No, I don't make errors. The nail polish is literally trying - making a strenuous effort to succeed against insurmountable odds, with the goal of inspiring the wearer to tackle challenges in their own life with the same fortitude. "Do, or do not -- there is no try."

Well how do you think they're selected? Even if some subset of tweets show up in "top tweets" without having been tweeted by high-profile users or being retweeted many times, that doesn't mean the tweets got selected on the basis of appealing to the highest percentage of users. Maybe there's some randomness in the process and those users just got lucky.

One benefit of random-sample-voting is that if Twitter did use it, they could tell us. When you use random-sample-voting as your algorithm, you can be completely transparent about it, and it's still not possible for someone to game the system. The only way to "win" is to create something that will appeal to the highest percentage of people.

I like cheese.

Except my recommendation is not to push tweets to the top that are the most favorited or most recommended, because this favors people who (a) game the system by having friends or sockpuppets like or retweet their posts, or who (b) simply have lots of organic followers, but their posts might not be the most interesting in and of themselves. (If a person with 1,000,000 follows posts something that 50% of users would consider insightful, and another person with 1,000 followers posts something that 80% of users would consider insightful, then I'd rather see the latter post in my feed.) That's why I suggested the random-sample rating system instead.

So, you quoted something that I wrote, but you interpreted it as the opposite of what I actually said. Welcome to Bennett-commenting crowd, you'll fit right in!

Oops. I'll see if I can get them to change it.

That's interesting. Possible counter-points: (1) It could be argued that if the mayor accessed her work Dropbox account from her home IP address, then that introduces her home IP address into the public record, and if she didn't want that, she shouldn't have accessed it from home. (2) I don't think revealing a person's IP address is quite as bad as revealing a person's home address, because there are "attacks" you can mount against someone once you know their home address (e.g. robbing or vandalizing their house) but the point of the article is that there's relatively little you can do to someone just by knowing their IP address. (3) As I noted, someone could also get the mayor's home IP address anyway just by finding a way to contact her and telling her to visit a certain website.

I don't think it achieves anything to group all of someone's personal information under a heading like "PII" with the implication that it should all be treated the same way. Some things deserve more privacy protection than others, depending on what harmful things someone would be able to *do* with that information.

I think it has to be interpreted in context; what he presumably means here is that the IP address can be used to help identify specific computers (with a certain degree of probability, and maybe depending on whether the ISP has retained customer logs), but emphasizes that that's all it can be used for, and knowing the IP address does not make it easier for an attacker to breach the system's security, so it wasn't a valid reason to reject the FOIA request.

The survey I posted on Mechanical Turk was limited to U.S. users.

I'm not sure how Amazon verifies that it's "U.S. users" are really from the U.S., but there's little incentive for an MTurk user to lie about that, since (1) most tasks posted on MTurk are not limited to U.S. users, so whether you claim to be from the U.S. or not, you're always going to be eligible for more tasks than you can possibly complete anyway, and (2) Amazon can see your IP address *and* knows where it's sending payments for your work. Obviously, you can use a VPN to appear to come from a U.S. IP address, and have your payments sent initially to a U.S. account and then transferred out of the country, but as I noted, there's little incentive for an MTurk worker to do that.

You're right, I'm NOT A STATISTICIAN (although I got my Master's degree in math), but incidentally, I do know how to calculate all the probabilities in the article directly, by calculating the standard deviations and using cumulative normal distributions. I just decided to leave that out of the article because I figured that for the typical Slashdot reader, it would be more convincing to read the perl script, to verify that it's checking what it claims to be checking, and then run it to verify that in 1 million trials, you only get the observed result about 2.5% of the time.

First, that's "bated breath". You don't catch fish with your breath.

Now, as several other commenters have already noted, the probability of getting 38 or more heads in 54 coin flips is about 0.00075, so the results "are unlikely to come from a random generator."

Also, I included a link to a perl script which you can use to verify that, positing a 20-percent gap between the two probabilities in the background population, there was only about a 2.5% chance of observing results similar to what I saw, so that passes the 95% test.

Your writing is very polished -- it's probably the way that most readers wish that I would write -- but the math is wrong.

Well yes, it's a sample only of Washington State small claims court judges. I had just assumed that was obvious enough that I didn't emphasize it, but it's true that you can't generalize from there to, say, federal judges.

In fact, you bring up a good point, which is that multiple lawyers agreed with me that small claims judges were often rather slapdash, but told me that if I ever found myself in a higher-level court, I might be pleasantly surprised that the higher-ranked judges weighed all parties' arguments more seriously. (Of course, as a non-lawyer, it wasn't practical to bring the cases in any court higher than small claims court.)

However, I still say it's correct that even on the basis of a small sample, you can rule out claims about the background population.

Do you disagree with the following argument: Suppose I have a tank containing some mix of blue and red balls. I tell you that only one in a thousand balls in the tank is red, but you have no a priori knowledge as to my reliability. You reach in, grab a ball at random and pull it out, and see that it's red. Even on the basis of that sample of 1, you can still say that my statement is probably false.

It's trivially true that "any small sample is going to have some non-random attributes", but that doesn't mean the sample itself isn't random, or that you can't use it to rule out extreme statements about the background population.

As I said, I included the link to the perl script in the article, so that you don't have to take my word for it about the statistical calculations -- you can run one million trials of the experiment and verify that, under the posited hypothesis, a result similar to the one that occurred will only occur about 2.5% of the time. So the posited hypothesis is probably wrong.

What's an example of one?

Your second sentence is grammatically incorrect -- "I could spot" instead of "I spotted". ("I could spot" makes no sense, because it sounds as if you were able to spot a certain number of mistakes, but of course you could only know that if you actually did spot them.) It's an understandable error for a non-native-speaker, but it's the kind of thing you should watch out for when you are complaining about other people's grammar...

Which one? I wrote three: (1) the easiest way to get to Burning Man (pre-pay for a membership in a camp -- not a turnkey camp, but a camp that requires volunteer hours so that you're contributing meaningfully -- and then pre-arrange for a bike rental, so that you can fly down and take the shuttlebus instead of loading a week's worth of belongings onto a truck); (2) a thought experiment about designing a faster exodus process; and (3) an argument that the ice vendors should have ice pre-loaded onto the vending table, so that people could wait minutes, instead of almost an hour, to get through the ice-purchasing line.

I included the link to the perl script precisely so that you don't have to take my word for it on the statistical calculations. You can run one million trials of the experiment yourself, and verify that under the posited 20% difference in proportion, the frequencies that I observed in my survey would only be that close together about 2.5% of the time.