Skip to content

Latest commit

 

History

History
747 lines (577 loc) · 14.1 KB

File metadata and controls

747 lines (577 loc) · 14.1 KB

Part 4 Causal Structure with Evidence Distribution

This document shows the complete causal dependency structure linking all questions, with per-book evidence counts under each question.


LAYER 0: Framework Validity (Foundation)

Purpose: Tests whether the two-pattern dichotomy (maximum-leverage vs capability-enhancement) is real and dominant.

Dependencies: None (foundational layer)

Critical: If Layer 0 fails, the entire framework collapses.

P0a: Do the patterns exist as distinct optimization targets?

P0a_pure_maximum_leverage: Pure maximum-leverage cases (centralized coercion + information control)?

  • blum: 106
  • acemoglu: 48
  • lessig: 46
  • scott: 46
  • graeber: 43
  • dictators-handbook: 38
  • ferguson: 37
  • tufekci: 17
  • han-feizi: 13
  • weber: 13
  • piketty: 8
  • girard: 7
  • ostrom: 5
  • perez: 3
  • schmitt: 1

Total: 431 evidence items

P0a_pure_capability_enhancement: Pure capability-enhancement cases (distributed capability + transparency)?

  • lessig: 26
  • ostrom: 16
  • graeber: 14
  • ferguson: 8
  • scott: 8
  • tufekci: 6
  • acemoglu: 5
  • weber: 5
  • blum: 3
  • dictators-handbook: 2
  • han-feizi: 1
  • kahneman: 1
  • perez: 1
  • piketty: 1
  • yudkowsky: 1

Total: 98 evidence items

P0a_patterns_coherent: Are these coherent, stable categories or do definitions blur?

  • weber: 9
  • ferguson: 6
  • lessig: 5
  • scott: 5
  • dictators-handbook: 3
  • girard: 3
  • graeber: 3
  • acemoglu: 2
  • han-feizi: 2
  • kahneman: 2
  • piketty: 2
  • blum: 1
  • ostrom: 1
  • perez: 1
  • schmitt: 1

Total: 46 evidence items

P0b: Are they alternatives or do they combine?

P0b_adversarial_relationship: Adversarial relationship (when one increases, does other decrease)?

  • ferguson: 13
  • acemoglu: 9
  • dictators-handbook: 7
  • lessig: 7
  • han-feizi: 5
  • scott: 5
  • blum: 3
  • perez: 3
  • weber: 2
  • graeber: 1
  • ostrom: 1
  • tufekci: 1

Total: 57 evidence items

P0b_patterns_combine: Can patterns combine in hybrid systems?

  • ferguson: 38
  • blum: 35
  • lessig: 33
  • acemoglu: 32
  • graeber: 27
  • ostrom: 19
  • weber: 19
  • scott: 18
  • dictators-handbook: 15
  • piketty: 15
  • tufekci: 13
  • perez: 9
  • girard: 6
  • kahneman: 3
  • han-feizi: 2
  • schmitt: 1
  • yudkowsky: 1

Total: 286 evidence items

P0b_hybrid_outperforms: Do hybrid systems outperform pure implementations?

  • acemoglu: 9
  • ferguson: 6
  • ostrom: 4
  • graeber: 3
  • perez: 2
  • piketty: 2
  • blum: 1
  • scott: 1
  • weber: 1

Total: 29 evidence items

P0c: Are these THE dominant patterns?

P0c_explain_variance: Do these patterns explain most variance in outcomes?

  • dictators-handbook: 6
  • acemoglu: 4
  • ferguson: 2
  • schmitt: 1
  • scott: 1

Total: 14 evidence items

P0c_legitimacy_dominant: Is legitimacy/ideology more dominant than structural patterns?

  • blum: 47
  • ferguson: 13
  • weber: 11
  • piketty: 10
  • lessig: 9
  • graeber: 7
  • acemoglu: 5
  • girard: 5
  • scott: 4
  • dictators-handbook: 3
  • ostrom: 3
  • tufekci: 3
  • han-feizi: 2
  • perez: 2

Total: 124 evidence items

P0c_efficiency_dominant: Is efficiency/rationalization more dominant?

  • scott: 26
  • piketty: 25
  • blum: 21
  • weber: 21
  • lessig: 17
  • ferguson: 16
  • perez: 16
  • dictators-handbook: 14
  • acemoglu: 12
  • graeber: 7
  • han-feizi: 7
  • ostrom: 7
  • kahneman: 4
  • tufekci: 4
  • girard: 2
  • yudkowsky: 2

Total: 201 evidence items

P0c_other_patterns: Other optimization targets that matter more?

  • piketty: 39
  • graeber: 32
  • perez: 31
  • girard: 28
  • ferguson: 24
  • acemoglu: 23
  • scott: 22
  • lessig: 20
  • ostrom: 20
  • kahneman: 17
  • weber: 16
  • tufekci: 15
  • blum: 9
  • dictators-handbook: 8
  • yudkowsky: 4
  • schmitt: 3
  • han-feizi: 1

Total: 312 evidence items


LAYER 1: Historical Premises

Purpose: Tests specific historical claims about pattern dominance and constraints.

Dependencies: Requires Layer 0 (framework must be valid)

Critical: Establishes baseline of historical pattern dynamics that AI might change.

P1: Cognitive bottlenecks were limiting

P1_cognitive_bottleneck: Did coordination fail due to lack of cognitive capacity (not just political will)?

  • kahneman: 71
  • scott: 7
  • ferguson: 5
  • ostrom: 5
  • han-feizi: 3
  • perez: 3
  • weber: 3
  • acemoglu: 1
  • girard: 1
  • lessig: 1
  • piketty: 1
  • tufekci: 1
  • yudkowsky: 1

Total: 103 evidence items

P1_information_insufficient: Was information available but comprehension/synthesis lacking?

  • kahneman: 33
  • ostrom: 12
  • scott: 7
  • tufekci: 7
  • blum: 4
  • ferguson: 2
  • lessig: 2
  • perez: 2
  • piketty: 2
  • yudkowsky: 2
  • girard: 1
  • graeber: 1

Total: 75 evidence items

P1_political_will_dominant: Did political disagreements dominate even when cognitive capacity was adequate?

  • blum: 67
  • acemoglu: 32
  • piketty: 29
  • scott: 20
  • graeber: 18
  • lessig: 17
  • ferguson: 16
  • yudkowsky: 12
  • perez: 11
  • dictators-handbook: 10
  • kahneman: 10
  • ostrom: 8
  • han-feizi: 6
  • tufekci: 4
  • weber: 4
  • girard: 2
  • schmitt: 2

Total: 268 evidence items

P2: Maximum-leverage historically won

P2_max_leverage_wins: Cases where centralized coercion beat distributed capability?

  • blum: 57
  • acemoglu: 36
  • graeber: 35
  • ferguson: 26
  • lessig: 25
  • dictators-handbook: 21
  • scott: 20
  • han-feizi: 12
  • tufekci: 10
  • piketty: 8
  • perez: 7
  • weber: 7
  • ostrom: 5
  • girard: 1
  • yudkowsky: 1

Total: 271 evidence items

P2_capability_wins: Cases where distributed capability beat centralized coercion?

  • acemoglu: 18
  • ostrom: 17
  • ferguson: 13
  • scott: 12
  • tufekci: 11
  • dictators-handbook: 10
  • lessig: 10
  • blum: 5
  • graeber: 5
  • weber: 2
  • han-feizi: 1
  • perez: 1
  • piketty: 1
  • yudkowsky: 1

Total: 107 evidence items

P2_conditional_factors: What conditions determine which pattern wins?

  • blum: 56
  • acemoglu: 42
  • ferguson: 29
  • graeber: 29
  • dictators-handbook: 27
  • scott: 26
  • perez: 21
  • ostrom: 20
  • lessig: 17
  • piketty: 15
  • tufekci: 15
  • weber: 15
  • girard: 7
  • han-feizi: 7
  • yudkowsky: 2
  • kahneman: 1
  • schmitt: 1

Total: 330 evidence items

P3: Existential risks are coordination-dependent

P3_technical_solutions_exist: Are threats technically solvable with existing knowledge?

  • acemoglu: 3
  • ostrom: 3
  • piketty: 3
  • dictators-handbook: 2
  • graeber: 2
  • blum: 1
  • yudkowsky: 1

Total: 15 evidence items

P3_coordination_prevents: Does coordination failure (not technical limits) prevent deploying solutions?

  • blum: 12
  • ostrom: 12
  • piketty: 9
  • girard: 8
  • graeber: 7
  • perez: 7
  • acemoglu: 6
  • dictators-handbook: 4
  • yudkowsky: 4
  • kahneman: 2
  • ferguson: 1
  • lessig: 1
  • schmitt: 1
  • tufekci: 1

Total: 75 evidence items

P3_scarcity_vs_coordination: Is absolute scarcity the problem, or coordination failure?

  • piketty: 6
  • graeber: 3
  • acemoglu: 2
  • blum: 2
  • girard: 1
  • ostrom: 1
  • perez: 1
  • scott: 1

Total: 17 evidence items

P4: Maximum-leverage is operationally corrosive

P4_operators_experience_cost: Do operators experience exhaustion, inability to trust, emotional toll?

  • blum: 30
  • graeber: 12
  • acemoglu: 5
  • ferguson: 5
  • lessig: 4
  • scott: 4
  • tufekci: 3
  • dictators-handbook: 2
  • han-feizi: 2
  • girard: 1
  • ostrom: 1

Total: 69 evidence items

P4_corrosion_when_winning: Is corrosion present even when system is winning competitively?

  • graeber: 4
  • acemoglu: 3
  • scott: 1
  • weber: 1

Total: 9 evidence items

P4_rationalize_successfully: Do operators rationalize/adapt successfully without corrosion?

  • blum: 9
  • dictators-handbook: 3
  • graeber: 3
  • acemoglu: 2
  • han-feizi: 2
  • ferguson: 1
  • kahneman: 1
  • perez: 1
  • piketty: 1
  • tufekci: 1
  • yudkowsky: 1

Total: 25 evidence items

P5: 'This time is different' usually fails

P5_previous_claims: Were similar transformative claims made for previous technologies?

  • perez: 18
  • lessig: 10
  • tufekci: 9
  • ferguson: 8
  • piketty: 6
  • acemoglu: 3
  • blum: 3
  • graeber: 3
  • schmitt: 3
  • yudkowsky: 2
  • han-feizi: 1
  • scott: 1

Total: 67 evidence items

P5_claims_failed: Did those claims fail to change fundamental coordination dynamics?

  • piketty: 11
  • perez: 10
  • lessig: 6
  • ferguson: 3
  • tufekci: 3
  • blum: 2
  • dictators-handbook: 2
  • graeber: 2
  • yudkowsky: 2
  • acemoglu: 1
  • schmitt: 1

Total: 43 evidence items

P5_claims_succeeded: Were there cases where "this time" actually was different?

  • graeber: 2
  • perez: 2
  • piketty: 2
  • yudkowsky: 2
  • acemoglu: 1
  • girard: 1
  • kahneman: 1

Total: 11 evidence items


LAYER 2: AI Mechanism

Purpose: Tests what AI actually does differently from previous technologies.

Dependencies: Requires Layers 0, 1 (framework valid + historical patterns established)

Critical: If M1 fails (AI is just more information), the entire 'this time is different' claim fails.

M1: AI amplifies comprehension (not just information)

M1_previous_info_only: Did previous technologies only transmit information faster/wider?

  • ferguson: 10
  • perez: 7
  • tufekci: 6
  • lessig: 5
  • acemoglu: 2
  • blum: 1
  • kahneman: 1
  • piketty: 1
  • schmitt: 1

Total: 34 evidence items

M1_previous_comprehension: Did previous technologies also amplify comprehension capabilities?

  • tufekci: 3
  • ferguson: 2
  • perez: 2
  • graeber: 1
  • kahneman: 1
  • lessig: 1

Total: 10 evidence items

M1_qualitative_difference: Qualitative difference between information transmission and comprehension amplification?

  • kahneman: 8
  • yudkowsky: 7
  • tufekci: 2
  • acemoglu: 1
  • girard: 1
  • weber: 1

Total: 20 evidence items

M2: Sweet spot exists (amplifies without replacing)

M2_amplification_window: Historical examples of tools amplifying without replacing human agency?

  • kahneman: 3
  • ferguson: 2
  • perez: 2
  • tufekci: 2
  • yudkowsky: 2

Total: 11 evidence items

M2_replacement_threshold: What causes a tool to replace rather than amplify?

  • yudkowsky: 7
  • perez: 2
  • ferguson: 1
  • kahneman: 1
  • scott: 1
  • tufekci: 1

Total: 13 evidence items

M2_window_closes: Evidence of amplification windows closing (tool begins replacing)?

  • yudkowsky: 4
  • tufekci: 1

Total: 5 evidence items


LAYER 3: Competitive Shift

Purpose: Tests whether AI changes competitive balance between the patterns.

Dependencies: Requires Layers 0, 1, 2 (framework valid + historical baseline + AI mechanism understood)

Critical: If C1 fails, capability-enhancement remains competitively inferior even with AI.

C1: Capability-enhancement becomes competitive

C1_distributed_scales: Can distributed coordination scale effectively, or does it fragment?

  • ferguson: 28
  • ostrom: 20
  • tufekci: 18
  • lessig: 13
  • graeber: 11
  • acemoglu: 9
  • piketty: 7
  • perez: 5
  • dictators-handbook: 4
  • girard: 3
  • kahneman: 3
  • scott: 3
  • weber: 2
  • han-feizi: 1

Total: 127 evidence items

C1_centralization_fragility: Does centralized control become more fragile as it scales?

  • acemoglu: 20
  • ferguson: 18
  • blum: 13
  • scott: 12
  • graeber: 10
  • perez: 9
  • dictators-handbook: 8
  • weber: 8
  • tufekci: 7
  • ostrom: 6
  • piketty: 6
  • lessig: 5
  • yudkowsky: 2
  • girard: 1
  • han-feizi: 1
  • kahneman: 1
  • schmitt: 1

Total: 128 evidence items

C1_operational_costs: Do operational costs favor one pattern?

  • blum: 42
  • acemoglu: 18
  • graeber: 17
  • ferguson: 16
  • lessig: 16
  • ostrom: 15
  • dictators-handbook: 13
  • tufekci: 13
  • scott: 12
  • weber: 11
  • piketty: 8
  • perez: 3
  • girard: 2
  • han-feizi: 2
  • yudkowsky: 1

Total: 189 evidence items


LAYER 4: Risk Reduction

Purpose: Tests whether distributed AI actually reduces existential risk.

Dependencies: Requires Layers 0, 3 (framework valid + competitive shift established)

Critical: If C2 fails, distributed AI might be competitive but still dangerous.

C2: Distributed capability reduces existential risk

C2_distribution_enables: Does distribution enable coordination on existential risks?

  • ostrom: 1
  • piketty: 1
  • tufekci: 1
  • yudkowsky: 1

Total: 4 evidence items

C2_distribution_arms_races: Does distribution create dangerous arms races?

  • ferguson: 4
  • tufekci: 3
  • yudkowsky: 2
  • blum: 1
  • girard: 1
  • lessig: 1

Total: 12 evidence items

C2_centralization_needed: Is centralized control needed to manage certain threats?

  • ferguson: 2
  • acemoglu: 1
  • girard: 1
  • ostrom: 1
  • yudkowsky: 1

Total: 6 evidence items


LAYER 5: Motivation

Purpose: Tests whether current events provide unprecedented motivation for change.

Dependencies: Requires Layers 1, 3 (historical patterns + competitive shift)

Critical: If L5 fails, even if shift is possible, motivation may be insufficient.

L5: Current events significance

L5_historical_precedent: Historical precedents for current scale of maximum-leverage demonstrations?

  • perez: 8
  • blum: 7
  • graeber: 2
  • lessig: 2
  • weber: 2
  • dictators-handbook: 1
  • ferguson: 1
  • piketty: 1
  • tufekci: 1

Total: 25 evidence items

L5_unprecedented_scale: Is current scale/visibility of demonstrations unprecedented?

  • tufekci: 4
  • yudkowsky: 3
  • blum: 1
  • girard: 1
  • lessig: 1
  • perez: 1

Total: 11 evidence items

L5_motivation_creates_change: Did similar demonstrations create switching incentives historically?

  • blum: 3
  • perez: 3
  • yudkowsky: 3
  • graeber: 2
  • tufekci: 2
  • acemoglu: 1
  • lessig: 1

Total: 15 evidence items


Causal Chain Summary

Layer 0 (Framework Valid?)
    ↓
Layer 1 (Historical patterns established?)
    ↓ ↘
Layer 2 (AI different?)   →   Layer 5 (Unprecedented motivation?)
    ↓                              ↑
Layer 3 (Competitive shift?)  ----↗
    ↓
Layer 4 (Risk reduction?)

Full argument requires ALL layers to succeed:

  1. Layer 0: The dichotomy is real and empirically identifiable
  2. Layer 1: Maximum-leverage historically won, cognitive limits mattered
  3. Layer 2: AI amplifies comprehension (not just information)
  4. Layer 3: This makes capability-enhancement competitive
  5. Layer 4: And this reduces existential risk
  6. Layer 5: Current events provide motivation to switch

Evidence Summary:

  • Layer 0 (Framework): 1,598 evidence items - 90% SUPPORTS
  • Layer 1 (Historical): 1,485 evidence items - 93% SUPPORTS
  • Layer 2 (AI Mechanism): 93 evidence items - 83% SUPPORTS
  • Layer 3 (Competitive): 444 evidence items - 67% SUPPORTS, 33% CONDITIONAL
  • Layer 4 (Risk): 22 evidence items - 100% SUPPORTS
  • Layer 5 (Motivation): 52 evidence items - 100% SUPPORTS