Lie to Me! or: Building a Polygraph
In my view: the world needs more engineers to solve all the problems.
Question: How to recruit future engineers?
Answer: Show young people that engineering is a lot of fun!
It is probably the same thing in many countries: universities and companies have challenges to attract enough (and good) engineers. Maybe young people think engineering is just about math or something else they believe is boring? Or because they think engineering is only about machines and not about feelings or how humans interact?
I think it is important to show them that engineering is about creativity, or using existing things and invent something new out of it. It is about collaboration, about human aspects, and it has relevance to the society. And not to forget: it is about having fun .
To bridge the gap between perception and reality, the Swiss Academy of Engineering Sciences is organizing special events in local schools to motivate young people for science and technology. Together with Professor Rolf Mettler, we had the pleasure to contribute to multiple such events. I thought why not showing here what we are doing?
How does such an event work? The school is organizing several workshops, with different engineering and science topics. From ‘Are mobile phones dangerous’ to ‘Can solar system save the world’. And the students can pick what is of interest. Our topic was “Seeking the Truth”. Does not sound very technical, right? It is about what is true or false, about lying or saying the truth (and all the gray zones in between).
The question is: Is it possible to measure the truth? How reliable would that be? Students have to investigate this and build a ‘lie detector’ to answer those questions.
For this, we set up an experiment, which is very simple: Money has been stolen! We know that there are suspects, but we do not know who did it. The students have to build a Polygraph (Lie Detector) and have to test who is guilty or not. With this, they can learn a lot about engineering and science: From building a claim, to think about how to prove that claim, to measure and rate things, up to coming to a conclusion. And of course to learn new things in the lab.
Typically we can take up to about 24 students for the lab. We divided them into groups of 5-6 people. Out of each group, a ‘suspect’ is selected. Each suspect can decide if he wants to take some money. For this, outside of the room there is a closed box with real money. Each suspect leaves the room, one after each other, and could open the box and remove money. Nobody in the room knows who did it, and it could be that there are multiple offender.
Is a Truth or Lie making a Difference?
Now it is up to each group to test their suspect, if he took the money or not. Each group has to think about how to ‘measure’ the difference between lying and saying the truth: can it cause physiological effects? Can it be measured or observed? How? What takes more efforts (lie or truth)? The claim is that lying is not easy, it creates ‘work’ for you, and you start to sweat, you heart rate goes up, etc.
To make the experiments ‘real’, as instructors we had to increase the pressure on the students: If the ‘police’ correctly identifies the ‘criminal’, the ‘police’ gets rewarded with confiscated money and the ‘criminal’ goes to ‘Jail’ for four weeks.. If the ‘police’ wrongly accuses as suspect, then the innocent person has to go to ‘jail’ for a week (until identified as ‘innocent’), and the ‘police’ has to go to ‘jail’ for two weeks (‘bad press’). Of course there is no real ‘jail’: Instead they have to help the school Janitor to clean the toilets for that amount of weeks!
The ‘Janitor Jail’ was only a psychological trick to put pressure on the students in the experiment. It should not be an ‘easy’ thing for the innocent suspects: they need to be worried. And the guilty suspects need to fear the ‘Jail’ too. They are informed about this at the end, and that this ‘jail’ is not real. But the motivation with the money is real: the can keep the money! .
The Question List
Each group has to build a question list. And as with real polygraph tests, the questions are discussed with the suspect. They learn, that the way how things are asked, it has a big impact on the results. They learn about ‘irrelevant‘, ‘relevant‘ and ‘control‘ questions, that there is only ‘yes’ and ‘no’ as answer, and that the answers are already known (!!!):
- Irrelevant: To allow the body to calm down between questions and at the start. They are not relevant to what we are investigating. Example: “Are you 36 years old?”. The answer is always ‘yes’.
- Relevant: They are directly about the topic investigated, and could be about something only the offender knows about. Example: “Do you know how much mony was in the box?”, and the answer is always ‘no’.
- Control: These questions are usually the hardest to formulate. They are not directly related to the topic, but in the same domain. Example: “Have you ever stolen sweets from your brother after the age of 10 years?”, and the answer has to be the truth and ‘yes’.
The idea behind control questions is that an offender does not much care about them (because the focus is on lying to the relevant questions). The innocent suspect however really wants to answer the control questions honestly (he knows that he is innocent, so he does not focus on the relevant questions). We tell the students this subtle difference after the measuring phase.
The questions are asked with enough time between, and they are put in sequence: after a relevant question, there is a control question, then again a relevant question, and so on. In between, an irrelevant question can be added to allow the suspect to ‘normalize’:
The expectation is that for each kind of question, the reaction will be different, depending on if the suspect is is guilty or not:
So it is less about if the suspect says the truth or not: it is more about how he reacts to the different kind of questions. And with the sequence and kind of questions, there should be a different pattern of up and downs.
Next step is to build the measurement ‘engine’. Each team has to build a PolyGraph (old Greek: Poly=Multi, Graph=Drawing). They have several instruments available: multimeters, medical electrodes, temperature sensors, heart rate monitors, or whatever is available otherwise. As a highlight, we had boards, development tools and software to build an ECG (Electro-Cardio-Graph) available:
Three medical ECG electrodes are aheared to the chest of each ‘suspect’:
Students can place electrodes, combine boards with a DSC (Digial Signal Controller), a MCF51MM256 microprocessor and a touch screen LCD to build a simple ECG and Heart Rate Monitor:
The software is built with the Eclipse based CodeWarrior. Using Processor Expert graphical UI Library it is easy to build a system in a very short time.
Another approach used was to measure skin temperature. For this, a K Thermocouple element gets attached to the skin. The claim is that you cannot control the temperature of your nose, and there could be an effect. The following picture shows such a sensor attached to a ‘suspect’:
Another measurement used was to measure the skin resistance with a multimeter, assuming that sweatening will lower the resistance:
But they used other ways to measure or see the difference between a lie or the truth:
- Voice: keep your eyes closed, just listening to the answers. Does it sound differently? How much?
- Body Language: any movements? What is the person doing? Nervous by which scale? How cooperative is the person?
- Facial color: getting red? or white? What is it telling to you?
- Eyes: Can you see inside the soul? What are the eyes telling you? Can the person look into your eyes or not?
Of course there are many more ways. There is the American crime drama television series “Lie to me” which is an inspirational source. The challenges with these is: putting things into wording or numbers, so it is ‘comparable’. If you are a parent, then you know that these kind of things can be strong indicators . But it requires good observation skills:
To avoid an undesired feedback, the suspect was not able to read the measured values collected:
Measurement is performed several times to get a good number of data samples. But before starting to ask the ‘real’ questions, a ‘calibration’ needs to be performed. For this, the ‘suspect’ draws a random number (e.g. from 1 to 9). Then he is asked ‘is it number 1?’ and the suspect has to deny. And so on. Typically after one or two rounds the number is identified. That way we know that things are working, and it is a good training exercise.
‘Calibration’ has a psychological effect too: with this everyone nows ‘that this machine indeed works!’. There are some tricks possible to make the effect a bit stronger.
Now a lot of data has been collected. Is the data correct and concise? Is the date collected correctly? Does everyone see the same pattern? Here the group needs to come to a consensus: ‘guilty’ or ‘not guilty’, and need to give an indication how sure they are. This part includes interpretation of the data, as things usually are not very clear. It is interesting to see how much some groups come to a conclusion, based on which ‘facts’ or ‘soft facts’. Or that measurements with the ‘technical’ methods are contradictory to the ‘soft’ factors.
We did the above experiment in several schools, with multiple groups. In some cases, all groups 100% correctly identifying if the person was lying or telling the truth. Impressive! In other cases it was between 30% to 60%. So very mixed results. Well, this was a ‘self-built Polygraph’, and does not compare to a professional one with trained person. But regardless, the results where interesting. And I think it gave everyone a sense of what is important, not only from an engineering perspective:
- Without asking the right questions, you will not get the right answers.
- If you only trust the sensor values, you might be way off.
- If you only trust your gut feeling, that might work, or you can be fooled too.
- If you want to measure something, you better ask first what *exactly* you want to measure, and why.
- If you measure something, make sure that you can reproduce the result, and things are well documented.
- A single measurement does not say anything: you need to sample multiple times.
- Without calibration, you do not have a stable base.
- Measurement means measuring the noise with the signal.
- The temperature has a key impact on the measured value.
- And: Polygraphs are very controversal (and banned by law in many countries).
As for myself: I have two daughters in the teenager age. I have my own methods. And no, I’m not using a Polygraph . I admit: I’m not always right, and sometimes the truth is unknown. But that keeps life interesting, or not?