Tom Osborn: AI should support, not replace, human mental healthcare in Africa

Tom Osborn, a Kenyan social entrepreneur and Forbes 30 under 30 recipient, argues that artificial intelligence (AI) should not replace but rather support human mental healthcare services in Africa. Speaking at the One Young World Summit in Cape Town, Osborn, who founded the Shamiri Institute in Nairobi, contends that AI tools should be used as supplementary aids to counsellors, psychologists, and psychiatrists rather than stand-alone mental health providers.

Osborn emphasizes that AI should not be used without rigorous evidence and safeguards. His Shamiri AI platform is designed to train lay providers, monitor clinical quality, and support care referrals, rather than operating as an automated therapist. He stresses that before AI can be used as a direct-to-consumer mental health agent, it must meet the same standards as digital medical devices, which requires robust scientific inquiry.

Osborn's interest in mental health care was sparked by a personal family crisis. At 14, he earned a scholarship to a boarding school in Nairobi, but during this time, his mother developed a severe respiratory infection from cooking with firewood. This experience led him to study energy and economics, intending to pursue clean-tech ambitions. However, a second-year psychology class at Harvard University inspired him to shift his focus to clinical psychology and mental health interventions.

Osborn criticizes the global reliance on Western diagnostic frameworks like the DSM, arguing that these systems are flawed and do not accurately represent mental health in diverse cultural contexts. He highlights how the violent history of psychiatric care under African settler colonialism contributes to stigma around psychological treatment. To address this, Shamiri AI focuses on developing tools that meet people where they are, rather than relying solely on clinical terminology.

The AI platform is utilized in several key areas: automated triage to identify high-risk cases and refer them to qualified practitioners; developing or assessing AI systems in conversation with mental health professionals; and conducting research on the relationship between therapeutic intervention and provider-client interactions.

Osborn stresses that Shamiri limits data used for machine-learning research to controlled scientific studies conducted in randomised trials, ensuring privacy and ethical standards are maintained. All data is anonymized and stored in Kenya, with oversight from local health authorities and an ethics review board.