Can artificial intelligence reduce learning poverty?

Learning poverty remains one of the most urgent education challenges of the digital age, limiting millions of students' ability to read, write, and fully participate in modern knowledge economies. While artificial intelligence (AI) is rapidly transforming classrooms worldwide, a central question persists: can AI meaningfully reduce literacy gaps, or is it simply another layer of technology added to fragile education systems?

In the study Leveraging AI to Mitigate Learning Poverty in the Digital Era: The Impacts of Integrated AI Educational Tools on Students' Literacy Skills, published in the journal AI, researchers investigate whether structured integration of AI-powered educational tools can significantly improve literacy outcomes, using a university case study to examine how digital intervention may combat persistent learning poverty.

Learning poverty extends into university classrooms

Learning poverty is commonly defined as the inability to read and understand age-appropriate text. While often discussed in relation to primary education, the study highlights how literacy gaps persist into higher education. In many developing countries, large numbers of students enter university without mastering academic reading comprehension, grammar, vocabulary, and structured writing.

The problem is particularly acute in systems where English is the medium of instruction but not the first language for most students. Weak English as a Foreign Language literacy skills limit students' ability to analyze complex academic materials, construct coherent arguments, and engage in research writing. These limitations affect performance across disciplines, including engineering, social sciences, and health sciences.

The study identifies structural causes common to many developing education systems. These include shortages of qualified teachers, outdated instructional materials, uneven curriculum reform, limited access to digital infrastructure, and socio-economic inequality. In addition, rapid expansion of higher education enrollment has not always been matched by proportional investment in literacy support programs.

The Ethiopian case, used as an example in the study, reflects many of these systemic pressures. National data have shown persistent literacy deficits at earlier stages of schooling, and those gaps often follow students into university classrooms. However, the authors argue that this pattern is not unique to one country. Similar literacy bottlenecks appear across parts of Sub-Saharan Africa, South Asia, and other emerging education markets.

Testing AI integration in higher education

To evaluate whether AI tools could mitigate literacy gaps, the researchers designed a quasi-experimental time series study involving 46 second-year university students enrolled in a communicative English course. While the participants were based at an Ethiopian university, the instructional model tested in the study is transferable to other higher education contexts.

Over a three-month period, students engaged in reading and writing lessons supported by three AI-powered tools: NoRedInk, Rewordify, and LanguageTool. Each tool targeted different dimensions of literacy development.

NoRedInk provided adaptive grammar exercises and personalized feedback on mechanics and usage. Rewordify simplified complex texts to support comprehension and vocabulary expansion. LanguageTool assisted students in revising essays by identifying issues in grammar, punctuation, sentence structure, and coherence.

The intervention was embedded within regular coursework rather than delivered as a separate technological experiment. Students completed three baseline literacy assessments before the AI integration and three follow-up assessments afterward. Quantitative data were analyzed using repeated measures statistical techniques, while qualitative insights were collected through focus group discussions and reflective journals.

The results revealed substantial improvements in literacy performance following the intervention. Pre-intervention scores remained stable across baseline measurements. After AI integration, mean scores increased significantly, with statistical analysis showing a very large effect size. Differences between pre-intervention and post-intervention phases were far greater than minor variations observed during the baseline period.

The researchers emphasize that the improvements reflected overall literacy performance, not just isolated grammar corrections. Gains were observed in grammar accuracy, vocabulary development, reading comprehension, idea generation, content organization, and writing style.

Students report strong gains and increased engagement

Qualitative findings reinforced the statistical results. Students reported that before the intervention they struggled with complex academic texts and often found it difficult to organize essays coherently. After sustained use of the AI tools, many described improved comprehension, clearer expression of ideas, and greater confidence in revising their work.

Participants highlighted how simplified texts helped them engage with challenging academic content without losing meaning. Personalized grammar exercises allowed them to address recurring weaknesses. Automated revision support enabled them to refine sentence structure and overall coherence.

Importantly, students did not describe AI as replacing instructors. Instead, they viewed the tools as supplements that provided immediate feedback and additional practice opportunities. This structured integration aligns with Constructivist Learning Theory, which underpins the study's framework. The theory emphasizes active engagement, scaffolding, and iterative learning processes.

The authors note that some previous research has raised concerns about over-reliance on AI tools potentially reducing critical thinking. However, in this case, AI integration was carefully embedded within guided instruction. Students remained responsible for drafting, revising, and refining their own work, with AI serving as a feedback mechanism rather than an automated writing solution.

Students also perceived the literacy gains as transferable beyond the classroom. Improved reading and writing skills were viewed as essential for research assignments, professional communication, and long-term career development.

Implications for education systems beyond one country

Many developing nations share similar challenges: English-medium instruction combined with limited literacy foundations, expanding university enrollment, and uneven digital infrastructure.

The research suggests that targeted AI integration can be a cost-effective supplement to traditional literacy instruction when implemented strategically. Rather than viewing AI as a disruptive replacement for teachers, institutions can treat it as a structured enhancement tool.

The authors recommend embedding AI-supported activities directly into language curricula, providing training for instructors, and raising awareness among students about responsible AI use. They also emphasize the need for institutional support, including reliable internet access and digital literacy development.

The study also acknowledges limitations. The sample size was relatively small, and the intervention lasted only three months. Participants were university students with some degree of digital familiarity, which may not reflect conditions in rural or resource-constrained institutions. As a result, caution is needed when generalizing results across all contexts.

Future research is recommended to test AI integration across larger and more diverse populations, extend intervention durations, and explore impacts on other language skills such as speaking and listening.