There are Plenty of Skeptics
June 7, 2014 1:21 PM   Subscribe

 
yeah, good luck with that.
posted by KathrynT at 1:22 PM on June 7, 2014 [42 favorites]


this is serious
posted by philip-random at 1:25 PM on June 7, 2014 [3 favorites]


Right, because that's totally going to work and not cost $3 billion to create.
posted by 1adam12 at 1:26 PM on June 7, 2014 [6 favorites]


I'm sure it will be great at catching false negatives.
posted by Blazecock Pileon at 1:29 PM on June 7, 2014 [9 favorites]


Relax. Sit down. Have a HAMBURGER {\}.
posted by oneswellfoop at 1:30 PM on June 7, 2014 [7 favorites]


bool IsSarcastic ( const string & text ) {
    return text.find ( "{\\}" ) || text.find ( "{/}" || text.find ( "HAMBURGER" );
}
GIVE ME MONEY, US SECRET SERVICE
posted by Flunkie at 1:30 PM on June 7, 2014 [22 favorites]


Hamburger!
posted by Tomorrowful at 1:30 PM on June 7, 2014 [1 favorite]


Uh, and if they give me money, I'll even fix it by taking into account the fact that string::find() returns npos on lack of string.
posted by Flunkie at 1:31 PM on June 7, 2014 [6 favorites]


This is beyond HAMBURGER. This is Big Mac, Whopper, Thickburger, Double Double, Royale WITH CHEESE.
posted by oneswellfoop at 1:33 PM on June 7, 2014 [3 favorites]


In fact I'll even make successfully compile without syntax errors! For money.
posted by Flunkie at 1:34 PM on June 7, 2014 [2 favorites]


"Oh, a sarcasm detector. That's a really useful invention."
posted by strangely stunted trees at 1:35 PM on June 7, 2014 [38 favorites]


I think everyone just failed with their own sarcasm detection. I think it is a sarcastic request from them that demonstrates how hard it is to detect sarcasm. gj ss.
posted by BurnChao at 1:36 PM on June 7, 2014 [8 favorites]


from the Transcript of a Secret Service Executive Meeting...

"Okay, these scandals are making some people think of us as real human beings. What can we do to rebuild our reputation as unthinking, humorless drones?"
posted by oneswellfoop at 1:39 PM on June 7, 2014


Actually, I consider the acknowledgment that humor exists to be a serious step forward for the Secret Service.
posted by Nerd of the North at 1:47 PM on June 7, 2014 [10 favorites]


There is always a /s at the end. Right? .. Right?
posted by jgaiser at 1:48 PM on June 7, 2014 [1 favorite]


This can't possibly end badly.
posted by Etrigan at 1:49 PM on June 7, 2014 [3 favorites]


but can software detect irony and post-irony, and perhaps most important...

can it get secret service agents visiting foreign countries to PAY THEIR HOOKERS instead of just kicking them out of the hotel room so they can create a ruckus in the hallway?
posted by bruce at 1:51 PM on June 7, 2014 [4 favorites]


This is just bizarre.

In other recent bizarro-world USGov news, the CIA now has a Twitter account. First tweet: @CIA We can neither confirm nor deny that this is our first tweet.
posted by feckless fecal fear mongering at 1:52 PM on June 7, 2014 [4 favorites]


Cheap jokes aside, this does raise some interesting questions on the whole subject of sarcasm. I'd have thought the geeks among us would find it a fascinating, if possibly impossible, challenge. At least something to think about, since the jokes are so obvious as to not need expressing.
posted by IndigoJones at 1:55 PM on June 7, 2014 [1 favorite]


I know somebody who received a visit at work from the secret service as a result of an entirely innocent and sarcastic forum comment that was interpreted as a threat to the vice president.

I wonder how many man-hours per year they actually lose chasing down those kind of leads...
posted by 256 at 1:58 PM on June 7, 2014 [7 favorites]




This is so much a better use of federal funds than fixing infrastructure.
posted by arcticseal at 2:03 PM on June 7, 2014 [6 favorites]


fascinating, if possibly impossible, challenge

What differentiates an Onion piece that appears to be written from a conservative point of view, and Fox News?

Impossible doesn't begin to describe it. Nothing short of full Turing-level AI would suffice, and judging from the number of my fundie relatives posting fake-Obama-outrage Onion stories as real news, not always that.
posted by Ryvar at 2:04 PM on June 7, 2014 [4 favorites]


Is this connected to the CIA's first tweet?
posted by infini at 2:05 PM on June 7, 2014 [2 favorites]


Isn't the Secret Service one of the lead agencies involved in high tech crimes? Awesome to see them demonstrate such a deep understanding of technology. I'd be happy to build one, but I'll have to ask them to supply the AI.
posted by feloniousmonk at 2:05 PM on June 7, 2014


Forget Skynet, it looks like we're going to end up with Alpha 60 instead.

A while back there was a post about a guy who was pranking PRISM by emailing sections of Finnegan's Wake to suspect email accounts, and ended up getting into a heap of trouble and interrogated by FBI agents that had simply could not process the meanings of the words at all - it was so much like scenes from Alphaville is was creepy - even more so because the interrogations were by humans, and not a machine.

How they expect to build a machine that can successfully interpret meaning and significance and yet have the results pass an approval people by people that are selected for the cold, mechanical mindset of law enforcement is beyond me.

In any case, it seems what the world needs is a few more people like Lemmy Caution.

Awesome to see them demonstrate such a deep understanding of technology.
Now if they can match it to people, and not just criminals, that would be something.
posted by chambers at 2:11 PM on June 7, 2014 [3 favorites]


the CIA now has a Twitter account. First tweet: @CIA We can neither confirm nor deny that this is our first tweet.

The Secret Service's acceptance criteria for the sarcasm detecting software is that it must properly identify whether the CIA's first tweet counts or not.
posted by radwolf76 at 2:18 PM on June 7, 2014 [6 favorites]


I don't know, I'm not entirely sure it couldn't be done. There is a pattern to much sarcasm, though it's not always obvious. But say they are able to filter out a large amount of "noise", so that human readers only have to look at those that the machine couldn't determine.

But less tricky thank truly detecting sarcasm, I suspect looking at someone's post history and comparing to a threatening statement, one could determine if there is a pattern that needs to throw a red flag. Someone who posts racist screeds or long rants about guns: probably someone to worry about when they say they want to shoot up a school. Someone who posts puppies and kittens, not so much.

Same with temporal context. There are going to be off color jokes in the wake of tragedies. They may not be in the best taste, but I bet there is a certain level of predictability.

Of course, I'd rather not the NSA bee poking around and pattern matching everything we say.

But the task really seems doable. And an amazing study into language and psychology.
posted by [insert clever name here] at 2:20 PM on June 7, 2014 [5 favorites]


This sort of thing is, if not actually in my field, at least within shouting distance. There's quite a bit of actual science going on in the area of sentiment analysis and sarcasm detection right now. The usual approach is to train a machine learning algorithm on known sarcastic text and let it generalize from that, and there's one group of Dutch researchers who've gotten around 75% accuracy on test tweets with that method.
posted by Mr. Bad Example at 2:23 PM on June 7, 2014 [14 favorites]


In other recent bizarro-world USGov news, the CIA now has a Twitter account. First tweet: @CIA We can neither confirm nor deny that this is our first tweet.

Lulzy acknowledgment of clandestine psyops FTW.
posted by save alive nothing that breatheth at 2:28 PM on June 7, 2014 [5 favorites]


... the jokes are so obvious as to not need expressing.

Sure, mm-hmm. Absolutely.
posted by Greg_Ace at 2:30 PM on June 7, 2014 [3 favorites]


I think one of the issues with detecting sarcasm is that it requires additional knowledge not in evidence - so actually it's entirely possible that using algorithms of social media might allow one to have a broad enough net pattern of behavior that one, "Yeah, kill [important person]. Great idea." will stand out as sarcastic against the general flow of the persons' self-expression. (I think this is a lot of why sarcasm fails online - it's difficult to gauge the level of relationship one has with another person when it can be mostly asymmetrical).

That suggests a far wider and more cohesive net than I'm really comfortable with, though, especially given it perforce will include American Citizens and an uncomfortable level of surveillance on all of us, automatically.

Though I did like the CIA tweet. The question is - does that make me a shill, or is it just a really effective psyops campaign?
posted by Deoridhe at 2:32 PM on June 7, 2014


In other recent bizarro-world USGov news, the CIA now has a Twitter account. First tweet: @CIA We can neither confirm nor deny that this is our first tweet.

Wikileaks tries hamburger dressing.
posted by infini at 2:32 PM on June 7, 2014 [1 favorite]


They're looking for someone to design yet another ineffable black-box. If you think the headaches Matt's had with Adsense are bad, wait until you're detained and questioned based on a red flag thrown by some inscrutable, secret Secret Service software. Welcome to the future of law enforcment.
posted by klarck at 2:32 PM on June 7, 2014 [1 favorite]


Welp, its not really the future, not for us little brown funny looking furriners found in airports and suchlike.
posted by infini at 2:33 PM on June 7, 2014 [3 favorites]


I suppose the core of what I am getting at is how once you've crossed some invisible line, and become a suspect/person of interest/potential threat by those whose job it is to investigate such matters, it is very difficult for the investigators to perceive you outside of that context, as it is counter to their training. This issue is neither a modern or a specifically western practice - it's merely an aspect to the role of the investigator that has been evolving over the history of civilization. Those same skills that blind them to nuance and alternate interpretations is not an bad thing on its own - it's necessary to break through a fake story and prevent being tricked by a clever or determined criminal.

My earlier comment was in regards to the future of this technology. I'm sure that basic stuff can be filtered but as this technology develops, things get more complex. There is a conflict between the specifications of the program and how it is built, the results of the programming, and the mindset of an investigator. If you make a machine that is designed to detect something you need to make perceptual assumptions about what you are going to find. In this case, those assumptions are about language, society, and personality - those are some pretty dynamic, ever-changing, and often vague things. Therein lies the first problem. The second is, that even if you build the best possible software, will the results be trusted or relied upon by the investigators? As a society, do we want to put our trust in that program defending us from misinterpretations by the government? Mistakes and errors will be made at all levels, but how will those failures affect the investigator's opinions? Do they trust the machine, or their gut instinct in regards to how they see the danger from a single offhand comment?

It's a fascinating thing to think about. It's times like these I wish we still had William S. Burroughs around to hear his take on this.
posted by chambers at 2:45 PM on June 7, 2014 [2 favorites]


It Took 2 Hours for @CIA to Get Dragged Into Its First Twitter Fight

Welcome to Twitter, @CIA. Perhaps some of the intelligence you've gathered over the years will tell you how to deal with trolls.




he he he hamburger font
posted by infini at 3:07 PM on June 7, 2014 [4 favorites]


I think sarcasm is trickier than some folks think. I try hard to not use it online for complex reasons that boil down to I have enough trouble being understood when I am playing things as straight as I can manage.

Aspies have trouble with the entire concept of sarcasm. My oldest son tells me he only figured out sarcasm because of my Dripping Sarcasm Voice. I have a tone I use exclusively to signal that I am being sarcastic and it is pretty over the top exaggerated. He says he could tell there was something funny going on there and that led him to eventually figure out sarcasm or he would not have ever gotten it.

I suspect that quirk of mine -- of having a Dripping Sarcasm Voice -- grows in part out of being one of the top students of my graduating high school class and too often finding myself in situations where people had no idea I was being sarcastic because they lacked the chemistry or physics or whatever to get, without an explanation, that X could not possibly work. It causes a great many problems if people take me seriously when I am not being sarcastic and there seems to be no good way to back pedal. So I just learned to do a voiced form of FLASHING NEON SIGN to signal my intent.

I have never learned an internet equivalent of that. So I try like hell to just not go there. I think sarcasm is just really hard online, in part because you can connect to so many people whose lives are so vastly different from yours that it muddles the question of context upon which comprehending sarcasm typically hangs. I have seen foreigners who speak English as a second language totally fail to get such subtleties in an American dominated forum and then no one had the heart (or perhaps manners) to explain "um, no, we were actually saying..."


{/}Thus, I am all for simplifying this issue by outlawing sarcasm on the internet. {/}

(Perhaps mefi will give me new bad habits since it does, in fact, have a cultural artifact which can serve as a substitute for The Dripping Sarcasm Voice.)
posted by Michele in California at 3:13 PM on June 7, 2014 [5 favorites]


Dear 1979,

This cyberpunk future is great. I wish you could see it. Everyone has a supercomputer video phone the size of a candy bar in their pocket! Almost every computer is connected to every other computer in a system that entertains and allows anyone to communicate with everyone.

Of course, you have to be careful what you say since the government uses a super-super computer to read all of these communications to decide if you're a threat. They haven't worked all the bugs out yet. There are some other problems too, but, again, I don't want to say too much here.

Well, I'm off to golf on the moon! (just kidding.)
posted by double block and bleed at 3:23 PM on June 7, 2014 [16 favorites]


I'm pretty sure I could write this code. In fact, I'm pretty sure anyone could write the code that the SS is looking for. My algorithm boils down to this:

If someone made a threat to the POTUS on Twitter or Facebook or Google+ or whatever, IT'S SARCASM. Because NOBODY WHO ISN'T HARMLESSLY INCOMPETENT USES SOCIAL MEDIA TO PLAN AN ASSASSINATION.
posted by um at 4:01 PM on June 7, 2014 [7 favorites]


Of course you do realize that after it became publically known that the secret service completely dismisses public threats on social media as not a serious plan, that then becomes THE way to plan an assassination of the POTUS, far more secure than anything else.
posted by Michele in California at 4:12 PM on June 7, 2014 [3 favorites]


Once you announce your channel for propaganda on the news, then what?

No wonder you have to go looking for the hamburger meat.
posted by infini at 4:17 PM on June 7, 2014


NOBODY WHO ISN'T HARMLESSLY INCOMPETENT USES SOCIAL MEDIA TO PLAN AN ASSASSINATION.

I think you may be overestimating the average level of competence in this country. (no sarcasm)
posted by Greg_Ace at 4:38 PM on June 7, 2014 [7 favorites]


I don't know why everyone is so down on this. Like Velcro and Tang, I believe could have wide application beyond its original military spec'ed purpose. Imagine a sarcasm detector built into your grandparents Internet Explorer. Not looking so stupid now, eh?
posted by fshgrl at 4:46 PM on June 7, 2014 [3 favorites]


NOBODY WHO ISN'T HARMLESSLY INCOMPETENT USES SOCIAL MEDIA TO PLAN AN ASSASSINATION.

I disagree. You could be delusional but still quite dangerous. Prime example? That college kid in Santa Barbara. He posted videos on YouTube for Pete's sake. His target of choice was all the girls who rejected him & not the President but that's just a detail, it could just as easily have been a threat against somebody under the Service's protection.

I don't think they'll ever rely on software to be the sole factor in determining whether an apparent threat is serious enough to investigate or not but as a tool to aid human judgement in sifting through the mountains of tweets & other social media posts it could be useful, especially if it can articulate why it thinks this post is threatening & that one isn't.
posted by scalefree at 5:02 PM on June 7, 2014 [3 favorites]


Well it certainly seems more reasonable and effective to spend money on detecting sarcasm on social media than increase funding to help mental illness.
posted by entropicamericana at 5:06 PM on June 7, 2014 [5 favorites]


"detecting sarcasm online". It would help, of course, if they [whoever they 'is'] could detect it offline. They have yet to be cured of their dumbness.
posted by SteveLaudig at 5:06 PM on June 7, 2014 [2 favorites]


I see Stage 1: Collecting a Representative Sample is coming along smoothly.
posted by pwnguin at 5:12 PM on June 7, 2014 [6 favorites]


But what if it works too well?

What happens when we detect hidden sarcasm in the US constitution? Or the Bible? I'm imagining the headlines now -- "Four out of Ten Commandments are Sarcastic, Study Says"
posted by serif at 5:25 PM on June 7, 2014 [14 favorites]




Given how hard it is for computers to understand people, and how hard it is for people to understand sarcasm, I'm sure this is going to go well.

Not one, but two, applicable Burnistoun skits. This is going to go well. :)
posted by adamt at 6:38 PM on June 7, 2014 [3 favorites]


All content on the internet should be assumed to be sarcastic until proven otherwise. Except this comment.
posted by _aa_ at 7:15 PM on June 7, 2014


First tweet: @CIA We can neither confirm nor deny that this is our first tweet.

haha the CIA made a funny tweet, how cryptofascist could they really be!!
posted by threeants at 7:52 PM on June 7, 2014 [2 favorites]


turn that frown upside down, Chileans
posted by threeants at 7:52 PM on June 7, 2014 [1 favorite]


Cheap jokes aside...

Na, cheap jokes above, below, around. and over.

USSS, FLEA, CIA, FBI for the LULZ.
posted by BlueHorse at 8:48 PM on June 7, 2014


. There's quite a bit of actual science going on in the area of sentiment analysis and sarcasm detection right now.

Yes. It's part of the metadata that they collect. All the better to determine whether you're angry enough to march on Washington or Wall Street. There will be no revolution to televise.
posted by five fresh fish at 10:37 PM on June 7, 2014


I'm sure none of the people making fun of the Secret Service were complaining about how the police should have seen Elliott Rogers's videos and stopped him before he killed all those people.
posted by straight at 10:51 PM on June 7, 2014 [1 favorite]


SEMANTIC WEB
posted by klangklangston at 12:34 AM on June 8, 2014


<sarcasm>
posted by klangklangston at 12:37 AM on June 8, 2014 [1 favorite]


This is really curious to me, as an Australian. It's a loooong held and widely perpetuated belief here that while Americans are clever and funny and awesome at a wide range of things.... that, generally speaking, they REALLY don't get sarcasm.

Maybe the reason all the Americans here seem to get it and have fun with it, is that MeFites are the best at everything. Hell, yeah!


Although I frequently wonder if it's just that Australians can be insensitive, casually racist and vulgar and we're all like, "Maaaate, it's a joke" and Americans actually generally have better manners than us and say "Hey cut it out, that's not cool, joke or not." And then we're all like, "Humourless Yanks."

Not sure, but MetaFilter ruins the old Australian "Americans don't do sarcasm" trope.
Be proud, possums!
posted by taff at 1:01 AM on June 8, 2014 [1 favorite]


Sarcasm online, eh? Cos I wasn't quite sure what to make of this headline: US begins satellite attack on Boko Haram •To launch N960m tv station in Northern Nigeria, given what it says in the body of the article.

they outlined challenges the new satellite project would face in a region with low levels of infrastructure, public services, literacy and security, especially where only few people own or have access to televisions. They pinpointed the electricity challenges of the country as one of the critical challenges.
posted by glasseyes at 2:12 AM on June 8, 2014 [1 favorite]


That's why I linked to it mama
posted by infini at 3:02 AM on June 8, 2014


This is really curious to me, as an Australian. It's a loooong held and widely perpetuated belief here that while Americans are clever and funny and awesome at a wide range of things.... that, generally speaking, they REALLY don't get sarcasm.

Oh, they get it, so long as it's done very broadly with a lot of winking and nodding and at least two neon signs with arrows pointing "SARCASM!" at it and a laugh track afterwards. Otherwise there's just no way to know. That's why we need a special sarcasm font for the Internet, one other than Times New Learn How To Write.
posted by Legomancer at 5:28 AM on June 8, 2014


isn't that what comic sans is for?
posted by feckless fecal fear mongering at 5:59 AM on June 8, 2014


The Secret Service should just bail out metafilter. The Mod squad is well versed in detecting all nuances of human digital expression and they are just about capable of passing the Turing test.
posted by Renoroc at 7:34 AM on June 8, 2014


isn't that what comic sans is for?

No, it's for pissing off hipster typographic geeks.
posted by Greg_Ace at 12:15 PM on June 8, 2014 [1 favorite]


Detecting sarcasm will undoubtedly be an easy chore for today's smart COBOL programmers. And I personally admire the secret service, they seldom stay up all night drinking anymore.
posted by Twang at 12:56 PM on June 8, 2014 [2 favorites]


COBOL? I thought this kind of federal government project had to be written in ADA, the language for programmers who feel that COBOL is just way too flexible. Or was that just DoD contracts?
posted by radwolf76 at 1:11 PM on June 8, 2014


A while back there was a post about a guy who was pranking PRISM by emailing sections of Finnegan's Wake to suspect email accounts

Which was a piece of fiction, in case you didn't realize.

ADA, the language for ...

Ada, if it were invented today, would be hailed by the HN crowd as a pretty great modern language. It's quite unfairly maligned.
posted by dmd at 5:14 PM on June 8, 2014


In general, I think a sarcasm font would be useful. It would keep me out of trouble.
posted by cherrybounce at 9:26 PM on June 8, 2014


In general, I think a sarcasm font would be useful. It would keep me out of trouble.

Oh, if only life were that simple. A font that magically shields one from trouble. Can I have a big box of that??
posted by Michele in California at 11:53 AM on June 9, 2014


A while back there was a post about a guy who was pranking PRISM by emailing sections of Finnegan's Wake to suspect email accounts

Which was a piece of fiction, in case you didn't realize.


I had completely forgotten that it ended up being fictional - thanks for the correction. In any case, he depiction of the FBI agents rings very true to me. I've encountered that demeanor in a few (not all) people that work in law enforcement, but thankfully I've not had to be in an interview room with them.
posted by chambers at 11:35 AM on June 10, 2014


They need to hire this guy as a consultant.
posted by homunculus at 1:31 AM on June 11, 2014


« Older Hons and rebels   |   My posse's got an orchestra! Newer »


This thread has been archived and is closed to new comments