How could you resolve AI inequality in higher education ?

AI inequality in Higher education

Marc LAURENT - Science Po Bordeaux ; Jade JEAN-ELIE - Université de Bordeaux ; Julia

Research (anything related to the use of AI in a research context)

AI is no longer an emerging technology (it is a rising star), now deeply embedded in our societies and educational systems. Trying to block its development or ban it from education would be pointless; it is already too integrated into our daily lives to be ignored. The real challenge is learning to live with it responsibly. We’ll look at this situation in higher education. The objective is to find solutions to AI inequality in higher education ?

The main problem is that AI doesn't just reflect existing inequalities, it actively amplifies them, particularly, in higher education. A concrete example is the gap between free and paid AI models: premium tools like GPT-4 or Claude Pro deliver significantly more accurate, nuanced, and reliable outputs than their free counterparts, yet they remain financially out of reach for many students. This creates a direct performance gap. Beyond access, the ability to prompt effectively has become a new form of capital, unevenly distributed along socioeconomic and linguistic lines. An English-speaking student with digital literacy will consistently extract higher-quality, more nuanced responses than a student prompting in Swahili or a regional dialect. Meanwhile, the concentration of digital infrastructure in urban centers drains rural areas of skilled talent, creating a self-reinforcing cycle of exclusion. Access alone, without education and linguistic inclusion, reproduces rather than corrects inequality.

Four angles were explored. Jade examined global AI governance highlighting how regulatory divergence (EU protecting citizens, US protecting business, China protecting the state) means geographic location alone determines access to the most powerful tools. Marc argued that access without education is insufficient — knowing how to use AI, particularly prompting skills, is what truly determines outcomes. Julia connected the rural/urban divide to the linguistic dimension, noting that these fractures exist both within countries and between them, and that even technically accessible AI delivers inferior results in low-resource languages.

A key tension emerged : should priority go to solving technical access (connectivity, licenses) or skills education (prompting, critical AI literacy)? For universities : -> Negotiate group licenses for "pro" versions of AI tools -> Train professors to design assignments that encourage thoughtful AI use rather than passive reliance. -> Integrate AI literacy courses or activities into curricula: learning to prompt effectively, identify biases, verify outputs, and cite AI correctly (training critical users, not dependent ones). For governments : -> Invest in rural digital infrastructure and developing the internet media, a prerequisite for any meaningful AI access. -> Fund the development of multilingual models that include low-resource languages such as Swahili, Arabic dialects, and regional languages. -> Work toward international regulatory alignment to prevent legal borders from becoming educational inequality borders. → Establish national AI education strategies with dedicated funding lines, rather than leaving adaptation entirely to individual institutions. For civil society and NGOs: → Develop open-source, low-bandwidth AI tools adapted to constrained connectivity environments → Build peer-educator networks, training students to train other students in AI literacy,