 Hi, I'm Giggs welcome to when electronics and deception collide a short talk about historical and modern uses of technology to mislead and deceive I'm the founder of pound pound electronics on free node. May she rest in peace We've moved over to libero chat if you haven't moved with us yet I've been a volunteer at the HHV for about eight years ish now Regarding this talk keep I might be wrong about something So what can be seen cannot be unseen. We'll obviously get a lot more into this later But there are elements to polygraph testing that don't really work right if a person being examined knows how they work If you need to do this for your job consider whether you should continue watching this or not That said everything in this presentation is public information. There shouldn't be anything secret You have been warned Related to the idea that knowing how a test works can ruin the test back in 2009 I used to edit Wikipedia. I noticed the debate raging on the back in discussion pages about the Rorschach inkblot test Regarding whether the images for the test which had fallen into the public domain should be published You may think that a test where you look at random images of inkblots wouldn't have the correct answers or incorrect answers But that's not the case the test actually has a set of sane answers and a set of insane answers And those are the incorrect and incorrect answers. I Wrote up a small blurb about the debate for slash dot which got posted on the front page in 2009 I expected that would be the end of it a Geeky story about slash dot or for slash dot about an argument within Wikipedia and that's it turns out That noam Cohen reporter for the New York Times also apparently read slash dot he picked up the story it caused the big media flurry There were morning news shows where they were broadcasting the actual Rorschach images where they were giving each other the test playing games Seeing if they would get the right answers or not Despite this leak in the strong evidence that the test is super scientific. It's still sometimes used in court cases So onto the actual topic of the talk electronic deception One of the earliest writings that documents electricity other than lightning is already ascribing powers to it that it doesn't have Plenty of the elder wrote extensively on natural remedies around 77 ad including the electric torpedo fish Which he said could numb the most vigorous arm and rivet the feet of the best runner He then goes on the list nine very questionable medical treatments based on the mysterious powers of this fish including one for precedence of the rectum Which apparently either means prolapse or hemorrhoids. It wasn't entirely clear and it's treated by applying the electric fish Topically, I don't know if they mean alive or ground up, but either way sounds pretty gross Skipping forward 1800 years or so not very long after the early work in harnessing electricity and magnetism had just been done One of the first non-scientific uses of electricity was in the stage show of Robert Huden His trick with the light and heavy chest was to invite a boy and a strong man from the audience Small boy could lift the chest while the strong man could not Relied on a strong electromagnet under the stage and metal steel plate in the bottom of the box And he would turn the electromagnet on and off to control whether the box was lighter heavy Most of the audiences at the time had no idea what electromagnetism was it had only really been discovered and documented 30 years before this Moving on to a slightly more controversial topic Nikola Tesla Tesla invented improved numerous things We still use today, but he was also intent especially toward the end of selling this idea of long-range wireless electricity He gave shows where he would hold lamps in strong electric fields And they would like despite not being connected to anything most of us have seen similar demonstrations of fluorescent tubes under The problem is the electric fields drop off with the inverse square of the distance despite Tesla's claims otherwise So was Tesla being deceptive? It's hard to say. He probably believed what he was saying That's the problem that runs deep when you look at these issues Separating the Charlotte in from the true believer is not always easy What we're about to look at are a few technologies that certainly have their share of true believers If you yourself happen if you yourself happen to be a true believer Don't take this as a personal attack, but maybe take a step back and evaluate whether your belief is misplaced or not So first a personal anecdote my uncle recently called the carpet cleaning service And they brought in these high-powered UV lights to detect cat urine They proceeded to say his entire house was covered in cat urine that they'd need to tear out all the carpets and replace At least the pads My mom showed me these pictures the thing is the UV light causes many things to fluoresce not just urine spilled food Lakewood detergents club soda lemon juice toothpaste. It's a pretty big list When you know that carpets pretty dirty, but maybe give the cat a break So what makes a good modern deception prop we can look at the field of ghost hunting for some easy examples It should be something that produces noisy fuzzy output with lots of artifacts preferably not totally random But with some kind of opaque internal functioning The vague property of being sciency helps lend credibility as well. People want to see patterns They want to confirm what they already believe These sorts of devices are tools of confirmation bias that let people find the patterns that they're looking for So you got a list here You can read it They all have the same sort of thing night vision video artifacts audio recorders Let's move on to the fMRI the functional f the functional MRI It's a tool that was developed in the 1990s and measures blood oxygenation in the brain. That's all it does Lot oxygenation is somewhat decent proxy for telling us what areas of the brain are being more active But that's about where the good parts and it collects an image of between 130 and a hundred and sixty thousand voxels with noise inherent in the capture. It is time-based blurring Which means that when a part of the brain demands more oxygen it takes several seconds to ramp up the detection on the MRI When the demand of blood goes down it keeps showing up for a while as well Head movements can throw it off in spatial resolution But the main problem is that it's generating a large and noisy data set with over a hundred thousand potential correlations P hacking is the term given to the idea that if you try enough independent tasks You'll eventually find a positive answer if your p-value threshold is 5% and you try something 20 times You'd be expected to find one false positive so This is a fish in an fMRI notably it's a dead fish a salmon Craig Bennett of UC Santa Barbara was frustrated with the situation of fMRI studies being released that we're not doing multiple Comparison corrections, which are mathematical techniques don't negate the inherent p-hacking that fMRI correlations give So he went to the store bought a dead salmon put it in the fMRI and then asked the dead salmon to explain what emotions were being depicted in the pictures that he showed to the dead salmon The fMRI did not fail to show the response of the dead fish. You can see the voxels lighting up there So the study making fun of the lack of use fixed fMRI, right? well Fast forward to 2016 Anders Eklund ran some tests on large sets of control data But the p-value of 05 5% false positives were expected instead. He found false positive rates as high as 70% Turns out the normal the noisy data with opaque software still wasn't working, right a Staggering 40,000 fMRI studies were potentially used use the software that he tested and could be made very easily To show positive results when none existed. This was on the heels of a year prior when a separate study Failed to replicate 50% of the published studies and psychology that they examined Obviously, I'm not accusing all the scientists involved in malpractice But the drive to find a positive result is a strong one and what the science is used for is not always benign These sorts of studies are used and abused to advance policy arguments Even in some cases directly in politics with studies saying things like this political group So there's more activity in their brain and this part of the brain than that political group and things like that The sheen of science and hard data lends credibility to these uses and that may not always be deserved So this is this year's tour badge CS and I built it for the tour project It's a mini polygraph with two channels galvanic skin response and heart rate Someone on Twitter said how can you sell this without warning people that polygraph is bullshit? Well, I'm here getting the stalks So you've been warned it's bullshit So about the polygraph Here's some stuff. You probably already know it's rolled in and missable in court Which is unlike the Rorschach test, which is apparently still mostly admissible measures several body bodily parameters that in theory are correlated with stress the APA among other groups say there's little Evidence that polygraph can accurately detect lies The inventors of the polygraph are more interesting than the machine itself Nard is on the left. His name was Leonard Keeler, but he preferred being called Nard He was into stage magic on the right is John Larson. He was the first cop in the US ever to hold a PhD which he got in physiology and 1921 these guys work together to invent the polygraph. It's a fitting invention a stage musician and someone with a PhD in medical field Both become cops and invent the thing. That's the intersection of all those things It is interesting to note that the John Larson there he he said later on that he regretted ever inventing the polygraph that it was it was abused for In ways that he never intended originally So here are the major categories of testing that are used with polygraph the relevant Irrelevant control question test guilty knowledge has peak of tension test They all have some traits in common. They all mostly work because the subject believes that they work It's the main job of the examiner to convince the subject that they work All of the tests are biased against the innocent. We'll talk more about that later and For in the results for the most part involves subjective judgment The oldest and simplest test is probably the conception of the lay public has for how polygraph works Some irrelevant questions are asked and they're known to be irrelevant by the examiner and the subject Didn't the relevant questions are asked this version has the most subjective interpretation of the results It varied so much from examiner to examiner that it was impossible to compare anything This version of the test is rarely rarely used today The control question test this is the big one in the us and has been for a long time This one works in a slightly more nuanced way the pre interview is conducted where the examiner goes over every question with the subject The questions might be something like What is your name if you ever stolen anything? Were you at home on the night of the bank robbery? The examiner will say The what is your name question is the control question, but it's not the control question. It's an irrelevant question The true control question is have you ever stolen anything? You may confess to a few things like stealing pens from work or a candy bar when you were a kid or whatever It doesn't really matter how much you confess So then during the test they changed the question slightly to other than what you've already told me Have you ever stolen anything? The premise is that it's impossible to fully confess to the question It's a probable lie question one that forces you into lying even if you don't intend to The probable lie questions established the baseline for what a lie looks like for that person on the graph The problem with this is that the cleaner your conscience is the more likely you are to fail the test I've written an example where this was used Where this particular question was used as the probable lie question And here it is other than what you've told me. Have you ever done any illegal drugs even once? The subject failed the polygraph because they were a super straight-edge person who could with great confidence answer that that question Which set their lie baseline really low That's just one way the test is biased against the innocent The stem test mentioned here on the slide is just more theater The examiner instructs the subject to lie on a question and That they tell the subject that that's establishing the baseline, but it really isn't We all know at this point that that's going to be treated as an irrelevant question The probable lie question is the real baseline the subject answers and then the examiner says Yeah, yeah, look like you can see here. This this machine can tell you're lying But it's all theater to convince the subject that the machine works So that that's the stem test So is there anything better? Is there a better methodology? Yeah, but only slightly so In japan they use a test called the guilty knowledge test in this test often the examiner isn't even aware of the facts of the case on purpose Is much less involved in the process the examiner may still use some theater to convince the subject to the validity of the test, but Questions are all multiple choice The subject is presented with the question and answers Which includes the correct answer? That's a fact of the crime That's unknown to the public and several in is several incorrect answers The idea is that the subject will show a greater response when the correct choice is presented And that this response will correlate over a large number of questions This test is still biased against the innocent since the guilty can seek out knowledge on how the test works And ensure that their response is less to the correct answer than it is to the other answers That the guilty know what the correct answers are after all Related to the guilty knowledge test is the peak of tension test This is often used with sequential quantities such as was the amount stolen one two three or four thousand The idea is that the tension builds up as the correct as the correct answer is approached and then will release once it's surpassed So what goes up must come down? So that's it for the polygraph This is far from a complete discussion on the topic in this short talk But it should give you some idea of the deception that underlies the machine itself On this slide. I put some honorable mentions You may or may not recognize the machine on the left. It's a Scientology emeter The glorified love tester that tells you about your Painful emotion angrams and the invisible aliens that are attached to you And on the right is DD Palmer the founder of chiropractic That's another field where the technology is used to give a sheen of credibility to the underlying pseudoscience So the next time you're using an electronic tool question if you're being deceived either by someone else or by yourself I'm gigs and you can probably find me in the hard-racking village if you want to discuss Enjoy the rest of the f-con And I apologize for the bug zapper. Lots of moths tonight