This is an unusually clear and serious piece. The way you trace the lineage of the “Average Patient” — from Lind through Kefauver-Harris to the modern regulatory fortress — gets at something many of us sense but rarely articulate this precisely: that the statistical ghost was a necessary compromise, not an error, and that biology has now outgrown it.
I was especially struck by your insistence that the current regulatory signals are seeds, not solutions. Treating cases like Baby K.J. as miracles rather than signals would be the real failure of imagination. The challenge, as you frame it, isn’t whether we can act on the particular, but whether we can build systems that learn responsibly from the N of 1 without abandoning rigor or trust.
This feels like a departure point rather than a conclusion — and I hope the conversation around it widens.
This is so informative and well written. Thank you. I’m dealing with this on a very personal level in cancer treatment It was eye opening for me to learn that the drug trials excluded people like me.
Regarding the topic of the article, what a super insightful read. That whole idea of the 'Average Patient' as a ghost that both saves and erases is so well articulated. It really makes you think about the biases we've baked into our models. I mean, AI today is all about pushing for more personalized, individual data points, right? Feels like we're finally grapling with that initial sacrifice.
Great piece Sean the road begins.
This is an unusually clear and serious piece. The way you trace the lineage of the “Average Patient” — from Lind through Kefauver-Harris to the modern regulatory fortress — gets at something many of us sense but rarely articulate this precisely: that the statistical ghost was a necessary compromise, not an error, and that biology has now outgrown it.
I was especially struck by your insistence that the current regulatory signals are seeds, not solutions. Treating cases like Baby K.J. as miracles rather than signals would be the real failure of imagination. The challenge, as you frame it, isn’t whether we can act on the particular, but whether we can build systems that learn responsibly from the N of 1 without abandoning rigor or trust.
This feels like a departure point rather than a conclusion — and I hope the conversation around it widens.
This is so informative and well written. Thank you. I’m dealing with this on a very personal level in cancer treatment It was eye opening for me to learn that the drug trials excluded people like me.
I wrote about it if you’re interested.
https://kaliakali.substack.com/p/dear-doc-please-dont-write-refuses?r=5dfy&utm_medium=ios
Regarding the topic of the article, what a super insightful read. That whole idea of the 'Average Patient' as a ghost that both saves and erases is so well articulated. It really makes you think about the biases we've baked into our models. I mean, AI today is all about pushing for more personalized, individual data points, right? Feels like we're finally grapling with that initial sacrifice.