No 2 (2026)

Abstract: Problem. Achieving technological sovereignty requires an analysis of the factors that ensure an effective return on R&D investment. Aim. To assess quantitatively the contribution of STEM education quality to the transformation of R&D expenditure into patent activity and to identify institutional factors that ensure a positive deviation of actual indicators from predicted ones (using China as a case study). Methods. The study combines qualitative analysis of China’s institutional model (including the “Made in China 2025” strategy, the Belt and Road Initiative, the “Double First-Class” program) with quantitative econometric modeling. The empirical basis comprises data from 77 countries for the period 2019–2023. Using regression analysis (OLS), the elasticity of patent activity (number of resident patent applications per capita) with respect to R&D expenditure (as a percentage of GDP) was estimated. Results. A direct relationship was identified between the quality of STEM education (PISA math 2018) and the efficiency of transforming R&D into patents: a 10-point increase in the PISA score is associated with a 6.4 % increase in patent activity, holding R&D expenditure constant. China is highlighted as an in-depth case study that verifies the model’s conclusions. A positive and statistically significant elasticity was established: a 1 % increase in R&D expenditure is accompanied by an average 1.21 % increase in patent activity, confirming a high return on investment in the scientific sphere. Based on country clustering, three groups were identified: high-efficiency countries (actual indicator close to predicted), medium-efficiency countries (actual indicator below predicted), and limited-efficiency countries (significant lag behind predicted). China, along with South Korea and Switzerland, is assigned to the first group. Conclusions. The key factors of the Chinese model are: large-scale training of STEM specialists (up to 10 million annually) with leading positions in PISA; institutional integration of education, science, and business according to the triple helix model; and the use of fiscal incentives based on the innovation funnel principle, combining direct government funding (2.5 % of GDP for R&D) with tax incentives.

Abstract: Problem. The proliferation of digital agents in business processes of companies produces multidirectional effects on their competitiveness, ranging from productivity growth to the loss of employee skills. A measurement apparatus that covers both aspects simultaneously and explains their origins does not yet exist, and this gap becomes critical as implementations scale up. Aim. To develop a categorical framework and a three-tier metric system that allows measuring the dual effects of digital agents, based on the causal mechanism of their generation and on the distinction of who makes the decision – a human or a digital agent. Methods. The work is carried out in a theoretical-methodological genre; its foundation consists of the four-step transfer of established frameworks in the economics of innovation from humans to digital agents while preserving the core of each original approach. The conceptual basis is formed by agency theory and the triad of artificial intelligence properties (autonomy, learnability, and opacity), with the unifying distinction being who makes decisions in the business process (a human or a digital agent). Results. A causal scheme is constructed in which the positive and negative sides of each effect arise within the same link of the mechanism. Eight channels of influence of digital agents on company competitiveness are identified, ranging from productivity to the composition of innovation authors. Three tiers of metrics are proposed, which differ in the degree to which the outcome depends on the decision-making subject. Each channel is assigned a primary measurement tier, and when an agent transitions to independent action, some channels move to higher tiers. Conclusions. The duality of the effects of digital agents is demonstrated as a structural property of a single mechanism, rather than a balance of independent factors. A three-tier classification of metrics is proposed based on sensitivity to who makes the decision – a human or a digital agent – and eight channels of influence are linked to these tiers. The resulting framework creates a foundation for the quantitative assessment of the impact of digital agents on company competitiveness, taking into account the observed heterogeneity of their implementation.