"AI is changing the very nature of knowledge creation." Discuss the implications for national accounting and for evidence-based policymaking in India.

AI systems now write code, text and designs. In doing so they create knowledge "beyond conventional human knowledge creation", as MoSPI's Committee on Knowledge Systems (chaired by Ratan P. Watal) notes [1][2]. This makes knowledge harder for GDP to value. It also gives India a chance to target its policies better.

Implications for national accounting

• New asset class: The 2025 SNA was adopted by the UN Statistical Commission [3]. It counts the information content of electronic data used in production as an asset, which the 2008 SNA did not [4].
• No market price: Firms use most of their data and AI models themselves, so the data is rarely sold. The OECD finds that existing statistics capture data only partially [5]. Valuing it by sum of costs undervalues a cheaply built dataset that later trains a valuable model.
• Double-counting risk:

Data (input) → AI model (asset) → Code/text/designs (output)

Fig: the same value can be counted at each step. The paper's boundary conditions of "knowledge as input" and "economic value" need clear rules on what is counted at which step [1].

• Codification bias: The tests of codification and identifiability favour digital knowledge with a clear owner. Oral traditional knowledge may be left out [1].
• Response: MoSPI proposes a family of SNA 2025-consistent thematic accounts, with AI systems as a separate priority area, rather than one "knowledge GDP" figure [1].

Implications for evidence-based policymaking Opportunities

• Mapping gaps: The accounts would show India's strengths, gaps and priorities [1]. This can guide where R&D and skilling money goes.
• Federated model: Sectoral ministries measure their own areas, and States are key partners [1]. This is cooperative federalism in statistics.

Concerns

• Measuring a gap does not close it: GERD (total R&D spending) was estimated at 0.64% of GDP in 2020-21. It has been falling since 2009-10 and is far below the 2% goal of the STI Policy 2013 [6].
• Self-reporting bias: Ministries would measure the sectors they are judged on. The paper does not give the integrator any power to audit their estimates [1].
• Measurability bias: Under the "cost-effective measurability" criterion, areas with little data, such as MSME and informal knowledge, may be measured last [1].

AI therefore widens what the national accounts must capture, and it also gives policymakers sharper tools. The way forward is to:

• publish each account's methods before its numbers;
• label the AI accounts "experimental" until their methods are tested;
• give States a funded role in sharing data;
• report each ministry's account next to the budget it controls.

Done this way, measurement anchored in SNA 2025 can make India's knowledge economy visible and inclusive, in line with SDG 9.