Your call for health benchmarks that can redirect the AI race toward measurable human benefit raises an essential question: What should count as winning?
Disease detection, diagnostic accuracy, treatment effectiveness, healthy lifespan, and healthcare cost are critical benchmarks. But they do not fully capture whether people are actually becoming healthier in their daily lives—psychologically, behaviorally, relationally, and functionally.
I would like to propose an additional benchmark category for the Health Moonshots Race: The Human Adaptive Well-Being Benchmark
The purpose of this benchmark would be to evaluate how effectively AI-enabled systems help people move from distress, dysfunction, and avoidable risk toward greater well-being, adaptive functioning, agency, resilience, connection, and sustained health behavior.
This would complement disease-specific moonshots by addressing the human processes that influence nearly every health outcome, including how people interpret their circumstances, manage emotions, cope with stress, solve problems, make decisions, seek support, change behavior, and sustain improvement over time.
Why this benchmark is needed
Many of the most consequential health problems are shaped not only by biology or access to treatment, but by the interaction of:
• stressful life and work conditions;
• maladaptive interpretations and mindsets;
• emotional and physiological responses;
• coping strategies;
• social and organizational environments;
• health-related decisions and behaviors; and
• the person’s ability to translate knowledge into sustained action.
AI may provide information, recommendations, coaching, monitoring, or personalization. But unless we can measure whether those capabilities deliver meaningful improvements in people’s lives, we risk mistaking engagement, output volume, or technical sophistication for health impact.
I contend that the benchmark should therefore ask: Does the system help people function and live better, and can it demonstrate how, for whom, under what conditions, and for how long?
Thanks for writing this. I really like the idea of shifting the AI race from technical scores to real health outcomes. Competing on things families can actually feel is the right direction.
What I keep coming back to, though, is that this still sounds like a speed problem. Faster detection, cheaper trials, better diagnostics. All important. But they don’t explain why care still breaks down for so many people.
Across health systems, the deeper issue I see is drift. The same patient is assessed differently across settings. Decisions get fragmented across hand-offs. A few weeks later, no one can clearly explain why something happened.
So, alongside moonshots like early detection, I’d love to see benchmarks that track:
How often do patient journeys break across teams
How long it takes to reconstruct a clinical decision
Whether care actions are traceable to a clear intent and responsibility
Fixing drift is what changes care in real life. Speed helps. Clear, continuous judgment is what builds trust. That’s also the missing foundation of real value-based care. That’s the race that matters.
Your call for health benchmarks that can redirect the AI race toward measurable human benefit raises an essential question: What should count as winning?
Disease detection, diagnostic accuracy, treatment effectiveness, healthy lifespan, and healthcare cost are critical benchmarks. But they do not fully capture whether people are actually becoming healthier in their daily lives—psychologically, behaviorally, relationally, and functionally.
I would like to propose an additional benchmark category for the Health Moonshots Race: The Human Adaptive Well-Being Benchmark
The purpose of this benchmark would be to evaluate how effectively AI-enabled systems help people move from distress, dysfunction, and avoidable risk toward greater well-being, adaptive functioning, agency, resilience, connection, and sustained health behavior.
This would complement disease-specific moonshots by addressing the human processes that influence nearly every health outcome, including how people interpret their circumstances, manage emotions, cope with stress, solve problems, make decisions, seek support, change behavior, and sustain improvement over time.
Why this benchmark is needed
Many of the most consequential health problems are shaped not only by biology or access to treatment, but by the interaction of:
• stressful life and work conditions;
• maladaptive interpretations and mindsets;
• emotional and physiological responses;
• coping strategies;
• social and organizational environments;
• health-related decisions and behaviors; and
• the person’s ability to translate knowledge into sustained action.
AI may provide information, recommendations, coaching, monitoring, or personalization. But unless we can measure whether those capabilities deliver meaningful improvements in people’s lives, we risk mistaking engagement, output volume, or technical sophistication for health impact.
I contend that the benchmark should therefore ask: Does the system help people function and live better, and can it demonstrate how, for whom, under what conditions, and for how long?
Found this super interesting, thanks for writing this. Nuanced thoughts
Thanks for writing this. I really like the idea of shifting the AI race from technical scores to real health outcomes. Competing on things families can actually feel is the right direction.
What I keep coming back to, though, is that this still sounds like a speed problem. Faster detection, cheaper trials, better diagnostics. All important. But they don’t explain why care still breaks down for so many people.
Across health systems, the deeper issue I see is drift. The same patient is assessed differently across settings. Decisions get fragmented across hand-offs. A few weeks later, no one can clearly explain why something happened.
So, alongside moonshots like early detection, I’d love to see benchmarks that track:
How often do patient journeys break across teams
How long it takes to reconstruct a clinical decision
Whether care actions are traceable to a clear intent and responsibility
Fixing drift is what changes care in real life. Speed helps. Clear, continuous judgment is what builds trust. That’s also the missing foundation of real value-based care. That’s the race that matters.