🧠 AI Ed Wiki

📄 Research Article

Structural Silence: When AI Infrastructure Fails Speakers of Underrepresented Languages

● high · 2026-08-13equity, multilingual-learning, language-learning, digital-divide, global-south, ai-education, inclusive-ai, higher-ed

Synthesis: Roy & Roy (2026) argue that the infrastructure of AI — training corpora, tokenization, benchmarks, deployment architectures — systematically disadvantages speakers of underrepresented languages before a model is trained, reframing dataset scarcity as a structural barrier rather than an isolated technical limitation. Using Bengali as a case in AI-assisted education, they document four interlocking failures: a web-presence gap (<0.5% of global content for ~4% of the population), a 67:1 English↔Bengali training-token deficit, a tokenization penalty from the alphasyllabary script, and connectivity exclusion (36.5% rural vs 71.4% urban internet penetration). They position offline-first design as an equity-oriented infrastructure strategy. The work connects to Equity, Language Learning, and Digital Divide debates in educational AI.

Four Interlocking Infrastructure Failures

The paper identifies four structural barriers that compound to exclude underrepresented languages from AI-assisted education:

  • Web presence gap: Bengali accounts for under 0.5% of global web content despite representing nearly 4% of the global population.
  • Training-token deficit: a 67:1 deficit between English and Bengali in major multilingual corpora.
  • Tokenization penalty: Bengali's alphasyllabary script compounds the data deficit through higher token fertility.
  • Connectivity exclusion: individual internet penetration is 36.5% in rural areas versus 71.4% in urban areas.

    • These failures reflect longstanding resource-allocation decisions, institutional priorities, and design defaults that did not center underrepresented languages in mainstream AI development.

    Reframing Scarcity as Structure

    The authors argue dataset scarcity should be understood as a structural barrier rather than an isolated technical limitation. They recommend treating offline-first design as an equity-oriented infrastructure strategy for AI-assisted education in low-connectivity environments, and outline directions for linguistics and AI research aimed at reducing these structural inequalities.

    Connected Concepts

  • Equity
  • Language Learning
  • Language Learning
  • Digital Divide
  • Equity In AI Education
  • Equity In AI Education
  • AI Education
  • Higher Ed
  • Privacy
  • Accessible Learning

    • Connected Articles

  • AI Scoring Language Bias Physics
  • Sec AI Literacy Narrative Review 2026
  • Data Comics For Education Evaluating Effectiveness Benefits Ethics

    • Citation

    Roy, A., & Roy, P. (2026). Structural silence: When AI infrastructure fails speakers of underrepresented languages. arXiv:2608.12278.