context-access-divide-agentic-inequality-architecture

name	context-access-divide-agentic-inequality-architecture
description	Formalizes the Context Access Divide (CAD) as a dimension of agentic inequality operating at the interaction level. Dynamic Context Retrieval vs Manual Attachment causes combinatorial collapse in task-success probability. Proposes contextuality as complement to Sharp et al.'s framework. Activation: agentic inequality, context access, interaction-level architecture, agent fairness, AI equity.
metadata	{"arxiv_id":"2607.08495","published":"2026-07-09","authors":"Masahiro Fujita","tags":["agentic-inequality","context-access","interaction-level-architecture","agent-fairness","ai-equity"]}

The Context Access Divide: Interaction-Level Architecture as a Complementary Dimension of Agentic Inequality

Overview

Two users with nominally equivalent agent access may experience qualitatively different AI utility depending on whether the system can autonomously retrieve context from the user's knowledge corpus (Dynamic Context Retrieval) or requires the user to manually identify and attach relevant documents at each query (Manual Attachment). This paper formalizes the Context Access Divide (CAD) as a structural dimension of AI-mediated inequality.

Key Innovations

Context Access Divide (CAD)

• Interaction-level dimension of agentic inequality
• Complements Sharp et al.'s (2025) availability, quality, quantity dimensions
• Dynamic Context Retrieval vs. Manual Attachment as a qualitative threshold
• For knowledge workers with tens of thousands of files, CAD determines AI usefulness

Probabilistic Model

• Grounded in the fan effect literature from cognitive psychology
• Shows manual context attachment leads to combinatorial collapse in task-success probability
• Collapse worsens as corpus size and task conjunctivity grow
• Dynamic retrieval architectures are structurally insulated from this collapse

Technical Analysis

• Analyzes CAD's technical basis in Model Context Protocol (MCP) and RAG architectures
• Examines implications for knowledge-work stratification
• Connects to AI platform governance

Methodology

1. Formalization: Define CAD as probabilistic model with fan effect grounding
2. Analysis: Compare Dynamic Context Retrieval vs. Manual Attachment
3. Technical Mapping: Map to MCP and RAG architectures
4. Implications: Analyze knowledge-work stratification and governance

Implications

• Agent access equality requires context access infrastructure, not just model access
• Manual context attachment is a structural barrier to AI utility
• MCP and RAG architectures have equity implications
• Platform governance must consider contextuality as a design dimension

Pitfalls

• Probabilistic model makes simplifying assumptions about user behavior
• Fan effect literature may not directly translate to AI-assisted tasks
• Binary framing (dynamic vs. manual) may oversimplify spectrum of solutions
• Governance implications are theoretical without empirical validation

Activation Keywords

agentic inequality, Context Access Divide, contextuality, dynamic context retrieval, manual attachment, MCP, RAG, knowledge-work stratification, AI equity, fan effect

Paper Reference

arXiv:2607.08495 - "The Context Access Divide: Interaction-Level Architecture as a Complementary Dimension of Agentic Inequality" (Jul 2026)