[diego898]

Purely empirical observation, in my own life, make no claim as to humanity/society/etc.:

It's interesting how often fermi estimation problems are used as proxy's for "intelligence". Something like: 'let's assess how well "they can think" - how many golf balls fit in a baseball stadium?' etc.

Often, doing well in these kinds of problems can more than makeup for a lack of specific knowledge in something someone is interested in assessing!

[miobrien]

This reminds me of a question from my first interview as a college grad: estimate the number of taxis in New York City. I was totally baffled by it.

[singleshot_]

I’ll simplify for manhattan and extrapolate for the four outer boroughs. Ten avenues, a hundred streets. A thousand blocks? One cab per block? One thousand cabs in manhattan? 5,000 total?

There are about 13,500 taxi medallions.

[saghm]

That sort of estimation feels a lot easier to me than the "golf balls in a baseball stadium one" that was mentioned by the parent because it's dealing with quantities I can recall having heard before like "how many streets are there in Manhattan" rather than measurements that personally would never stick in my head like "how wide is a golf ball". I'm not sure why, but I've always been awful at making even rough estimates of units. If you gave me the diameter of a golf ball and the dimensions of the stadium, I could do some basic calculations, but even though I physically know about how large a golf ball is, I couldn't tell you whether it's more likely that its diameter is 0.5 or 1.5" (and not having looked it up, I would believe you if you told me it wasn't even within that range)! This gets worse with units I can't visualize (like weight), and when the sizes get larger than I can easy relate to; if you asked me questions like how much a car weighs or how long the Brooklyn Bridge is, I'm doubtful I'd even be within a factor of 2 more often than not.

I'm probably taking this more seriously than it was intended above, but the idea that this is some sort of proxy for "thinking" or "intelligence" feels off to me; doing the math given the size of something might be thinking or intelligence, but knowing roughly "how big" something is seems more like intuition.

[stavros]

I kind of figure a centimeter is about the width of a pinky, then I try to gauge how many pinkies fit in some distance, and go by that. I'd imagine a golf ball is about 4cm in diameter, though I haven't seen one in years.

[saghm]

Yeah, I don't feel confident in the idea that my pinky is around 1/4 of a golf ball. It could be 1/6, or 1/2, or nowhere close to either. A lot of this seems to be an exercise in how confident someone is about arbitrary guesses, and it seems weird to me that a higher willingness to make assumptions is somehow correlated with raw intellect. If someone wants me to do the math with a completely made up number, I can do that, but at that point it seems like the true test is figuring out whether the person asking the question actually cares about the accuracy of the answer or not, and that seems more about the social aspect of an interview. That isn't to say that what it's measuring isn't useful, but I think as someone on the spectrum, it's hard for me not to have a strong reaction to the idea that it's purely a measure of intelligence.

[stavros]

I get it, but it looks like a golf ball is 4.3cm, so I was pretty close for it to be that arbitrary.

[darkerside]

Maybe take something small you do know the size of and estimate how many golf balls fit into that

[Breza]

I had a great boss who really liked that kind of question. I disagreed with him. I would rather have someone who knows how to find the official answer online and verify the quality of the source.

[kragen]

A crucial tool for verifying the quality of the source is noticing when the answers given by the source are clearly wrong by orders of magnitude.

[toss1]

THIS!!

The ability to estimate within an order of magnitude or within 2X is vastly more valuable, and beyond being able to have a sense of whether the "official" answer is likely accurate or off by orders of magnitude.

During most of the process of designing anything in or that touches the physical world, you are using rough figures.

Taking time to get the fully accurate and precise answer for every question is a waste of time as you don't need that many decimals of precision to move forward. Every decimal of precision in the answer takes more time and there are MANY of those questions, so being 100% accurate in every answer does not scale.

Of course, when it gets to the end of the process, the accuracy & precision requirements increase, but the emphasis needs to be placed where needed, not everywhere.

Plus, you are not going to find the "official" and accurate number of golf balls in the particular school bus you want to model. You'll find some vaguely similar answer or set of sub-answers, so sure, those will be fully accurate and precise, but THEN you must take those as inputs for your estimate, and we're back to the skill of estimating being most critical.

Being able to estimate and do sound back-of-the-envelope calculations is the far more critical skill, at least on any team I'm building.

[nocoiner]

Ideally I think the ability to come up with a quick off-the-cuff guess that is correct to one order of magnitude and then find and verify the specific answer are nicely complementary.