Software Needed To Detect If This Post Is Or Is Not True
from the quandary dept
No matter what advances technology throws at us, people remain fascinated with developing the ability to detect when people are lying. Polygraphs remain largely inaccurate (and easily gamed), so researchers focused on the legal, security and defense markets remain busy, while others explore new ways to detect lies and and learn more about people in other fields as well. While we've seen before applications for mobile phones that purport to be able to detect lies, some researchers at Cornell now think they can develop software that will be able to detect lies in emails and text messages. They say they can use linguistic information like word choice, shifts in verb tense and use of the passive voice to detect lies, and they've analyzed materials such as emails from the Enron fraud case to hone their methods. They plan to spend the next three years working on a system to evaluate the content and context of communications, with a view to training software to be able to detect subtle changes that may indicate a lie. Or at least that's what the article says. After all, they could be lying.Thank you for reading this Techdirt post. With so many things competing for everyone’s attention these days, we really appreciate you giving us your time. We work hard every day to put quality content out there for our community.
Techdirt is one of the few remaining truly independent media outlets. We do not have a giant corporation behind us, and we rely heavily on our community to support us, in an age when advertisers are increasingly uninterested in sponsoring small, independent sites — especially a site like ours that is unwilling to pull punches in its reporting and analysis.
While other websites have resorted to paywalls, registration requirements, and increasingly annoying/intrusive advertising, we have always kept Techdirt open and available to anyone. But in order to continue doing so, we need your support. We offer a variety of ways for our readers to support us, from direct donations to special subscriptions and cool merchandise — and every little bit helps. Thank you.
–The Techdirt Team
Reader Comments
Subscribe: RSS
View by: Time | Thread
Put Your Money Where You Mouth is :)
[ link to this | view in thread ]
Re: Put Your Money Where You Mouth is :)
Anyway.
I can't see how this could possibly work. The reason detecting lying in speech is possible (maybe not 100% accurate and there are ways to cheat, but it can work at times) is that you more or less instinctively tell the truth. When you lie, you're trying to override that instinct and since its all in real-time, you don't have that much time to prepare yourself. Hence its easier to detect someone lying on the spot as opposed to someone who was prepared to lie. When you type out an e-mail, you usually look over it a couple times, proofread it, make sure it all sounds good, etc. The only thing I can see being possible to pick up is when someone lies about something they know nothing about. Doesn't matter how much you proofread in a situation like that. It may sound good, but there may be contextual flaws or contradictions which you're not aware of. That would seem hard to write an algorithm for since it may require special knowledge of what's being talked about.
This probably just picks up on buzzwords that managers throw around when they have no clue what they're talking about. They should try it on non-business documents and more personal ones to see if it still works.
[ link to this | view in thread ]
Of course it can't work
[ link to this | view in thread ]
wtf
[ link to this | view in thread ]
Detect lying in emails?
[ link to this | view in thread ]
"This sentence is False"
If the UTM says it is true then the UTM lies because it says it is false, if that is a true statement it must be false... if UTM says it is false UTM loses again becuase the the fact that it is not false makes it a true statement...
No logic circuit can beat it, dualistic thinking does not explain all that is under the sun my friends.
[ link to this | view in thread ]
Type I Error
Will software like this tell humans when someone really is telling the truth, and the skeptics are wrong?
[ link to this | view in thread ]
Re: Of course it can't work
I'm not sure how makeup matters... A RUSE on the other hand...
[ link to this | view in thread ]
Oblig Star Wars reference?
[ link to this | view in thread ]
This will work.... Until you reach someone like me
"Me fail English? thats unpossible!" ~ Ralph Wiggum
[ link to this | view in thread ]
Re: Oblig Star Wars reference?
Who would use it? Boyfriends and husbands trying to see if their significant other is cheating on them.
[ link to this | view in thread ]
can't even catch me in real life..
I had a friend who was in Army Intelligence tell me how they learn to lie to a lie detector: they cut a hole in the bottom of one of their socks and put bologna in their shoe, then when they answer questions they press their foot firmly into the bologna, say to themselves "I would never do that"(with the "that" meaning put bologna in their shoe but the brain plays with the ambiguity of the word and applies it to the question) and then answer the question how they wish. The body apparently learns the response and you learn to lie like Gonzales, err, any politician..at least for a polygraph.
I lie on a daily basis...I'm a magician and looking someone square in the eye while telling them that "this next trick was my grandfather's favorite card trick" while knowing that one wouldn't touch a deck of cards and the other was nothing close to a role model for me, is nothing new.
Given the fact that email isn't a real-time happening, I don't see this software working very well on those who will spend an hour retyping a single sentence until it reads exactly what they want it to say. I also don't see it working well with stories that shift tense because they are describing a chronology...e.g. "I was going to go to the store to pick up some milk. I stopped at the stop light and while looking for my sunglasses, the other car hit me from behind". Knowing that to date there's no way that computers can comprehend what we're really saying, someone is going to find themselves at the wrong end of a defamation lawsuit over this system.
[ link to this | view in thread ]
Re: people are educated
[ link to this | view in thread ]
It won't work
I mean seriously this is one of those non-arguments for people who believe the world (as with lying) is black and white; when its truly in shades of gray.
Did you do "XYZ"? Lying to that sort of question is a great deal different than answering to, "Do these jeans make me look fat?"
Some things, such as facts, can be checked where as other things are far more subjective. Its like these fools who run around being crude saying "I'm only telling the truth", the truth according to whom? A rude asshole?
And on that note........no those jeans don't make your ass look fat, your fat ass makes your ass look fat. ;-)
[ link to this | view in thread ]
Re: Re: Put Your Money Where You Mouth is :)
[ link to this | view in thread ]
Re: Re: Re: Put Your Money Where You Mouth is :)
Yes, someone who lies does tend to do this, but many things written are also presented in passive voice to generalise a topic for consumption en masse; and the passive voice is also quite common to many lacking self-confidence in their writing (even in cases where it is entirely accurate) to dissociate themselves from it, but not necessarily because it's untrue.
[ link to this | view in thread ]
why do the research when posters here already know
from the article:
In the past three years, professors and students from several universities have coordinated with counterparts at Cornell to conduct experiments, analyze language, and compile a list of indicators of written deception. They have drawn from 40 years of research in linguistics and lies, including recent work in the context of computer media and reviews of Enron e-mails.
The National Science Foundation awarded the researchers $680,000 last summer to develop software that would sift through text and flag messages in which in the authors had lied.
"There's still an open question of whether that is actually possible or not," said Jeff Hancock, a communications professor and information science faculty member at Cornell. "Our research suggests that it is."
[ link to this | view in thread ]
A RUSE on the other hand...
[ link to this | view in thread ]
re: dorpus
[ link to this | view in thread ]
Re: Re: people are educated
[ link to this | view in thread ]
Re: Put Your Money Where You Mouth is :)
[ link to this | view in thread ]
come on now
;]
[ link to this | view in thread ]
Re: why do the research when posters here already
***********************
Jeez, what a bunch of morons these folks are, spending their lives perform detailed analytical experiements, when you all already know for sure!
*****************
Ever had a computer say you didn't pay a bill when you had? Had Windows crash?
'Nuff said. We know it won't work because even HUMANS can't tell reliably so acting as though software can do this IS retarded.
[ link to this | view in thread ]
DID YOU EVER
[ link to this | view in thread ]
Re: DID YOU EVER
[ link to this | view in thread ]
Re: Re: Re: Re: Put Your Money Where You Mouth is
But a change in tense or "voice" is a recognized "tell" for liars in court. The place where this could be useful would be to analyze deposition transcripts looking for the changes in word choice and tense.
One of the classic examples of word choice giving away the truth is when a child goes missing: if the parent speaks of the child as already being dead, before the body is found, then the parent is probably implicated in the disappearance; while if the parent speaks in terms appropriate for a belief that the child is still alive then it is less likely that the parent is involved in the disappearance.
[ link to this | view in thread ]
Wouldn't work
For an email message, the filters would likely only catch off-the-cuff lies, where the sender has not bothered to think throught the lie. If he does think things through, then he is more likely to make sure his message matches a normal one in stlye and proofreading quality. This would make it very hard for an AI to find the lies, unless it had a very good plain-English handler and good "undrstanding" of semantics, enabling it to find inconsistencies between differrent messages.
[ link to this | view in thread ]