Hacker News Reader

Comment by maze-le | Hacker News Reader

I don't think that is entireley true. While the Schengen-Zone (EU border agreements) has established that there are no physical borders, where you have to show an id, there is a 50km (maybe more maybe less I don't know exactly) zone around those borders, where Identity checks are permitted. Police is allowed to control your identity in those zones if they have a "reasonable cause" to think yo may be an illegal immigrant...

Of course "reasonable cause" can (and usually will) mean, that if you happen to have the "wrong" skin-tone, hair-color etc. you will be controlled. The Police justifies this racial profiling with "experience" and "statistics" (at least in Germany)

View on HN

In theory, they can't do that here in the US either.

> Border Patrol, nevertheless, cannot pull anyone over without "reasonable suspicion" of an immigration violation or crime (reasonable suspicion is more than just a "hunch"). Similarly, Border Patrol cannot search vehicles in the 100-mile zone without a warrant or "probable cause" (a reasonable belief, based on the circumstances, that an immigration violation or crime has likely occurred).

> In practice, Border Patrol agents routinely ignore or misunderstand the limits of their legal authority in the course of individual stops, resulting in violations of the constitutional rights of innocent people. These problems are compounded by inadequate training for Border Patrol agents, a lack of oversight by CBP and the U.S. Department of Homeland Security, and the consistent failure of CBP to hold agents accountable for abuse. Thus, although the 100-mile border zone is not literally "Constitution free," the U.S. government frequently acts like it is.

https://www.aclu.org/constitution-100-mile-border-zone

The justification is not "experience" and "statistics". The justification is that it works well. Just like the language of an email can help identifying spam, skin color can help identifying illegal immigrants. I would say that 99% of all spam I receive is written in English, whereas about 80% of my personal emails are in German. Thus, even though there is nothing wrong with that language in general, English words can be used as one factor of many to help identifying spam.

> I would say that 99% of all spam I receive is written in English, whereas about 80% of my personal emails are in German.

You're missing some important numbers here for this to be meaningful: Base rates.

If you're getting 10 spam emails per month, and 99% of them are in English, but you're receiving 1000 legitimate personal emails per month, and 20% of them are in English, that's a total of 210 English emails on average per month. In this case, while technically true that an email written in English is more likely to be spam than an email written in German, it's also such a very poor indicator that you'd be better off using another indicator entirely.

On top of that, you've only pinpointed two categories out of many more possible categories of emails. What if someone is getting 50 spam emails per month, 5 job offer emails per month, 10 updates relevant to their profession per month, 100 professional emails per month, and 50 personal emails per month, and the English:German ratios for those are all 99:1, 9:1, 4:1, 1:1 and 4:1 respectively? Suddenly just "this email is written in English" isn't all that important anymore.

Base rates are very, very, very, super important when discussing stats like these.

That's why I said "one factor of many". If you have ten factors like that and combine them, you'll get an excellent spam filter. That's how Bayesian spam filters work. They combine a number of factors that are not very significant on their own, but using all of them adds up.

Reply on news.ycombinator.com