White Rose is a protest blog collective focusing on civil liberties in the UK.
It was set up to point a finger at the erosion of personal freedom in the UK.
Government's active measures introduce new means of control such as identity cards and surveillance cameras, the passive measures such as weakening of double jeopardy and presumption of innocence.


The arguments
The resistants
Gabriel Syme and Perry de Havilland of Samizdata.net to rally the Anglosphere behind the UK.
White Rose contributors are those bloggers and non-bloggers who oppose restrictions on personal liberties.

To find out how to become a White Rose contributor, please go here.
Friday, October 31, 2003
Lie detection software for phone conversations
Brian Micklethwait (London)
(0)   Technology

I did a posting yesterday on Transport Blog about how they're now using lie detection software to monitor phone conversations from insurance claimants, to flag up potential liars, and then "give them the opportunity to change their story". The result is a fall in insurance claims, and hence, presumably, potential cheaper car insurance.

I have a the overwhelming feeling that this procedure will bring bad news as well as good, in a White Rose Relevant way, when governments start using stuff like this for instance, as I dare say many have. But what form will this bad news take? I can't think of any obvious badnesses, but I feel sure there are some. Comments please.

One suggestion. The insurance companies mentioned in this story are all saying at the start of their conversations that "this call is being monitored", although I don't believe they say straight out that this means a lie detection machine. Clearly others will not be so scrupulous, and will simply monitor all conversations and flag up what the machines says are lies, all the time. What are the White Rose Relevant implications of that?

On the face of it, I think I have the right to buy a machine that helps me decide whether I trust someone at the far end of a phone line. I could simply say "Is this a junk phone call?" every time I suspect it is, and if they say no but my machine goes "ping", then down goes the phone. At present the danger is that with our own more fallible bullshit detection software that we all have in our brains, we do this to "real" phone callers who are merely a bit clumsy in identifying themselves, or whom we are a bit clumsy in identifying.

Presumably what makes this so much more usable now is that the kit has got a lot cheaper, and it supplies answers straight away, while the conversation is still going on.

Techo-food for thought here, I think.


Comments

well 1) lie detectors only detect when people think they are lying, so "is this a junk phonecall" will only work if the person on the other end thinks its junk... It's easier to just make up your mind and not have to ask.

2) They probably generate just as many false positives as anything else that claims to tell when people lie.

Since they aren't perfect, they'll only end up being determiners of action for groups that rely on statistics anyway (like insurance companies, something which gives them a better idea where to look will save them money, and they aren't interested in spending the money to get it perfect).

The only real danger is that it becomes a psuedo-scientific measurement of truthfulness and could be used as probable cause for a search if used by a police department. I would hope a statement that rang the bell would be taken as 'no statement' legally, but I'm worried that a judge might be willing to accept a "he's lied here, lets find out what else he's hiding" argument. (of course the police would look harder at it, that's prefectly fine, but as far as a search or arrest warrant goes it should be null)

Posted by: Michael on October 31, 2003 04:54 PM

As Michael says, they're far from perfect. "Lie detectors" don't detect lies, they detect stress. Making an insurance claim can be very stressful. An elderly person who's never made a claim in their life before is more likely to be stressed than someone who's done it several times and knows how to milk the system.

The insurance companies aren't stupid, they know that. I wonder how much of this is just an attempt to persuade people not to try it on? A bitf like having an empty burglar alarm box or speed camera.

Posted by: Trevor Mendham on October 31, 2003 07:47 PM

Do these "lie detectors" read minds? No, ergo they can't really detect lies. They detect stress. I wonder if anyone compared the procedure where they challenge people based on the stress these voice analysers detect with simply randomly challenging the same percentage of people. ISTM the reduction in claims may be due to people being intimidated and the chancers being deterred, not because lies are really detected but because they are being challenged.

As for "white rose relevance", ISTM that if such "lie detectors" are used to screen e.g. potential employees or people checking in to board planes, then they could have an adverse impact on people's freedom.

Posted by: James Hammerton on November 1, 2003 01:15 AM
Post a comment








    


    •