What is your calcium?

Somebody introduced me to a song a long time ago. Seven minutes of a calm voice telling a graduating class how their lives are going to go. Wear sunscreen, floss, stretch, do one thing every day that scares you. If you are roughly my age (or one of those poor souls I shared the song with) you can probably still recite almost all of it.

Recently I wanted to know which class of 1999 was lucky to get this speech. Turns out, none. Kinda sad.

It was never a speech at all

It is a newspaper column. Mary Schmich, the Chicago Tribune, 1 June 1997, headlined "Advice, like youth, probably just wasted on the young". Within weeks it was circling the internet as Kurt Vonnegut's commencement address at MIT. He had never been MIT's commencement speaker. Schmich called him up and wrote the whole business up that August. What he told her was "it was very witty, but it wasn't my wittiness".

MIT's real speaker that June was Kofi Annan, who, as she pointed out, did not mention sunscreen. Also sad!

So my favorite graduation speech was written by a columnist, published under the wrong name, delivered by email. The gown and the podium were added later by strangers. So 'fake news', a term I could live without and do not remember hearing back then. It is the same move I wrote about with the founder stories we file under inspiration, nobody used that word at the time and the ending got attached once it paid, talking about the column not the 'fake news' term - the latter is just stupid.

She vouched for one line

Well at least she claimed scientific evidence.

"If I could offer you only one tip for the future, sunscreen would be it. The long-term benefits of sunscreen have been proved by scientists, whereas the rest of my advice has no basis more reliable than my own meandering experience."

So why not use science to ruin a perfectly good piece of art? Let's do it! (Not actually going to ruin it, this song got me through a lot of dark times, but let's have a little bit of fun with it).

The one she vouched for held. In 1992, five years before the column, 1,621 people in Nambour, Queensland were randomly assigned to daily sunscreen or to using it when they felt like it. Ten years after the trial finished, the daily group had 3 invasive melanomas against 11 in the other group. A separate arm measured 24% less skin ageing over four and a half years.

With one little problem. In June 1997 it was not actually proved. The first randomized skin cancer result arrived in August 1999 and the melanoma result took until December 2010. She was right about thirteen and a half years before anyone could show she was right. I told you the lyrics are visionary. Also, anybody that says AI is useless, should tell me how long it would have taken them to research and cross check these studies. For me it took a few prompts and verifications: 10 min. tops. I challenge you to beat that with traditional research methods.

Now let's look at the rest of the claims

Stretch. A meta-analysis of 25 randomized trials, 26,610 people, 3,464 injuries. Stretching scored RR 0.963, which is nothing. Strength training cut injuries to under a third. In that entire analysis, stretching is the only thing that did not work, and it is the line everyone can repeat, and entire sections in gyms are dedicated to it. :)

Get plenty of calcium. Across 59 randomized trials, more calcium moved bone density by 0.6% to 1.8%, which the authors call "unlikely to lead to a clinically significant reduction in risk of fracture". In the four trials at lowest risk of bias, 44,505 people, no effect on fracture at any site. The money I wasted on calcium pills - and yet - I keep doing it. Try to stop me.

Floss. Cochrane called the plaque evidence "weak, very unreliable" in 2011, and the 2019 replacement rates it low certainty and says the effects may not be clinically important. The word "floss" appears once in the 2010 US Dietary Guidelines and zero times in the 2015-2020 edition. That comes from searching the documents themselves rather than the reporting about them. The 2010 hit proves the search works, so the zero is real. Again, beat that without AI agent support, and mine often have names.

BUT, not everything she disclaimed by her own words fell over. Be kind to your knees turns out close to literal: in one long-running US cohort, the lifetime risk of symptomatic knee osteoarthritis was 44.7%, but near enough a coin flip. Mine are still fine, but kind to them I was not.

So the scoreboard is upside down. The single claim she staked her credibility on survived 29 years of trials. Two of the ones she explicitly disowned have been dismantled. The most accurate sentence is the one nobody even remembers, knees are knees.

This wouldn't be a DETGAAO blog if there wasn't an angle, come on you didn't think I am going to dig on one of my favorite songs if there wasn't a good reason?

Every industry has a stretching

Maybe that is the thing to take away here.

Stretching did not spread because it worked. It spread because it was easy to say, easy to picture, and everybody was already doing it. It got repeated until repetition looked like evidence. Then somebody eventually tested it, and the counting arrived twenty years after the habit. (Get it? Testing is important, not just two decades after spending the money or time or effort.)

Every field I have worked in has one of those. In marketing it is the channel everybody buys because everybody buys it, or your VP or CEO saw a Twitter post about it. (... cough hyper segmentation cough... ) Right now, in AI, it is most of what gets said at conferences. I spend a lot of my time at DETGAAO doing exactly this job for companies: taking the things they have been told about AI and sorting the ones with a trial behind them from the ones that are somebody's meandering experience, sincerely offered and sounding wise.

The tell is not confidence. Schmich was confident and she was right. The tell is whether anybody ever went back and tested, or evaluated the environment, which is the same reason I built a tool to count how many rules I had to give twice. Nobody counts the unflattering number unless they decide to. Or I largely stopped giving my AI gates and turned them into flexible rules to be re-evaluated at frequent turns.

So, what is your calcium?

I am not saying 'doubt everything'. It is to pick the process or practice you would defend hardest in a meeting, the one you would be slightly annoyed to have questioned, and go and look, test it.

Two things happen. Usually it holds, and you find out the evidence under it was much much thinner than anyone repeating it knew, which is worth knowing. But sometimes it's the calcium or stretching. Then the decision is, what wins inertia and pride or the result in front of you? I worked for a women's clothing retailer once (there are a few so good luck guessing). Presented sales numbers in an exec meeting that clearly showed that the new merch and pricing strategy hurt conversions and led to customer churn. The comment from the most Sr. leader in that room was: "I know that's what the numbers say, but I don't believe it" ... I can still hear it and it has been years. It was the 'sign' I needed to find new employment... what happened to the retailer? You guess. ... so ... really 'doubt everything' :)

Either way you now know which one you are standing on. If "I know that's what the numbers say, but I don't believe it" is going to be your answer forget you read this post, hey, the song is still right about the sunscreen.

To be serious and clear about this, this song is amazing, and no it isn't about whether the facts all check out, because its main point absolutely holds... it doesn't matter when you look back from, things were not as bad as they seemed in that moment, and you were NOT as fat as you imagined.