Comment by StandardFuture | Hacker News Reader

Comment by StandardFuture | Hacker News Reader

Parent Full thread

While, I agree this COULD all be a sham story for someone who may be a bit of an attention seeker ... I would like to counter your arguments anyways to show that this story is actually perfectly believable.

In the article, the 'Task Force' (http://en.wikipedia.org/wiki/Joint_Terrorism_Task_Force) probably has a set of "Tasks" that they must accomplish to be ... well ... worth being a department with a budget. And I imagine one of those is following up on "profile-perfect" households or individuals.

So what would make this household 'profile-perfect'? Well it is true that there were searches for Pressure cookers and backpacks but that alone won't mean much due to the massive amounts of searches for these products.

But BOTH searches were done; and yet still I have to agree that the intersection still does not provide an un-biased view.

However, what if we take into consideration the timescale at which these searches (and possible purchases) took place? We know that the searches were made WITHIN the same week. This factor could possibly give us stronger signals for a profile-match.

Now what if we couple that with 'international travel'(e.g. how often a passport was used in a system for out-of-country travel)? The profile match would become higher due to the husbands business trips.

And I don't know where these people live, but location could also be a factor as well (e.g. the closer you live to a major city the greater your risk potential factor).

Now the cross-section for these factors would have such precision that doing a dragnet on this profile-match set would be feasible. And I imagine it would amount to less than 100 households a week (although I think if the story is true that the agent said this just to 'console' the family that this is common practice and there is nothing to worry about).

What do you guys think?

EDIT: Also not to mention that another factor could be 'Copy-cat-syndrome'(made up the term) e.g. how many things were occurring at this household to make them appear like potential Boston-bomber copycats?

You're not engaging with my point. Manpower and expense is a problem with this story, yes, but not the main problem; the main problem is that it's a strategy that is practically guaranteed to betray the existence of a whole new domestic Internet surveillance program while at the same time being almost mathematically guaranteed never to find a single terrorist.

She's inferring that her searches are the cause. It's not clear that the agents ever explicitly stated that they used search history to profile the family. They asked, "Have you ever looked up pressure cooker bombs?" They might ask this question of everyone.

Someone posted this below which I find hilarious.

"Section Chief: You dispatched field agents without understanding the base rate theorem!? You're a loose cannon, McGillicutty...

McGillicutty: ...I get results!

Section Chief: Mathematically, you shouldn't. Ergo, you don't."

Actually I addressed both the Financial and Mathematical points of your initial comment quite avidly/thoroughly.

No, you think you have, but because you don't understand the base rate theorem you're just talking past me. It's not enough to filter the population down to 100 households. Nothing you do with a filter changes the base rate of terrorist plots, which is extremely low. Profile 100 households, you get 100 false positives. Profile 50 households, you get 50 false positives. The likelihood of any one house you investigate housing an actual plot is always going to be very, very low.

I don't think I could facepalm any harder. What you are basically stipulating is that because pursuing certain leads could be almost always wrong that this would prevent a government department from following up on them? Are you kidding me?

>Nothing you do with a filter changes the base rate of terrorist plots, which is extremely low.

In that case, I imagine the base rate of 'post-successful terrorist act plotting' is higher than 'pre-successful terrorist act plotting,' which would eliminate your argument.

Also, what you would be studying is the TYPE of terrorist activity not total plot numbers e.g. the stages of activity. If a successful terrorist attack was pulled off not to long ago in the past then that would call for an increase in the base rate of the TYPE of plotting. More terrorists would be more invigorated to pursue/look into similar methods as their next plot.

Maybe the cross-section of the plotting-type-pivot is what you could match against?

The simplest way to express my facepalm is: you are tremendously oversimplifying things.

Surely any increase in wannabe-copycat terrorists Googling for items used in recent terrorist attacks would be more than cancelled out by a much larger proportional increase in the much larger group of people that follow the news and don't want to be terrorists also Googling for items used in recent attacks

Section Chief: You dispatched field agents without understanding the base rate theorem!? You're a loose cannon, McGillicutty...

McGillicutty: ...I get results!

Section Chief: Mathematically, you shouldn't. Ergo, you don't.

I do not think so.

Filtering to target at high-risk sub-groups - which, among themselves, have different base rates than the general population- is entirely effective.

It is precisely the reasoning under which medical screening - ineffective, generally, due to the base rate - is aimed at, say, "Men Over 50", where the base rate grows sufficiently large to make screening worthwhile.

What trouble do you have with the fact that there aren't 100 terrorist plots in NYC? There might not even be 2 of them. If they're filtering down to 100 visits a week, something close to 99% of the visits are false positives.

tptacek seems hell-bent on not giving the side for which he has no factual data any mindshare at all--there are arguments for that mindset.

As for your filtering observation, yeah, I have no idea if he just mis-stated his position, is using different definitions, or today just can't into math.

One of two possibilities arise from this comment. The first is that you think that there might be many many terrorist plots ongoing at any time. The other is that you don't understand the simple idea of a base rate and how it applies to statistical filters.

So, your first possibility is incorrect.

The second is worth investigating, and in the interest of teaching others I'll provide a helpful link:

http://en.wikipedia.org/wiki/Base_rate_fallacy

It's unfortunate, because that article and the base rate article have some contradictory conclusions--but, the linked article shows its math.

What's really troubling me here though is that you seem unwilling even as a thought experiment to entertain the notion that this sort of behavior could be purposeful by the government. In many, many prior discussions you've held this view, you've advocated for stripping rights, you've advocated for screwing over people scraping public unauthenticated APIs, you've advocated a great many things against the general notion of the public good, of privacy, and of exploration.

Perhaps come down here and--in the name of intellectual honesty--try and understand what exactly has so many of these people bothered. As it is, due to your choice of message and presentation, you often come off as at best aloof and at worst willfully ignorant.

I like how this comment goes from acknowledging that the author didn't know about an extremely basic and fundamental concept of statistics to a whole paragraph of (frankly weird) assertions about my character and beliefs. It's almost gymnastic.

I think one place to start moving forward is to acknowledge that you really have no idea what I do or don't believe.

How would you stack this against the Guardian article that was quoted further down the page?

A spokesman for the FBI told to the Guardian on Thursday that its investigators were not involved in the visit, but that "she was visited by Nassau County police department … They were working in conjunction with Suffolk County police department."

You're suggesting that the FBI always does the optimal thing. This may not be the case. To put it mildly. I refer you to Hanlon's razor: http://en.wikipedia.org/wiki/Hanlon's_razor

Surveillance of Google search terms wouldn't be revealed unless someone told the suspects directly (which these bumbling police officers essentially did).

It is entirely plausible that searches for "pressure cookers" and backpacks are some of several factors used to flag suspects. There is almost certainly a mandate to prevent pressure cooker bombings. I would guess they allotted some resource budget to that task and then simply created search criteria that narrowed the set of cases into one that could be investigated using those resources. I am pretty certain they didn't make any calculations of posterior probabilities. There aren't nearly enough samples to make any kind of meaningful statistical assessment of efficacy, even if you believed that that was the primary decision making tool of the security agencies.

Assuming the story is true the NSA or FBI probably flagged these individuals using a number of factors (search terms, "anti-government" writing, etc) but not enough to merit use of limited federal resources. Still they hit on several risk factors so local police are informed, they may have additional information about these people and have boots on the ground to keep on eye on them (if something did happen then NSA and FBI could wash their hands: "We told the local police, they did nothing"). The comparatively unsophisticated local police, partly motivated by the same fear of being responsible if something did happen, then ham-fisted the investigation.

Agree on plausibility. Also, don't forget the 20-year-old son. Bet they checked the medicine cabinet.

Reply on news.ycombinator.com