The Arrow Impossibility Theorem

What Hacker News says

1. somberiFeb 2017
   For someone interested in a contemporary view of his "Impossibility Theorem", especially in policy planning, I recommend:

   https://www.amazon.com/Arrow-Impossibility-Theorem-Kenneth-L...

   thread
2. PaulHouleJun 2020
   OKRs leave a terrible taste in my mouth from experience.

   I saw them used at a struggling startup, where it seemed we had just one real goal (e.g. "find product-market fit") rather than a large number of objectives. We were already zig-zagging a lot, and we just didn't need the distraction of creating and maintaining OKRs, never mind yet another tool to create confusion about what was essential and what was not.

   My first thought when I heard about it was the book "The Goal" where the protagonist's friend immediately says that a business in crisis has one goal: quantitative methods are important to find and remove the bottlenecks for that goal. (In that a startup hasn't established a profitable business at first, it shares quite a bit with a failing car factory or restaurant.)

   Going back to Pareto, social scientists have established that you really can't find a total ordering for the world's utility function. A world where everything else is the same and I am $1 richer is a better world (partial ordering), but you can't compare that objectively to a world where you got $1 and I didn't.

   This connects to a group of results (such as the Arrow Impossibility Theorem) that show that social science is limited in the way that mathematics is limited by Godel's theorem, the halting problem, etc. There is no voting system, economic system, etc. which provably produces "just" or "correct" results and in institution design we are stuck with limiting the damage that can be done by bad people.

   A world where I hit 10 out of 10 OKRs is better than any in which I fail OKRs, in the real world that I hit 5 out of 10 it is hard to compare that to other 5 out of 10 outcomes that I could have had if I had made different choices.

   There are some jobs that you can break down into pieces, say a supermarket cashier, the store has a model of what makes a good cashier, they have control charts at the front or in the break room for each of them on two metrics, one of which is the scan time (tied to the profitability of the business), another is the % of times you hit the "No Sale" button to open the cash drawer between customers, stop the scan clock, and cheat on the first metric.

   That's profitable thinking because somebody thought it up a long time ago and the store manager knows it and the checkout manager who trained you knows it, and I thought it was fun when as a teen I could try but never beat certain middle-aged women who reminded me of my mother.

   To expect every person to make a quantitative model of their work and have their manager review it, etc. is ridiculous. Many people don't think that way but do great at their jobs. People who have PhD's in quantitative thinking can also go into the weeds (e.g. either their first response is what I gave you above, they will pick one thing to maximize, present a Pareto Frontier of choices that is somebody elses problem, or just will go in circles until they get lucky, or learn the first answer the hard way.)

   If they can pass as a neurotypical they'll get control of their systematizing urge and bullshit it the same way the sales person and art director do and get back to doing work done.

   Show morethread