tversky, Amos (1979) · Prospect Theory: An Analysis of Decision under Risk
Sweller, John (1988) · Cognitive Load During Problem Solving: Effects on Learning
Sweller, John and van Merri{\"e}nboer, Jeroen J. G. and Paas, Fred (2019) · Cognitive Architecture and Instructional Design: 20 Years Later
Danziger, Shai and Levav, Jonathan and Avnaim-Pesso, Liora (2011) · Extraneous Factors in Judicial Decisions
Gl{\"o}ckner, Andreas (2016) · The irrational hungry judge effect revisited: Simulations reveal that fast, frugal, and fair strategies explain the data
Hagger, Martin S. and Chatzisarantis, Nikos L. D. and Alberts, Hugo and Anggono, Calvin Octavianus and Batailler, C{\'e}dric and Birt, Angela R. and Brand, Ralf and Brandt, Mark J. and Brewer, Gene and others (2016) · A Multilab Preregistered Replication of the Ego-Depletion Effect
Carter, Evan C. and McCullough, Michael E. (2014) · Publication Bias and the Limited Strength Model of Self-Control: Has the Evidence for Ego Depletion Been Overestimated?
Iyengar, Sheena S. and Lepper, Mark R. (2000) · When Choice is Demotivating: Can One Desire Too Much of a Good Thing?
Schwartz, Barry (2004) · The Paradox of Choice: Why More Is Less
+31 more citations