Science is often described as a universal enterprise, but for many researchers in low- and middle-income countries (LMICs), especially across Africa, it does not feel universal at all. It feels fragmented, under-resourced, and difficult to enter on equal terms. The barriers are not only financial or technical. They are built into the infrastructure, incentives, and systems that shape who gets to participate fully in research and who is left working around the edges. As artificial intelligence (AI) begins to reshape scientific work, it offers both a new opportunity and a new test: whether emerging tools will help repair those inequities or deepen them.
Artificial intelligence offers a powerful opportunity to mend this fragmentation. It can help researchers work around barriers that have long limited participation. But that promise is not guaranteed. AI is still shaped by the same inequalities embedded in the broader scientific system, and without more deliberate investment in access, infrastructure, and inclusion, it may end up reinforcing the very gaps it seems to address.
As part of my work with the Bioinformatics Outreach Nigeria (BON), a community-driven effort to build capacity in bioinformatics and open science, we conducted a survey of 212 respondents in our community. This survey offers insights into a few of the systemic issues facing LMIC-based scientists and how AI affects these dynamics.
The everyday barriers of scientific work
For researchers in LMICs, scientific work and access to learning opportunities often look very different from what researchers in well-resourced countries experience. In our recent survey, one key baseline insight was the overall observation that LMIC-based bioinformatics researchers perceive their work as difficult, with 50% describing it as very difficult and another 33.5% as somewhat difficult. This is not difficult to understand as modern bioinformatics research relies on many factors that can be out of reach of many LMIC-based scientists: reliable internet, computational power, and stable electricity, etc. Slow connections turn simple tasks into ordeals, while power outages disrupt workflows entirely, as this makes research learning processes almost impossible, and research processes very daunting.
Just as important, even when infrastructure hurdles are cleared, knowledge itself stays locked behind paywalls. Subscriptions to major journals are often unaffordable, and open access publishing frequently brings high article processing charges that institutions cannot cover. This creates a paradox: researchers are expected to contribute to global science while being largely excluded from its latest developments. The result is skewed research agendas that prioritize high-income contexts, reduced visibility for local knowledge, and a persistent cycle where LMIC scientists remain primarily consumers rather than producers.
Beyond the difficulty of conducting research, access to training and mentorship forms another major bottleneck. The same survey showed strong demand for practical support, with the majority of respondents prioritizing workshops, requesting one-on-one training, and seeking online courses or tutorials. These types of structured opportunities for advanced skills are scarce in LMICs, and many online resources assume high-bandwidth access or prior expertise that beginners may lack. It further highlighted lack of training or knowledge (24.5%) and lack of access to resources or equipment (23.1%) as the top challenges when using bioinformatics tools. Without localized guidance, aspiring scientists often navigate complex fields in isolation. Community initiatives like BON address this by offering workshops, hands-on sessions, and mentorship tailored to real constraints. These efforts show how quickly capacity can grow when barriers drop. Yet they typically run on volunteer energy and limited funding, operating at the margins rather than as part of sustained national systems.
The survey also pointed to a deeper structural problem: the incentives surrounding scientific work are often poorly aligned with openness, collaboration, and capacity building. Systems reward high-impact publications but do little to recognize openness, data sharing, collaboration, or capacity building. Concerns about misuse or lack of credit further discourage open practices, even among motivated researchers. Together, infrastructure gaps, knowledge barriers, training shortages, and poor incentives reinforce one another, producing an ecosystem where global science participation stays uneven.
AI as Opportunity and Risk
Artificial intelligence offers a powerful opportunity to ease some of this fragmentation. AI can democratize knowledge by summarizing complex research papers, translating content across languages, and adapting formats for easier understanding, helping researchers in LMICs stay current with global developments when access is uneven. In resource-scarce environments like those highlighted in our survey, AI can act as a personalized tutor, providing real-time guidance through coding challenges, analytical methods, and bioinformatics workflows.
More specifically, what makes AI especially important in this context is that it maps so clearly onto the challenges our survey identified. When research is difficult because infrastructure is weak and access is uneven, AI can help lower some of the friction around discovery, analysis, and technical work. When training and mentorship are hard to access, it can offer a form of immediate, practical support. And when scientific incentives do little to reward openness or collaboration, AI can at least make some of the labor behind documentation, metadata, and reproducibility easier to manage. It cannot fix these deeper structural problems by itself, but it can help relieve some of the pressure they place on researchers. Yet these benefits are not automatic. AI systems trained predominantly on data from high-income countries risk embedding biases that perform poorly in LMIC contexts, potentially widening rather than closing existing gaps. Uneven access to advanced tools could deepen inequalities, and concentrated control by a few institutions or companies might centralize power in new ways. The real challenge lies in ensuring AI is deployed inclusively, with models developed or fine-tuned using diverse, locally relevant data.
Rebuilding science as a truly global enterprise will require more than technological fixes. It will require deliberate investment in reliable infrastructure, equitable models for knowledge access, institutionalized training programs, and incentives that reward openness and collaboration. Grassroots efforts like BON show that focus on meaningful change can begin locally through targeted mentorship and community programs. Scaling these efforts will require sustained funding, policy integration, and institutional support.
The fragmentation of science in LMICs is not inevitable. It reflects deliberate choices about infrastructure, access, training, and reward systems that have left too many LMIC-based researchers working at the margins. AI will not solve these problems on its own, but will create a real opportunity to ease some of the burdens they impose. The question is whether these tools will be developed and deployed in ways that make science more equitable, or whether they will simply make an already uneven system more efficient for those who are already well served by it.
Copyright © 2026 Seun Olufemi. Distributed under the terms of the Creative Commons Attribution 4.0 License.
