Gardiner develops knowledge theory of capital value

The arXiv submission A Knowledge Theory of Capital: The Value of Natural and Artificial Intelligence by Jeffrey Gardiner (submitted 12 Jun 2026) proposes a formal economics treatment of knowledge as stock. Per the paper's abstract, Gardiner starts from Adam Smith's theory of labour, stock, specialization, and market extent and asks what changes when knowledge becomes stock-like, mobile across forms, scalable, governable, recombinable, and imperfectly visible in accounting. The abstract states the paper introduces knowledge-bearing stock as the central analytic object and frames the argument as conditional and testable. The abstract lists new terminology: first conversion, cognitive enclosure, feedback capture, dark capital, and expected knowledge loss. The abstract also distinguishes five knowledge forms:

• •embodied
• •disembodied
• •institutionalized
• •commons
• •public

Finally, the abstract argues that modern wealth depends not only on capital accumulation but on how productive knowledge is governed, per the paper's abstract.

The paper reframes standard capital theory by making knowledge the inventory item analogous to physical stock, which shifts the unit of analysis toward reproducibility, mobility, and observability of productive processes. For practitioners, this matters because it foregrounds properties that also shape machine learning assets: reproducibility of models, data governance, feedback loops from deployed systems, and metrics for measuring intangible depreciation. The paper's terminology , for example, feedback capture and expected knowledge loss , maps onto operational concerns in ML lifecycle management such as model drift, data lineage, and intellectual property enclosure.

Observed patterns in comparable literature show increasing attention to governance, measurement, and institutional arrangements when intangible assets dominate productive capacity. Industry reporting and policy debates have similarly moved from treating software as capital to interrogating governance regimes for data and models. Gardiner's conditional, testable framing aligns with academic efforts to operationalize measurement of intangible capital rather than prescribe firm-level strategy.

Indicators that could validate or extend this framework include empirical measures of knowledge depreciation in deployed ML systems, case studies of cognitive enclosure (legal or technical), and econometric work linking knowledge governance to productivity. Researchers and data practitioners will likely look for the paper's proposed measurable constructs and suggested empirical tests in the full text and follow-up work.