In my earlier post on “doing science in the age of AI, when the system isn’t built for you,”. I argued that AI offers both opportunity and risk for researchers working in systems that were not built with them in mind. AI may help lower barriers to knowledge, training, analysis, and participation. Since then, a response from Daniela Saderi pushed me to think more deeply about what AI cannot provide on its own: a sense of belonging.
The more I think about it, the more convinced I become that this may be one of the most important conversations in modern science: Access to tools is not the same as belonging in science.
For that to occur, researchers need mentorship, community, and human connection. AI is changing how researchers write, analyze data, search literature, code, communicate, and even generate hypotheses. While for many scientists, especially in low-resource settings, these tools offer exciting opportunities to bridge long-standing gaps in access and productivity. Beneath the optimism lies a difficult question: if science becomes increasingly automated, what happens to the human relationships that make science meaningful, ethical, and inclusive? Because despite all its technological advancements, science has never been purely computational. Science is deeply social.
Science Beyond Automation
Scientific discovery depends on mentorship, collaboration, peer discussion, and shared intellectual spaces. Researchers are not simply trained through textbooks or software tools; they are shaped through conversations with mentors, feedback from colleagues, participation in communities, and exposure to diverse perspectives. Much of scientific growth happens informally, during laboratory discussions, journal clubs, conferences, collaborative projects, and moments of collective problem-solving. AI can accelerate access to information, but it cannot fully replace the human processes that cultivate judgment, creativity, and scientific identity.
One of my growing concerns is that excessive dependence on AI tools may gradually increase intellectual isolation among researchers. A generation of scientists could become highly efficient at generating outputs while becoming less connected to collaborative learning and critical engagement. Early-career researchers, in particular, may increasingly rely on AI systems for answers while missing opportunities for mentorship and scientific dialogue that traditionally shape independent thinking. Efficiency coming at the cost of intellectual community.
This concern becomes even more important in the Global South. For many researchers in low- and middle-income countries, barriers to participating in global science extend far beyond access to information. Challenges such as limited infrastructure, funding constraints, weak institutional support, and limited mentorship opportunities continue to shape scientific participation. While AI tools may improve access to knowledge, access alone does not automatically create inclusion.
Community Matters
The more I reflect on my own journey, the more I see community not as a supplement to scientific infrastructure but as infrastructure itself. My experience with Open Life Science (OLS) illustrates this clearly. I first joined OLS as a participant in the Open Seeds program, where I gained mentorship, training, and exposure to open science practices. Those experiences did more than teach me technical skills; they connected me to a global network of researchers who invested in my growth and development.
Later, I returned to the same community as an OLS Resident Fellow, supporting participants and helping create environments where others could learn, collaborate, and thrive. Through this role, I was able to further develop Bioinformatics Outreach Nigeria (BON), a community dedicated to supporting researchers in Nigeria and across Africa through training, mentorship, and open science advocacy.
What makes this story significant is not my individual progression from community member to fellow. It is what happened next. Through BON, we have been able to decentralize opportunities that might otherwise remain inaccessible to many early-career researchers in low resource settings. For example, in a recent full cycle example, two members of our community were selected into the OLS Nebula program. Drawing on the mentorship and support I had previously received, I was able to coach and mentor them on their projects and on integrating open science practices into their research. In other words, the value generated by one community experience did not stop with a single individual; it was transferred, adapted, and multiplied through another community.
This is why I increasingly view the community as the infrastructure. Infrastructure is not only computers, laboratories, datasets, or AI systems. Infrastructure is also the networks of trust, mentorship, and collaboration that allow people to access opportunities, develop expertise, and support others in turn. Communities create pathways that technology alone cannot build.
For many researchers, especially in the Global South, scientific advancement often depends less on access to information and more on access to people: mentors who provide guidance, peers who share experiences, and communities that create opportunities for participation. AI may help researchers find answers faster, but communities help researchers become scientists.
Mentorship Matters
We all know that AI systems are not neutral. Algorithms are trained on existing literature, datasets, and patterns of knowledge production that already reflect global inequities and underrepresentation. If left unexamined, AI will amplify dominant perspectives while further marginalizing voices from underrepresented communities and regions.
This is where mentorship becomes indispensable. Mentors do far more than transfer technical knowledge. They nurture confidence, challenge assumptions, support creativity, and help early-career researchers navigate complex academic environments. More importantly, mentorship helps preserve scientific agency. As AI-generated outputs become increasingly common, researchers must remain capable of independent reasoning, skepticism, and ethical judgment. The goal should not be to reject AI, but to ensure that human values continue to shape how these technologies are used within scientific communities.
Building Infrastructure
AI will continue to transform research and scientific practice. It will accelerate workflows, expand access to information, and reshape how knowledge is produced and shared. But technological advancement alone cannot guarantee equitable or inclusive science. Human relationships remain essential.
In the age of AI, mentorship, collaboration, and scientific communities may become some of the most valuable forms of scientific infrastructure we have. Because, while AI may transform how science is done, communities and mentorship will determine who still gets to belong in science.
Copyright © 2026 Seun Olufemi. Distributed under the terms of the Creative Commons Attribution 4.0 License.
