AI transforms personalization by shifting the cost of adaptation from physical retooling to concentrated cognitive work embedded in infrastructure, turning visible product standards into implicit rules controlled by platform and system designers.

Main Finding

AI fundamentally changes the role of standardization by expanding society's information-processing capacity along three dimensions—perception, computation, execution—so that complexity that industrial systems had to compress can now be accommodated. Standardization does not vanish; it shifts from explicit, product-level constraints toward implicit generation rules embedded in infrastructures. The central political-economic question thus moves from “whether to have commonality” to “who controls commonality.”

Key Points

• Core hypothesis: a society’s ability to accommodate individual differences is bounded by its information-processing capacity (perception, processing, execution). When capacity is limited, systems compress complexity via standardization; when capacity expands (AI era), systems can support richer individuality.
• Historical framing:
  • Pre-industrial: differences were embedded in craft experience and not objectified as operable information.
  • Industrial: severe limits in perception (sampling), processing (fixed plans), and execution (specialized tooling) made standardization the practical way to keep large-scale systems manageable (Taylorism, Fordism).
  • AI era: leap in sensing, computation, and flexible execution enables information-based adaptation and continuous objectification of differences.
• Two forms of standardization:
  • Explicit product-level standardization (industrial era): compresses output variety into discrete categories (S/M/L, fixed models).
  • Implicit infrastructural standardization (AI era): commonality lives as generative rules, APIs, models, protocols that generate individualized outputs—standardization is submerged into the stack.
• New analytical concepts:
  • Cognitive fixed cost: upstream concentration of cognitive labor (data labeling, model training, rules encoding) creates an upfront fixed cost that changes the cost structure of personalization; marginal cost of individualized outputs can be low once cognitive fixed costs are borne.
  • Information-based adaptation: personalization shifts from physical retooling toward informational instructions that reconfigure production/outputs.
  • Type implicitization: converting explicit discrete types into implicit generative procedures embedded in infrastructure.
• Political shift: the problem becomes governance of commonality (who controls embedded rules, data, models, APIs) rather than the existence of commonality itself.
• Methodological stance: the paper is descriptive/analytic, not normative; it places industrial standardization and AI-enabled personalization as institutional adaptations determined by information-processing constraints.

Data & Methods

• Conceptual operationalization:
  • Breaks "information-processing capacity" into three sub-capacities: perception (acquire/encode individual differences), processing (turn differences into production decisions), execution (physically enact differentiated instructions).
  • Uses Ashby’s Law of Requisite Variety, Simon’s bounded rationality/complexity, and Galbraith’s information-processing requirement to link variety, information needs, and system capacity.
• Method: comparative historical analysis across three periods (pre-industrial, industrial, AI era), drawing on technology and production-organization history (e.g., Taylorism, Ford’s Model T, Chandler on enterprise coordination).
• Evidence: qualitative/historical examples and logical extension of information-processing theory; no new quantitative dataset—argument is primarily theoretical and interpretative.
• Limitations acknowledged: does not claim AI is inherently liberatory or dystopian; focuses on changing capacity boundaries and institutional implications rather than moral evaluation.

Implications for AI Economics

• Changing cost structure and scale economics:
  • Cognitive fixed cost implies large upfront investments (data, models, rule-encoding) with low marginal cost for individualized outputs — leading to strong scale economies in cognitive inputs.
  • Firms that internalize or amortize cognitive fixed costs across many users can price personalized goods/services cheaply, generating winner-take-most dynamics.
• Firm organization and division of labor:
  • Expect centralization of cognitive tasks (model training, rule design) and decentralization/flexibility in execution (robotic/programmable manufacturing, digital delivery).
  • New organizational forms where firms act as infrastructure providers (models, APIs, design generators) and downstream firms/consumers compose individualized outputs.
• Market structure and platform power:
  • Control over generative infrastructure, data, and model parameters becomes the locus of market power. The question of “who controls commonality” creates strategic rents.
  • Lock-in and network effects intensify as more personalized outputs are produced on a common generative platform.
• Standards, interoperability, and regulation:
  • Traditional antitrust approaches focusing on product-level standardization need updating: regulators should consider control over embedded rules, protocol governance, data access, and model interoperability.
  • Policy levers may include mandates on data portability, model auditability, open interfaces, and standards governance to democratize control of commonality.
• Labor and skills:
  • Demand shifts from routine execution tasks toward cognitive design/configuration, dataset curation, and oversight roles (upstream cognitive labor).
  • Cognitive fixed costs favor a smaller number of high-skilled roles; displacement risk for execution workers can be mitigated if execution becomes lower-cost but still requires new supervisory/maintenance skills.
• Pricing, welfare, and distribution:
  • Personalization enables price discrimination and greater consumer surplus extraction but also creates opportunities for tailored welfare-improving products; distributional effects hinge on who captures cognitive rents.
  • Welfare analysis must consider both gains from better match between products and preferences and harms from concentrated control over personalization infrastructure.
• Innovation and product variety:
  • Ability to move from discrete to continuous objectification of differences increases product variety and niche markets, but many niches may be served by the same generative infrastructure.
  • Competition may shift to quality/ownership of generative models and data rather than incremental product specs.
• Research directions for AI economics:
  • Quantify cognitive fixed costs and their amortization across markets.
  • Model market power dynamics when commonality is embedded in shared AI infrastructure.
  • Study regulatory design for governance of implicit standards (data access, model transparency, interface interoperability).
  • Empirical work on labor reallocation between cognitive and execution tasks, and on distributional consequences of platform-controlled personalization.

Concluding note: The paper reframes standardization as an adaptive response to information-processing constraints. In the AI era, expansion of those constraints alters the institutional terrain: standardization persists but is reallocated into embedded generative rules, shifting economic stakes toward control of informational infrastructure.