Also, this is a problem about people and social norms as well as the limits of memory. It just happens to also be about math.
@evan I would say option 3 should be 3000, not 3400??
@virtuous_sloth @evan No! It's precisely right to not offer 3000. Obviously, 3000 would be the superior answer, but it would conflate the two competing principles that are tested here: Do we accept the order of magnitude as a correct answer or not, precisely when it does not happen to coincidentally also give 1 correct digit?
@paulbalduf @evan
Are you treating order of magnitude in base 10 or base 2?
Hah! I chose b) 2000 exactly because it was within an order of magnitude, but farthest off frim 3247.
@modulux 13000 is closer to 12345 than 10000 is.
@evan
I'd be ok with saying "a bit over 3000"
@TCMuffin I don't understand the question. Are you asking what I would have guessed the majority's answer was before I started the poll?
@TCMuffin I didn't have a good idea, which is why I made the poll. I can't unsee these results, so that probably clouds my judgment, but I think the last two options were the most probable, maybe the second for some very forgiving people.
Hey, all. So, I asked a question about podcasts recently. One of the things that makes me most nervous about doing live interviews is having to give numbers off the top of my head. How many people are active on the Fediverse? How many servers are there? What is their median size? How many codebases support the Fediverse? What percentage of servers have gone offline? How many servers make up 80% of the user accounts? 99%? That kind of thing.
I hold myself to what I think is a pretty fair standard -- I think being able to recall numbers to an order of magnitude and one significant digit is probably OK. So, if I say, "There are about 40,000 servers on the Fediverse", and someone says, "Actually, there are 43,000," well, I don't feel bad about it.
A lot of respondents said that expressing two significant figures if the second one was off was worse, even if the value is close to the correct value. I think that's probably true for scientific papers but less for casual discussions or press interviews.
And, of course, these values are conditional on the situation. I don't think having the OOM and first significant figure is sufficient for your partner's age, but it's probably fine for your aunt's or neighbour's. Knowing numbers to 2 or 3 sigfigs is a sign of care. Maybe knowing your KPIs at an organisation is similar.
And scale matters too. I'm a lot more forgiving about the difference between 4e-15 and 4e-16 than I am about 4e2 and 4e3. Not very logical, but that's how numbers work for me.
Anyway, I think I'm good with, in typical situations, same OOM and 1 sigfig the same.
These are:
- off by an order of magnitude
- same order of magnitude
- same order of magnitude, 1 significant digit
- same order of magnitude, 2 significant digits