Logo Lanfrica

Governing Generative AI Concentration: Policy Responses to Geographic, Epistemic, and Network Inequalities in Global ICT Innovation (2018–2024)

Domaine:

digital infrastructure

Type de record:

paper
Créateur:
SerESRFurAri
Éditeur:
Elsevier BV
Hôte:
This study examines policy implications of geographic, epistemic, and network concentration in generative artificial intelligence (GenAI) innovation. Analyzing 111,512 patent and publication records from 2018–2024 through bibliometric analysis, network science, and computational text analysis, we map three dimensions of concentration relevant to ICT governance. First, geographic distribution exhibits extreme inequality (Gini coefficient 0.762), with China and the United States jointly accounting for 58.1% of global GenAI innovation. Concentration intensifies over the observation period, with the Herfindahl-Hirschman Index rising from 1611.6 to 2358.6. Second, thematic analysis reveals that only 3.9% of documents address any inclusion theme under the validated baseline classification, with a pronounced hierarchy of attention: low-resource computing (1.5%) and accessibility (1.4%) dominate, while disadvantaged-population needs (0.3%) and digital-equity promotion (0.0%) are effectively absent. Third, international collaboration networks display preferential attachment dynamics (QAP β = 0.586, R2 = 0.723) consistent with cumulative advantage rather than equitable knowledge co-creation. We interpret these patterns through a decolonial framework that treats concentration as structural rather than incidental, while explicitly stating falsification conditions. We argue that the bipolar US–China configuration represents a new hegemon pattern in which a formerly peripheral state replicates rather than disrupts extractive dynamics. The paper develops concrete policy instruments across three governance domains: (i) data sovereignty frameworks (drawing on the Te Hiku Māori model and emerging African Union proposals); (ii) public investment in indigenous AI capacity (the Masakhane NLP and AI4D Africa models); and (iii) reciprocal collaboration funding (modelled on World Bank–AfDB co-investment instruments).

Similaires