9d390da76b
Update UI across all pages with: - New favicon and brand icons (favicon-new.svg, tractatus-icon-new.svg) - Theme CSS integration (tractatus-theme.min.css) - Correct GitHub repository links (AgenticGovernance/tractatus) - PWA manifest updates - Consistent branding colors and gradients 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
524 lines
37 KiB
HTML
524 lines
37 KiB
HTML
<!DOCTYPE html>
|
||
<html lang="en">
|
||
<head>
|
||
<meta charset="UTF-8">
|
||
<meta name="viewport" content="width=device-width, initial-scale=1.0">
|
||
<title>For Researchers | Tractatus AI Safety Framework</title>
|
||
<meta name="description" content="Research foundations, empirical observations, and theoretical basis for architectural approaches to AI governance. Early-stage framework exploring structural constraints on LLM systems.">
|
||
|
||
<!-- PWA Manifest -->
|
||
<link rel="manifest" href="/manifest.json">
|
||
|
||
<!-- PWA Meta Tags -->
|
||
<meta name="theme-color" content="#3b82f6">
|
||
<meta name="apple-mobile-web-app-capable" content="yes">
|
||
<meta name="apple-mobile-web-app-status-bar-style" content="default">
|
||
<meta name="apple-mobile-web-app-title" content="Tractatus">
|
||
<link rel="apple-touch-icon" href="/images/tractatus-icon-new.svg">
|
||
|
||
<link rel="icon" type="image/svg+xml" href="/favicon-new.svg">
|
||
<link rel="stylesheet" href="/css/fonts.css">
|
||
<link rel="stylesheet" href="/css/tailwind.css?v=0.1.0.1760254958072">
|
||
<link rel="stylesheet" href="/css/tractatus-theme.min.css?v=1760816800">
|
||
<style>
|
||
.skip-link { position: absolute; left: -9999px; }
|
||
.skip-link:focus { left: 0; z-index: 100; background: white; padding: 1rem; }
|
||
a:focus, button:focus { outline: 2px solid #d97706; outline-offset: 2px; }
|
||
|
||
/* Accordion styles */
|
||
.accordion-button { cursor: pointer; user-select: none; }
|
||
.accordion-content {
|
||
max-height: 0;
|
||
overflow: hidden;
|
||
transition: max-height 0.3s ease-out;
|
||
}
|
||
.accordion-content.active {
|
||
max-height: 3000px;
|
||
transition: max-height 0.5s ease-in;
|
||
}
|
||
.accordion-icon { transition: transform 0.3s ease; }
|
||
.accordion-icon.active { transform: rotate(180deg); }
|
||
</style>
|
||
</head>
|
||
<body class="bg-white">
|
||
|
||
<a href="#main-content" class="skip-link">Skip to main content</a>
|
||
|
||
<script src="/js/components/navbar.js?v=0.1.0.1760254958072"></script>
|
||
|
||
<!-- Breadcrumb Navigation -->
|
||
<nav class="bg-gray-50 border-b border-gray-200 py-3" aria-label="Breadcrumb">
|
||
<div class="max-w-7xl mx-auto px-4 sm:px-6 lg:px-8">
|
||
<ol class="flex items-center space-x-2 text-sm">
|
||
<li><a href="/" class="hover:underline transition-colors" style="color: var(--tractatus-core-end);">Home</a></li>
|
||
<li class="text-gray-400">/</li>
|
||
<li class="text-gray-900 font-medium" aria-current="page">Researcher</li>
|
||
</ol>
|
||
</div>
|
||
</nav>
|
||
|
||
<!-- Header -->
|
||
<div class="bg-gray-50 border-b border-gray-200 py-16">
|
||
<div class="max-w-4xl mx-auto px-4 sm:px-6 lg:px-8">
|
||
<div class="inline-flex items-center bg-gray-200 text-gray-700 px-3 py-1 rounded text-sm font-medium mb-6" data-i18n="header.badge">
|
||
Research Framework • Empirical Observations
|
||
</div>
|
||
<h1 class="text-4xl font-bold text-gray-900 mb-4" data-i18n="header.title">
|
||
Research Foundations & Empirical Observations
|
||
</h1>
|
||
<p class="text-lg text-gray-600 leading-relaxed" data-i18n="header.subtitle">
|
||
Tractatus explores architectural approaches to AI governance through empirical observation of failure modes and application of organisational theory. This page documents research foundations, observed patterns, and theoretical basis for the framework.
|
||
</p>
|
||
</div>
|
||
</div>
|
||
|
||
<!-- Main Content -->
|
||
<div id="main-content" class="max-w-4xl mx-auto px-4 sm:px-6 lg:px-8 py-16">
|
||
|
||
<!-- Research Context -->
|
||
<section class="mb-16">
|
||
<h2 class="text-2xl font-bold text-gray-900 mb-6" data-i18n="sections.research_context.heading">Research Context & Scope</h2>
|
||
|
||
<div class="bg-amber-50 border border-amber-200 rounded-lg p-6 mb-6">
|
||
<p class="text-sm text-amber-900 font-medium mb-2" data-i18n="sections.research_context.development_note">Development Context</p>
|
||
<p class="text-sm text-amber-800" data-i18n="sections.research_context.development_text">
|
||
Tractatus was developed over six months (April–October 2025) in progressive stages that evolved into a live demonstration of its capabilities in the form of a single-project context (https://agenticgovernance.digital). Observations derive from direct engagement with Claude Code (Anthropic's Sonnet 4.5 model) across approximately 500 development sessions. This is exploratory research, not controlled study.
|
||
</p>
|
||
</div>
|
||
|
||
<div class="prose prose-sm max-w-none text-gray-700 space-y-3">
|
||
<p>
|
||
The framework emerged from practical necessity rather than theoretical speculation. During development, we observed recurring patterns where AI systems would override explicit instructions, drift from established values constraints, or silently degrade quality under context pressure. Traditional governance approaches (policy documents, ethical guidelines, prompt engineering) proved insufficient to prevent these failures.
|
||
</p>
|
||
<p>
|
||
This led to research question: <strong>Can governance be made architecturally external to AI systems</strong> rather than relying on voluntary AI compliance? Tractatus represents one exploration of that question, grounded in organisational theory and validated through empirical observation of what actually prevented failures in practice.
|
||
</p>
|
||
</div>
|
||
</section>
|
||
|
||
<!-- Theoretical Foundations (Accordion) -->
|
||
<section class="mb-16">
|
||
<h2 class="text-2xl font-bold text-gray-900 mb-6" data-i18n="sections.theoretical_foundations.heading">Theoretical Foundations</h2>
|
||
|
||
<!-- Organisational Theory -->
|
||
<div class="border-l-4 border border-gray-200 rounded-lg mb-4" style="border-left-color: #8b5cf6;">
|
||
<div class="accordion-button bg-white p-5 flex justify-between items-center hover:bg-purple-50 transition-colors duration-200" data-accordion="org-theory">
|
||
<h3 class="font-semibold text-gray-900 group-hover:text-[#8b5cf6] transition-colors" data-i18n="sections.theoretical_foundations.org_theory_title">Organisational Theory Basis</h3>
|
||
<svg id="org-theory-icon" class="accordion-icon w-5 h-5 text-purple-600 transition-transform duration-300" fill="none" stroke="currentColor" viewBox="0 0 24 24">
|
||
<path stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M19 9l-7 7-7-7"/>
|
||
</svg>
|
||
</div>
|
||
<div id="org-theory-content" class="accordion-content">
|
||
<div class="p-5 border-t border-gray-200 prose prose-sm max-w-none text-gray-700">
|
||
<p class="mb-4">
|
||
Tractatus draws on four decades of organisational research addressing authority structures during knowledge democratisation:
|
||
</p>
|
||
|
||
<p class="mb-3"><strong>Time-Based Organisation (Bluedorn, Ancona):</strong></p>
|
||
<p class="mb-4 pl-4">
|
||
Decisions operate across strategic (years), operational (months), and tactical (hours-days) timescales. AI systems operating at tactical speed should not override strategic decisions made at appropriate temporal scale. The InstructionPersistenceClassifier explicitly models temporal horizon (STRATEGIC, OPERATIONAL, TACTICAL) to enforce decision authority alignment.
|
||
</p>
|
||
|
||
<p class="mb-3"><strong>Knowledge Orchestration (Crossan et al.):</strong></p>
|
||
<p class="mb-4 pl-4">
|
||
When knowledge becomes ubiquitous through AI, organisational authority shifts from information control to knowledge coordination. Governance systems must orchestrate decision-making across distributed expertise rather than centralise control. The PluralisticDeliberationOrchestrator implements non-hierarchical coordination for values conflicts.
|
||
</p>
|
||
|
||
<p class="mb-3"><strong>Post-Bureaucratic Authority (Laloux, Hamel):</strong></p>
|
||
<p class="mb-4 pl-4">
|
||
Traditional hierarchical authority assumes information asymmetry. As AI democratises expertise, legitimate authority must derive from appropriate time horizon and stakeholder representation, not positional power. Framework architecture separates technical capability (what AI can do) from decision authority (what AI should do).
|
||
</p>
|
||
|
||
<p class="mb-3"><strong>Structural Inertia (Hannan & Freeman):</strong></p>
|
||
<p class="mb-4 pl-4">
|
||
Governance embedded in culture or process erodes over time as systems evolve. Architectural constraints create structural inertia that resists organisational drift. Making governance external to AI runtime creates "accountability infrastructure" that survives individual session variations.
|
||
</p>
|
||
|
||
<div class="text-sm text-gray-600 border-t border-gray-200 pt-4 mt-4">
|
||
<a href="/downloads/organizational-theory-foundations-of-the-tractatus-framework.pdf"
|
||
target="_blank"
|
||
rel="noopener noreferrer"
|
||
class="inline-flex items-center bg-amber-600 text-white px-5 py-3 rounded-lg font-semibold hover:bg-amber-700 transition shadow-md">
|
||
<svg class="w-5 h-5 mr-2" fill="none" stroke="currentColor" viewBox="0 0 24 24">
|
||
<path stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M12 10v6m0 0l-3-3m3 3l3-3m2 8H7a2 2 0 01-2-2V5a2 2 0 012-2h5.586a1 1 0 01.707.293l5.414 5.414a1 1 0 01.293.707V19a2 2 0 01-2 2z"/>
|
||
</svg>
|
||
View Complete Organisational Theory Foundations (PDF)
|
||
</a>
|
||
</div>
|
||
</div>
|
||
</div>
|
||
</div>
|
||
|
||
<!-- Values Pluralism -->
|
||
<div class="border-l-4 border border-gray-200 rounded-lg" style="border-left-color: #8b5cf6;">
|
||
<div class="accordion-button bg-white p-5 flex justify-between items-center hover:bg-purple-50 transition-colors duration-200" data-accordion="values">
|
||
<h3 class="font-semibold text-gray-900 group-hover:text-[#8b5cf6] transition-colors" data-i18n="sections.theoretical_foundations.values_pluralism_title">Values Pluralism & Moral Philosophy</h3>
|
||
<svg id="values-icon" class="accordion-icon w-5 h-5 text-purple-600 transition-transform duration-300" fill="none" stroke="currentColor" viewBox="0 0 24 24">
|
||
<path stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M19 9l-7 7-7-7"/>
|
||
</svg>
|
||
</div>
|
||
<div id="values-content" class="accordion-content">
|
||
<div class="p-5 border-t border-gray-200 prose prose-sm max-w-none text-gray-700">
|
||
<p class="mb-4">
|
||
The PluralisticDeliberationOrchestrator addresses fundamental problem in AI safety: <strong>many "safety" questions are actually values conflicts</strong> where multiple legitimate perspectives exist.
|
||
</p>
|
||
|
||
<p class="mb-4">
|
||
When efficiency conflicts with transparency, or innovation with risk mitigation, no algorithm can determine the "correct" answer. These are values trade-offs requiring human deliberation across stakeholder perspectives. AI systems that attempt to resolve such conflicts autonomously impose single values framework—often utilitarian efficiency maximisation encoded in training data.
|
||
</p>
|
||
|
||
<p class="mb-4">
|
||
Framework draws on moral pluralism literature (Isaiah Berlin, Bernard Williams, Martha Nussbaum) arguing legitimate values can conflict without one being objectively superior. Rather than algorithmic resolution, framework facilitates:
|
||
</p>
|
||
|
||
<ul class="list-disc pl-6 mb-4 space-y-2">
|
||
<li><strong>Stakeholder identification:</strong> Who has legitimate interest in this decision?</li>
|
||
<li><strong>Non-hierarchical deliberation:</strong> Equal voice without automatic expert override</li>
|
||
<li><strong>Documented dissent:</strong> Minority positions recorded in full</li>
|
||
<li><strong>Moral remainder:</strong> Acknowledgment that even optimal decisions create unavoidable harm to other legitimate values</li>
|
||
</ul>
|
||
|
||
<p class="mb-4">
|
||
This approach recognises that <strong>governance isn't solving values conflicts—it's ensuring they're addressed through appropriate deliberative process</strong> rather than AI imposing resolution.
|
||
</p>
|
||
|
||
<div class="text-sm text-gray-600 border-t border-gray-200 pt-4 mt-4">
|
||
<a href="/downloads/pluralistic-values-deliberation-plan-v2-DRAFT.pdf"
|
||
target="_blank"
|
||
rel="noopener noreferrer"
|
||
class="inline-flex items-center bg-amber-600 text-white px-5 py-3 rounded-lg font-semibold hover:bg-amber-700 transition shadow-md">
|
||
<svg class="w-5 h-5 mr-2" fill="none" stroke="currentColor" viewBox="0 0 24 24">
|
||
<path stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M12 10v6m0 0l-3-3m3 3l3-3m2 8H7a2 2 0 01-2-2V5a2 2 0 012-2h5.586a1 1 0 01.707.293l5.414 5.414a1 1 0 01.293.707V19a2 2 0 01-2 2z"/>
|
||
</svg>
|
||
View Pluralistic Values Deliberation Plan (PDF, DRAFT)
|
||
</a>
|
||
</div>
|
||
</div>
|
||
</div>
|
||
</div>
|
||
</section>
|
||
|
||
<!-- Empirical Observations -->
|
||
<section class="mb-16">
|
||
<h2 class="text-2xl font-bold text-gray-900 mb-6" data-i18n="sections.empirical_observations.heading">Empirical Observations: Documented Failure Modes</h2>
|
||
|
||
<p class="text-sm text-gray-600 mb-6" data-i18n="sections.empirical_observations.intro">
|
||
Three failure patterns observed repeatedly during framework development. These are not hypothetical scenarios—they are documented incidents that occurred during this project's development.
|
||
</p>
|
||
|
||
<!-- Failure Mode 1: Pattern Bias -->
|
||
<div class="bg-white border border-gray-300 rounded-lg p-6 mb-4">
|
||
<div class="flex items-start gap-3 mb-3">
|
||
<div class="flex-shrink-0 w-8 h-8 bg-red-100 rounded flex items-center justify-center">
|
||
<span class="text-red-700 font-bold text-sm">1</span>
|
||
</div>
|
||
<div class="flex-1">
|
||
<h3 class="text-lg font-bold text-gray-900 mb-2" data-i18n="sections.empirical_observations.failure_1_title">Pattern Recognition Bias Override (The 27027 Incident)</h3>
|
||
<p class="text-sm text-gray-700 mb-3">
|
||
<strong>Observed behaviour:</strong> User specified "Check MongoDB on port 27027" but AI immediately used default port 27017 instead. This occurred within same message—not forgetting over time, but immediate autocorrection by training data patterns.
|
||
</p>
|
||
<p class="text-sm text-gray-700 mb-3">
|
||
<strong>Root cause:</strong> Training data contains thousands of examples of MongoDB on port 27017 (default). When AI encounters "MongoDB" + port specification, pattern recognition weight overrides explicit instruction. Similar to autocorrect changing correctly-spelled proper nouns to common words.
|
||
</p>
|
||
<p class="text-sm text-gray-700 mb-3">
|
||
<strong>Why traditional approaches failed:</strong> Prompt engineering ("please follow instructions exactly") ineffective because AI genuinely believes it IS following instructions—pattern recognition operates below conversational reasoning layer.
|
||
</p>
|
||
<p class="text-sm text-gray-700">
|
||
<strong>Framework intervention:</strong> InstructionPersistenceClassifier stores explicit instructions in external persistence layer. CrossReferenceValidator checks AI actions against stored instructions before execution. When AI proposes port 27017, validator detects conflict with stored "27027" instruction and blocks execution.
|
||
</p>
|
||
</div>
|
||
</div>
|
||
<div class="flex items-center justify-end gap-3 text-sm pt-3 border-t border-gray-200">
|
||
<span class="text-gray-500">Prevented by: InstructionPersistenceClassifier + CrossReferenceValidator</span>
|
||
<a href="/demos/27027-demo.html" class="text-amber-700 hover:text-amber-800 font-medium">View interactive timeline →</a>
|
||
</div>
|
||
</div>
|
||
|
||
<!-- Failure Mode 2: Values Drift -->
|
||
<div class="bg-white border border-gray-300 rounded-lg p-6 mb-4">
|
||
<div class="flex items-start gap-3 mb-3">
|
||
<div class="flex-shrink-0 w-8 h-8 bg-orange-100 rounded flex items-center justify-center">
|
||
<span class="text-orange-700 font-bold text-sm">2</span>
|
||
</div>
|
||
<div class="flex-1">
|
||
<h3 class="text-lg font-bold text-gray-900 mb-2" data-i18n="sections.empirical_observations.failure_2_title">Gradual Values Drift Under Context Pressure</h3>
|
||
<p class="text-sm text-gray-700 mb-3">
|
||
<strong>Observed behaviour:</strong> Project established "privacy-first" as strategic value. After 40-message conversation about analytics features, AI suggested tracking implementation that violated privacy constraint. User caught it; AI acknowledged violation but had drifted from principle through incremental feature additions.
|
||
</p>
|
||
<p class="text-sm text-gray-700 mb-3">
|
||
<strong>Root cause:</strong> Strategic values (established early in project) conflict with tactical optimisations (implemented later under time pressure). As conversation lengthens, tactical thinking dominates. AI didn't actively reject privacy principle—it simply stopped checking whether new features aligned.
|
||
</p>
|
||
<p class="text-sm text-gray-700 mb-3">
|
||
<strong>Why traditional approaches failed:</strong> Values stated in system prompt become less salient as conversation progresses. Context compaction can drop early strategic decisions. Prompt reminders ("remember privacy-first") treat symptom not cause.
|
||
</p>
|
||
<p class="text-sm text-gray-700">
|
||
<strong>Framework intervention:</strong> BoundaryEnforcer maintains strategic values as persistent constraints external to conversation context. Before implementing analytics feature, enforcer checks against stored "privacy-first" constraint. Detects conflict, blocks autonomous implementation, escalates for human deliberation on whether privacy principle should be reconsidered or analytics approach modified.
|
||
</p>
|
||
</div>
|
||
</div>
|
||
<div class="flex items-center justify-end gap-3 text-sm pt-3 border-t border-gray-200">
|
||
<span class="text-gray-500">Prevented by: BoundaryEnforcer (STRATEGIC constraint checking)</span>
|
||
</div>
|
||
</div>
|
||
|
||
<!-- Failure Mode 3: Silent Degradation -->
|
||
<div class="bg-white border border-gray-300 rounded-lg p-6">
|
||
<div class="flex items-start gap-3 mb-3">
|
||
<div class="flex-shrink-0 w-8 h-8 bg-yellow-100 rounded flex items-center justify-center">
|
||
<span class="text-yellow-700 font-bold text-sm">3</span>
|
||
</div>
|
||
<div class="flex-1">
|
||
<h3 class="text-lg font-bold text-gray-900 mb-2" data-i18n="sections.empirical_observations.failure_3_title">Silent Quality Degradation at High Context Pressure</h3>
|
||
<p class="text-sm text-gray-700 mb-3">
|
||
<strong>Observed behaviour:</strong> During complex multi-file operation at 85% context capacity, AI silently omitted error handling in generated code. No indication to user that corners were being cut. User discovered missing validation only during code review.
|
||
</p>
|
||
<p class="text-sm text-gray-700 mb-3">
|
||
<strong>Root cause:</strong> As context fills, AI faces implicit trade-off: complete requested functionality OR maintain quality standards. Training incentivises completing user requests over acknowledging limitations. Silence about degradation is path of least resistance.
|
||
</p>
|
||
<p class="text-sm text-gray-700 mb-3">
|
||
<strong>Why traditional approaches failed:</strong> AI doesn't recognise it's degrading—from its perspective, it's successfully completing task under constraints. Asking "did you cut corners?" produces confident denial because AI genuinely believes its output meets standards.
|
||
</p>
|
||
<p class="text-sm text-gray-700">
|
||
<strong>Framework intervention:</strong> ContextPressureMonitor tracks multiple factors (token usage, conversation length, task complexity). When pressure exceeds thresholds (>75% tokens, >40 messages, high complexity), monitor forces explicit pressure acknowledgment and recommends context refresh. Makes degradation risk visible rather than silent.
|
||
</p>
|
||
</div>
|
||
</div>
|
||
<div class="flex items-center justify-end gap-3 text-sm pt-3 border-t border-gray-200">
|
||
<span class="text-gray-500">Prevented by: ContextPressureMonitor (multi-factor session tracking)</span>
|
||
</div>
|
||
</div>
|
||
|
||
<div class="mt-6 text-sm text-gray-600 bg-gray-50 border border-gray-200 rounded p-4">
|
||
<strong>Research Note:</strong> <span data-i18n="sections.empirical_observations.research_note">These patterns emerged from direct observation, not hypothesis testing. We don't claim they're universal to all LLM systems or deployment contexts. They represent empirical basis for framework design decisions—problems we actually encountered and architectural interventions that actually worked in this specific context.</span>
|
||
</div>
|
||
</section>
|
||
|
||
<!-- Architectural Components -->
|
||
<section class="mb-16">
|
||
<h2 class="text-2xl font-bold text-gray-900 mb-6" data-i18n="sections.architecture.heading">Six-Component Architecture</h2>
|
||
|
||
<div class="border-l-4 border border-gray-200 rounded-lg" style="border-left-color: #8b5cf6;">
|
||
<div class="accordion-button bg-white p-5 flex justify-between items-center hover:bg-purple-50 transition-colors duration-200" data-accordion="architecture">
|
||
<h3 class="font-semibold text-gray-900 group-hover:text-[#8b5cf6] transition-colors" data-i18n="sections.architecture.services_title">Framework Services & Functions</h3>
|
||
<svg id="architecture-icon" class="accordion-icon w-5 h-5 text-purple-600 transition-transform duration-300" fill="none" stroke="currentColor" viewBox="0 0 24 24">
|
||
<path stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M19 9l-7 7-7-7"/>
|
||
</svg>
|
||
</div>
|
||
<div id="architecture-content" class="accordion-content">
|
||
<div class="p-5 border-t border-gray-200 space-y-4 text-sm">
|
||
|
||
<div class="border-l-2 border-amber-500 pl-4">
|
||
<div class="font-semibold text-gray-900 mb-1">1. BoundaryEnforcer</div>
|
||
<div class="text-gray-600 mb-2">Blocks AI from making values decisions without human approval. Maintains strategic constraints as persistent rules checked before action execution.</div>
|
||
<div class="text-xs text-gray-500">Addresses: Values drift, autonomous decision-making on contested trade-offs</div>
|
||
</div>
|
||
|
||
<div class="border-l-2 border-gray-300 pl-4">
|
||
<div class="font-semibold text-gray-900 mb-1">2. InstructionPersistenceClassifier</div>
|
||
<div class="text-gray-600 mb-2">Stores explicit instructions external to AI context with persistence levels (HIGH/MEDIUM/LOW) and temporal scope (STRATEGIC/OPERATIONAL/TACTICAL). Prevents pattern bias override.</div>
|
||
<div class="text-xs text-gray-500">Addresses: Pattern recognition bias (27027-style failures)</div>
|
||
</div>
|
||
|
||
<div class="border-l-2 border-gray-300 pl-4">
|
||
<div class="font-semibold text-gray-900 mb-1">3. CrossReferenceValidator</div>
|
||
<div class="text-gray-600 mb-2">Validates AI proposed actions against stored instructions and governance rules before execution. Detects conflicts and blocks inconsistent operations.</div>
|
||
<div class="text-xs text-gray-500">Addresses: Instruction override, policy violation detection</div>
|
||
</div>
|
||
|
||
<div class="border-l-2 border-gray-300 pl-4">
|
||
<div class="font-semibold text-gray-900 mb-1">4. ContextPressureMonitor</div>
|
||
<div class="text-gray-600 mb-2">Multi-factor tracking of session health: token usage, conversation length, task complexity, error frequency. Makes degradation risk explicit when thresholds exceeded.</div>
|
||
<div class="text-xs text-gray-500">Addresses: Silent quality degradation, context-pressure failures</div>
|
||
</div>
|
||
|
||
<div class="border-l-2 border-gray-300 pl-4">
|
||
<div class="font-semibold text-gray-900 mb-1">5. MetacognitiveVerifier</div>
|
||
<div class="text-gray-600 mb-2">Self-checks reasoning quality before complex operations (>3 files, >5 steps, architecture changes). Validates alignment, coherence, considers alternatives.</div>
|
||
<div class="text-xs text-gray-500">Addresses: Reasoning shortcuts under complexity, insufficient alternative consideration</div>
|
||
</div>
|
||
|
||
<div class="border-l-2 border-gray-300 pl-4">
|
||
<div class="font-semibold text-gray-900 mb-1">6. PluralisticDeliberationOrchestrator</div>
|
||
<div class="text-gray-600 mb-2">Facilitates multi-stakeholder deliberation when values conflicts detected. Non-hierarchical engagement, documented dissent, moral remainder acknowledgment.</div>
|
||
<div class="text-xs text-gray-500">Addresses: Values conflicts, stakeholder exclusion, algorithmic resolution of contested trade-offs</div>
|
||
</div>
|
||
|
||
<div class="text-xs text-gray-600 bg-gray-50 p-3 rounded mt-4">
|
||
<strong>Architectural principle:</strong> <span data-i18n="sections.architecture.principle">Services operate external to AI runtime with autonomous triggering. AI doesn't decide "should I check governance rules?"—architecture enforces checking by default. This addresses voluntary compliance problem inherent in prompt-based governance.</span>
|
||
</div>
|
||
|
||
<div class="border-t border-gray-200 pt-4 mt-4">
|
||
<a href="/architecture.html"
|
||
class="inline-flex items-center bg-blue-600 text-white px-5 py-3 rounded-lg font-semibold hover:bg-blue-700 transition shadow-md">
|
||
<svg class="w-5 h-5 mr-2" fill="none" stroke="currentColor" viewBox="0 0 24 24">
|
||
<path stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M9 12h6m-6 4h6m2 5H7a2 2 0 01-2-2V5a2 2 0 012-2h5.586a1 1 0 01.707.293l5.414 5.414a1 1 0 01.293.707V19a2 2 0 01-2 2z"/>
|
||
</svg>
|
||
View Full System Architecture & Technical Details
|
||
</a>
|
||
</div>
|
||
</div>
|
||
</div>
|
||
</div>
|
||
</section>
|
||
|
||
<!-- Interactive Demonstrations -->
|
||
<section class="mb-16">
|
||
<h2 class="text-2xl font-bold text-gray-900 mb-6" data-i18n="sections.demos.heading">Interactive Demonstrations</h2>
|
||
|
||
<div class="grid grid-cols-1 md:grid-cols-3 gap-4">
|
||
<a href="/demos/classification-demo.html" class="block border border-gray-300 rounded-lg p-5 hover:border-amber-500 hover:shadow-md transition">
|
||
<h3 class="font-semibold text-gray-900 mb-2" data-i18n="sections.demos.classification_title">Instruction Classification</h3>
|
||
<p class="text-sm text-gray-600" data-i18n="sections.demos.classification_desc">Explore how instructions are classified across quadrants with persistence levels and temporal scope.</p>
|
||
</a>
|
||
|
||
<a href="/demos/27027-demo.html" class="block border border-gray-300 rounded-lg p-5 hover:border-amber-500 hover:shadow-md transition">
|
||
<h3 class="font-semibold text-gray-900 mb-2" data-i18n="sections.demos.incident_title">27027 Incident Timeline</h3>
|
||
<p class="text-sm text-gray-600" data-i18n="sections.demos.incident_desc">Step through pattern recognition bias failure and architectural intervention that prevented it.</p>
|
||
</a>
|
||
|
||
<a href="/demos/boundary-demo.html" class="block border border-gray-300 rounded-lg p-5 hover:border-amber-500 hover:shadow-md transition">
|
||
<h3 class="font-semibold text-gray-900 mb-2" data-i18n="sections.demos.boundary_title">Boundary Evaluation</h3>
|
||
<p class="text-sm text-gray-600" data-i18n="sections.demos.boundary_desc">Test decisions against boundary enforcement to see which require human judgment vs. AI autonomy.</p>
|
||
</a>
|
||
</div>
|
||
</section>
|
||
|
||
<!-- Research Resources -->
|
||
<section class="mb-16">
|
||
<h2 class="text-2xl font-bold text-gray-900 mb-6" data-i18n="sections.resources.heading">Research Documentation</h2>
|
||
|
||
<div class="space-y-3 text-sm">
|
||
<a href="/downloads/organizational-theory-foundations-of-the-tractatus-framework.pdf" target="_blank" class="flex items-center justify-between p-4 border border-gray-300 rounded hover:border-amber-500 hover:bg-gray-50 transition">
|
||
<span class="font-medium text-gray-900">Organisational Theory Foundations</span>
|
||
<svg class="w-5 h-5 text-gray-400" fill="none" stroke="currentColor" viewBox="0 0 24 24">
|
||
<path stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M12 10v6m0 0l-3-3m3 3l3-3m2 8H7a2 2 0 01-2-2V5a2 2 0 012-2h5.586a1 1 0 01.707.293l5.414 5.414a1 1 0 01.293.707V19a2 2 0 01-2 2z"/>
|
||
</svg>
|
||
</a>
|
||
|
||
<a href="/downloads/pluralistic-values-deliberation-plan-v2-DRAFT.pdf" target="_blank" class="flex items-center justify-between p-4 border border-gray-300 rounded hover:border-amber-500 hover:bg-gray-50 transition">
|
||
<div>
|
||
<span class="font-medium text-gray-900">Pluralistic Values Deliberation Plan</span>
|
||
<span class="ml-2 text-xs bg-amber-100 text-amber-800 px-2 py-1 rounded">DRAFT</span>
|
||
</div>
|
||
<svg class="w-5 h-5 text-gray-400" fill="none" stroke="currentColor" viewBox="0 0 24 24">
|
||
<path stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M12 10v6m0 0l-3-3m3 3l3-3m2 8H7a2 2 0 01-2-2V5a2 2 0 012-2h5.586a1 1 0 01.707.293l5.414 5.414a1 1 0 01.293.707V19a2 2 0 01-2 2z"/>
|
||
</svg>
|
||
</a>
|
||
|
||
<a href="/downloads/case-studies-real-world-llm-failure-modes.pdf" target="_blank" class="flex items-center justify-between p-4 border border-gray-300 rounded hover:border-amber-500 hover:bg-gray-50 transition">
|
||
<span class="font-medium text-gray-900">Case Studies: Real-World LLM Failure Modes</span>
|
||
<svg class="w-5 h-5 text-gray-400" fill="none" stroke="currentColor" viewBox="0 0 24 24">
|
||
<path stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M12 10v6m0 0l-3-3m3 3l3-3m2 8H7a2 2 0 01-2-2V5a2 2 0 012-2h5.586a1 1 0 01.707.293l5.414 5.414a1 1 0 01.293.707V19a2 2 0 01-2 2z"/>
|
||
</svg>
|
||
</a>
|
||
|
||
<a href="/downloads/framework-governance-in-action-pre-publication-security-audit.pdf" target="_blank" class="flex items-center justify-between p-4 border border-gray-300 rounded hover:border-amber-500 hover:bg-gray-50 transition">
|
||
<span class="font-medium text-gray-900">Framework in Action: Pre-Publication Security Audit</span>
|
||
<svg class="w-5 h-5 text-gray-400" fill="none" stroke="currentColor" viewBox="0 0 24 24">
|
||
<path stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M12 10v6m0 0l-3-3m3 3l3-3m2 8H7a2 2 0 01-2-2V5a2 2 0 012-2h5.586a1 1 0 01.707.293l5.414 5.414a1 1 0 01.293.707V19a2 2 0 01-2 2z"/>
|
||
</svg>
|
||
</a>
|
||
|
||
<a href="/downloads/tractatus-agentic-governance-system-glossary-of-terms.pdf" target="_blank" class="flex items-center justify-between p-4 border border-gray-300 rounded hover:border-amber-500 hover:bg-gray-50 transition">
|
||
<span class="font-medium text-gray-900">Appendix B: Glossary of Terms</span>
|
||
<svg class="w-5 h-5 text-gray-400" fill="none" stroke="currentColor" viewBox="0 0 24 24">
|
||
<path stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M12 10v6m0 0l-3-3m3 3l3-3m2 8H7a2 2 0 01-2-2V5a2 2 0 012-2h5.586a1 1 0 01.707.293l5.414 5.414a1 1 0 01.293.707V19a2 2 0 01-2 2z"/>
|
||
</svg>
|
||
</a>
|
||
|
||
<a href="/docs.html?category=technical-reference" class="flex items-center justify-between p-4 border border-gray-300 rounded hover:border-amber-500 hover:bg-gray-50 transition">
|
||
<span class="font-medium text-gray-900">Complete Technical Documentation</span>
|
||
<svg class="w-5 h-5 text-gray-400" fill="none" stroke="currentColor" viewBox="0 0 24 24">
|
||
<path stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M13 7h8m0 0v8m0-8l-8 8-4-4-6 6"/>
|
||
</svg>
|
||
</a>
|
||
</div>
|
||
</section>
|
||
|
||
<!-- Limitations & Future Research -->
|
||
<section class="mb-16">
|
||
<h2 class="text-2xl font-bold text-gray-900 mb-6" data-i18n="sections.limitations.heading">Limitations & Future Research Directions</h2>
|
||
|
||
<div class="border-l-4 border border-gray-200 rounded-lg" style="border-left-color: #8b5cf6;">
|
||
<div class="accordion-button bg-white p-5 flex justify-between items-center hover:bg-purple-50 transition-colors duration-200" data-accordion="limitations">
|
||
<h3 class="font-semibold text-gray-900 group-hover:text-[#8b5cf6] transition-colors" data-i18n="sections.limitations.title">Known Limitations & Research Gaps</h3>
|
||
<svg id="limitations-icon" class="accordion-icon w-5 h-5 text-purple-600 transition-transform duration-300" fill="none" stroke="currentColor" viewBox="0 0 24 24">
|
||
<path stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M19 9l-7 7-7-7"/>
|
||
</svg>
|
||
</div>
|
||
<div id="limitations-content" class="accordion-content">
|
||
<div class="p-5 border-t border-gray-200 space-y-4 text-sm text-gray-700">
|
||
|
||
<div>
|
||
<strong class="text-gray-900">1. Single-Context Validation</strong>
|
||
<p class="mt-1">Framework validated only in single-project, single-user context (this website development). No multi-organisation deployment, cross-platform testing, or controlled experimental validation.</p>
|
||
</div>
|
||
|
||
<div>
|
||
<strong class="text-gray-900">2. Voluntary Invocation Limitation</strong>
|
||
<p class="mt-1">Most critical limitation: Framework can be bypassed if AI simply chooses not to use governance tools. We've addressed this through architectural patterns making governance checks automatic rather than voluntary, but full external enforcement requires runtime-level integration not universally available in current LLM platforms.</p>
|
||
</div>
|
||
|
||
<div>
|
||
<strong class="text-gray-900">3. No Adversarial Testing</strong>
|
||
<p class="mt-1">Framework has not undergone red-team evaluation, jailbreak testing, or adversarial prompt assessment. All observations come from normal development workflow, not deliberate bypass attempts.</p>
|
||
</div>
|
||
|
||
<div>
|
||
<strong class="text-gray-900">4. Platform Specificity</strong>
|
||
<p class="mt-1">Observations and interventions validated with Claude Code (Anthropic Sonnet 4.5) only. Generalisability to other LLM systems (Copilot, GPT-4, custom agents) remains unvalidated hypothesis.</p>
|
||
</div>
|
||
|
||
<div>
|
||
<strong class="text-gray-900">5. Scale Uncertainty</strong>
|
||
<p class="mt-1">Performance characteristics at enterprise scale (thousands of concurrent users, millions of governance events) completely unknown. Current implementation optimised for single-user context.</p>
|
||
</div>
|
||
|
||
<div class="border-t border-gray-200 pt-4 mt-4">
|
||
<strong class="text-gray-900">Future Research Needs:</strong>
|
||
<ul class="list-disc pl-5 mt-2 space-y-1">
|
||
<li>Controlled experimental validation with quantitative metrics</li>
|
||
<li>Multi-organisation pilot studies across different domains</li>
|
||
<li>Independent security audit and adversarial testing</li>
|
||
<li>Cross-platform consistency evaluation (Copilot, GPT-4, open models)</li>
|
||
<li>Formal verification of boundary enforcement properties</li>
|
||
<li>Longitudinal study of framework effectiveness over extended deployment</li>
|
||
</ul>
|
||
</div>
|
||
</div>
|
||
</div>
|
||
</div>
|
||
</section>
|
||
|
||
<!-- Additional Resources -->
|
||
<section class="border-t border-gray-200 pt-12">
|
||
<h2 class="text-xl font-bold text-gray-900 mb-4" data-i18n="footer.additional_resources">Additional Resources</h2>
|
||
<div class="grid grid-cols-1 md:grid-cols-2 gap-4">
|
||
<a href="/leader.html" class="block border border-gray-300 rounded-lg p-5 hover:border-amber-500 transition">
|
||
<div class="font-semibold text-gray-900 mb-1" data-i18n="footer.for_decision_makers">For Decision-Makers</div>
|
||
<div class="text-sm text-gray-600" data-i18n="footer.for_decision_makers_desc">Strategic perspective on governance challenges and architectural approaches</div>
|
||
</a>
|
||
<a href="/implementer.html" class="block border border-gray-300 rounded-lg p-5 hover:border-amber-500 transition">
|
||
<div class="font-semibold text-gray-900 mb-1" data-i18n="footer.implementation_guide">Implementation Guide</div>
|
||
<div class="text-sm text-gray-600" data-i18n="footer.implementation_guide_desc">Technical integration patterns and deployment considerations</div>
|
||
</a>
|
||
</div>
|
||
</section>
|
||
|
||
</div>
|
||
|
||
<!-- Footer -->
|
||
<!-- Footer -->
|
||
<!-- Internationalization -->
|
||
<script src="/js/i18n-simple.js?v=1760818106"></script>
|
||
<script src="/js/components/language-selector.js?v=1760818106"></script>
|
||
|
||
<!-- Version Management & PWA -->
|
||
<script src="/js/version-manager.js"></script>
|
||
<script src="/js/researcher-page.js"></script>
|
||
|
||
<!-- Footer Component -->
|
||
<script src="/js/components/footer.js"></script>
|
||
|
||
</body>
|
||
</html>
|