Results of LW Technical Background Survey
The main goal of the survey was to provide info for authors about their target audience, so here’s a high-level overview toward that end:
The average respondent is some kind of professional programmer, with an undergrad degree (or equivalent) in CS.
Most people have seen at least some economics and probability, but not at the level of a undergrad degree.
Almost everyone knows calculus, but linear algebra or differential equations will likely be lost on at least ~25% of respondents.
There are substantial zero-knowledge and high-knowledge counts for most areas.
About half of respondents had read the sequences in their entirety.
Here are charts of the responses to each question. I strongly recommend looking at them directly rather than just taking my summary at face value. As always, remember this is an opt-in survey without any sort of verification of responses, so take everything with a grain of salt.
One interesting note: we had a handful of respondents declaring very high skill levels (Nobel-level economists, Turing-level computer scientists, or primary developers of popular software). I’d personally be interested to hear what exactly those people work on, especially if they’re willing to occasionally field questions on their area of expertise. All y’all should leave a comment or something.
Actually, I’m curious what everyone works on, especially specialties for all the researchers. Feel free to leave a quick comment, especially if you’re able and willing to occasionally field questions in your area of expertise.
Thanks for doing this!
This looks like the median.
It’s taking the median across two different axes independently, then sticking the results together. In principle, if we measure x and y values in a population, there may not actually be anybody in the population with median x value and median y value. Point is, the concept of “median” doesn’t neatly generalize to multiple dimensions.
So I sneakily swept all that under the rug and fudged it by saying “average”.
I wonder what the downside of assuming it is the median, instead of the median in each group, for the purpose of writing posts is. And if there’s a convenient statistical measure of that.
ETA: If one assumed that, we’d figure that our posts would be read by professional programmer, who is a technical undergrad, who majored in CS, and took at least a course in economics, and probability, and read the sequences. If we assumed no correlation and treated the ratios as probabilities, then multiplied them together, the chance of a reader being exactly that would be 2.4%. The chance of them having at least that knowledge would be about 12%.* I was asking whether it’s actually higher or lower than that based on the survey data.
*(.504+.252)(1-.236-.195)(1-.089-.297)(1-.134)(.537+.272+.138)(.537)=0.11631737422