Monday, September 01, 2014

Notes on Seth Roberts Memorial Symposium

I had a long-planned family reunion in Maine the week of the Ancestral Health Symposium in Berkeley this month, so it was just not possible for me to attend the Seth Roberts Memorial Symposium, but thanks to Tess McEnulty there are videos of the public talks. The talks are well worth watching, but if you don’t have the time, here is my brief summary of the highlights:

Nassim Taleb explained how Seth’s philosophy (n=1) is the exact opposite of what you see in today’s fascination with Big Data. No matter how many data points you accumulate, a new theory can be disproven with a single counter example; and sometimes you can build a true theory based on a single example. (“OJ Simpson only killed once; does that mean you can’t prove he’s a murderer?”). In fact, the more variables you add to a model, the more likely you are to find spurious correlations, as he shows in this slide:Nassim Taleb: Tragedy of Big Data

Tim Ferris (Four Hour Work Week) credited much of his book The Four Hour Body to ideas he got from Seth, who taught him five things:
  • Extremes inform the means. New products and ideas rarely come from “normal” use cases. If you want to find something interesting, search for odd examples.
  • Choose fast results over big data. Look for quick-and-dirty experiments, not big-huge-complicated ones.
  • Track yourself regularly: don’t try to judge a soccer match from a single ultra-hi res photo; it’s much better to have multiple, low-res photos, so track what you can however you can. Seth tracked most stuff with pencil and paper.
  • Remember the “Minimum Effective Dose”: for example, he gets fantastic sleep by taking raw honey and a single tablespoon of apple cider vinegar before bed. No more.
  • Care about normal people: Seth didn’t care who you are or your background. You can learn something from anybody.
Gary Taubes says Seth was one of the only people he ever talked with (they hiked regularly in Berkeley). His talk started with an overview from 19th century doctor Claude Bernard, whose 1865 book Intro to the Study of Experimental Medicine "should be required reading for every med student”.  Key quote: "All human knowledge is limited to working back from observed effects to their cause”. The rest of the talk was a summary of the limitations of various approaches to scientific research (observational studies, randomized control trials, etc.). There are no really good solutions, other than the open-minded and humble approach of people like Seth.

There were several other speakers, like best-selling “fratire" author Tucker Max (who met Seth by randomly emailing him), Paleo author John Durant (who appreciates Seth’s example that you don’t need fancy equipment to do science), experimental psychologist Aaron Blaisdell (who founded the health crowdsourcing site thanks to collaborations with Seth), and many others.

So many great memories of Seth’s ideas, by people who knew him well. I wish I could have attended in person.