You actually just fell for it. Not only did you fall for it, but you also missed the point of the example.

You're thinking in the wrong way. You have a pool of people and an automated system of sniffing out bad guys. You detect one. You don't have the 90% chance of that person being a bad guy, but just a .3% chance due to him being selected from a pool.

This is exactly what's going on right now. The pool size is the entire United States of America. The entire pool is being sampled to find terrorists, and because the automated systems sampling the pool have error rates, they come up with false positives (which is the example of this article) an order of magnitudes more often than they come up with terrorists.

THis is a matter of false positives, not of catching bad guys. The concept in the united states is innocent until proven guilty, but due to massive size of the sample pool (310 million and growing), even small error rates generate far more false positives than correct matches.

99.7% of people detected in the city by the machine are 100% innocent and only .3% of the people detected by the machine are guilty. Those probabilities are the definition of doing more harm than good.

Nope. Hell, even only half of what I do on the web is monitored and sold. The vast majority of what I do on the Internet doesn't even involve the web.

Fun fact: the NSA destroys all non-encrypted files under PRISM, but keeps encrypted files for cryptographic anolysis (read: until they can crack them). This means that, for me at least, the NSA has a nice backup of all my computers

Ah, interesting, thank you. I'll be sure to google those.

I do know about the Rainbow Warrior, it was a Dutch ship after all. Quite a stir over here when that one became public.

No, I got it. You didn't. There are severals problems with his anology, but the most pertinent are 1. His example was based on there being only 1 terrorist, while in the real world, that number is undefined. We don't know how many potential terrorists are out there. All 301 positives could actually have been terrorists, for all we know, which would have made the accuracy 100%. What a win for us! 2. His example is predicated on identifying terrorists based on a single test. The government isn't going out and throwing people in jail for a single data point. In the example that started all of this, we know that those people were found through the conjunction of at least two different data points. Now, if it turns out that those were the only two data points they had, then all of the government agents involved deserve to be slapped upside the head. However, even if the people pinged on large number of data points that they were not aware of, it is not possible to eliminate error in the system. That is why they were only questioned rather than immediately executed.

I think we can both agree that you are hardly a typical web user.

That might come in handy if your computer and all your back-ups somehow manage to crash at the same time.

Yeah, not like anyone has ever been wrongfully convicted before.....oh wait.

To note: the reason the base rate fallacy applies is because of the huge gap in the pool size vs the representative size of the target group. If the United States was made up of 50% terrorists instead of some really small number (say, 50k terrorists in the USA, thus making up .00016% of the population) then the base rate fallacy wouldn't really be a problem. Thus it'll report non-target group members (non-terrorists) as a false positive many, many times more than it'll actually find a bona-fide target-group member.

It's just a mathematical fact.

The percentages aren't in that favor. We do have a pretty darn good idea as to the number, to say the number is undefined is a fallacy.

*sigh* I see you're still failing to grasp the idea behind the base rate fallacy. It's ok. I'm sure you also fail at other statistics like the http://en.wikipedia.org/wiki/Monty_Hall_problem. Statistics goes against common sense oftentimes.

The base rate fallacy has nothing to do with guilty/not guilty in jail/not injail, but about how often an experiment produces the wrong result (false-positive). Which is what this thread started with. The system labels an innocent person as a terrorist far more often than it labels a terrorist a terrorist. This has many social and economic downsides, which far outweigh the marginal benefits especially when you consider the privacy loss as well.

I am not saying that searching for bad guys is wrong, but rather that the automated systems being used today are fundamentally broken in both logic and actual mathematics. The whole system also throws the legal concept of innocent until proven guilty on its back.

See Spot, see Spot run, see Spot carry 2 kilo's of Semtex, run Spot run.

Now if we could only figure out what a normal family could possibly want with a pressure cooker or a backpack....

You don't watch or listen to a lot of news do you? Seems there are a lot of sheep that think the NSA spying is a good thing while the remainder of the wolves see it as illegal and infringing on their rights.

I'm the latter and feel that the NSA should be disbanded and every last one of them should be sent to prison.

Isn't it the wolves that the sheep need to be protected from?

Well yeah but I couldn't think of an animal that was both smart and cool that could be used to reference people with some intelligence to see that what the NSA is doing is illegal.

#Run on sentence FTW.

More like 99 sets of toes, each week, is how many they get wrong.

Understood. For me there's a difference between a person and a religion. Action vs. idea if that makes sense. This could be taken to PM if you wish, but I dont' really have the time to put into it (not meant to evade nor degrade, just a fact...it's time to cook supper

But frankly, my limited knowledge of and expereince with religion makes me opposed to them in genereal Again, another PM topic

It's not news over here.

Well, then I'm sure we should let all of the criminals go then. You never know, they might have been wrongfully convicted.

So give us the number.

No, I am quite familiar with that one. Nice try at using an ad hominem fallacy to support your argument, though.

An experiment does not automatically transfer over to real life. Just because your extremely limited example turns out that result does not mean that the actual practice has the same result. Especially since multiple levels of control can dramatically reduce the margin of error. Especially since talking to these people was another level of control and not the end result. This is why they were not executed on the spot.

The thing about statistics is that you can make them say what ever you want them to. I see your Benjamin Franklin and raise you a Mark Twain on this subject.

What the NSA is doing is within the bounds of US law. It is legal. Your trying to paint your opponents as the opposite of cool and intelligent does nothing to forward this conversation in any way.

Yep, the anology is usually the sheeple and the sheepdogs

Narf! poit

Ehm. Yes it does.

That is what statistics is for.

It's a practical application of mathematics.

You are positing here that every experiment ever conducted had its results undisputably confirmed when applied to real world situations? A very brave stance to take.

You'd have to ask the government for it, but it's well below 0.5%

No ad hominem in pointing out the fact that statistics goes against common sense using the example best known for making people's head explode.

Statistics is one of the few branches of anything where experiments DO automatically transfer over to real life.

Uhg, you still don't get it I see. This whole thread is the problem: too many innocents in the crossfire. There's way too much collateral damage for a very marginal return. Yes, these people aren't getting killed, but their lives are being ruined as well as the general public losing hundreds of billions of dollars due to spending on these automated systems that don't work, loss of trade secrets, and all sorts of other problems.

IF there were ABSOLUTELY no downsides to the practice besides the false positive problem, then maybe you'd have a leg to stand up on, but witch hunts create more problems than they solve. Always.

Bending statistics is what Mark Twain hated. Not statistical laws. Probabilities are written in stone.

Uhg. You missed his point too. Statistics is based in the real-world. Every single thing in scientific statistics (Which is where the base line fallacy exists) is 100% always replicable in reality to within error margin (which is to say, if you were to take samples of samples, it'd always average out to exactly what statistics tell you it should).

That's what makes statistics so useful.

I am positing no such thing my good sir, as you are undoubtably quite aware.

I am however not in the habit of showing people what it is exactly that they said to which my reply was relevant and accurate.

What I don't get is why hasn't eveyone in America as a protest just start doing searches for pressure cookers, backpacks, Uranium, Dirty Bombs, and a whole host of other things that would get the NSA or whoever else is spying all the time to go nuts.

But hey we live in a crazy world, I could say more but I will leave it with that.

I think that would be more something for the English, they seem to have a knack for cheery civil disobedience.

But hey, I think it's a great idea and a novel form of 21st century protest.

Perhaps some redditor can run with the idea or something.

Edit: Thanks for posting the http://en.wikipedia.org/wiki/Monty_Hall_problem Defron, that was a delightful read. I knew about the concept before this from QI, but nothing in-depth. This Vos Savant sounds like an amazing character.

When it comes to probabilities, yes. It's what makes statistics so powerful: statistics are the exact things that will happen in the real world to within error margin (which is to say, when you take multiple independent samples, you always will average to the statistical outcome)

The entirety of REAL statistics is built on this fact that is mirrors the real-world. If it didn't, statistics would be an entirely useless field of mathematics.

Other experiments aren't necessarily the case, but that's what makes statistics useful is that it IS the case.

Nah, all that we'd do is post about it relentlessly on Facebook to piss our friends off.

I think as more of these encounters come to light, this will start to happen. Remember the list of key words that the NSA triggers on? Two of them were Freedom and Liberty.

So you don't know.

Implying I am wrong because I, assumedly, do not understand an unrelated example is an ad hominem.

Please tell that to black people that are pulled over because black people are statistically more likely to commit crimes.

I would hardly say these people's lives were ruined. Yes, it was probably unconfortable being questioned by government agents, but I am sure they will get over it. Especially considering their snarky responses to said government agents.

I was once questioned by the police because I was walking through an area where a burglary had just taken place. It was annoying to have to explain what I was doing out and about at 11:30 and why I was carrying a bag of Taco Bell food*, but I was not scarred for life.

*Technically food, anyway.

Well, our legal system is immensely costly and innocent people do get sent to prison, even executed. If there were absolutely no downsides, then maybe the police would have a leg to stand on, but I say it is all just a witch hunt.

Only if you correctly account for all variables. I am saying that you haven't, therefore you are bending statistics to meet your own standpoint.

I said exactly what I said. That his experiment was limited by arbitrary factors that do not necessarily exist in real life and is therefor not necessarily representative of real life.

I searched both of those multiple times today. If anybody needs me, I will be down at NSA headquarters.