FAQ on Computerized Knowledge Assessment (CKA)
Written by
Brendan Allison, MS, PhD student
Brain computer interface project director
Cognitive Neuroscience Laboratory
Dept. of Cognitive Science
UC San Diego
Note: I do not work with Dr. Farwell, have never met him, and do not stand to benefit in any way from the success of his company. I do not have any involvement in any company involved in brainwave fingerprinting. My comments are inspired only by my firm belief that this technology needs to be discussed openly and intelligently.
Recently, much attention has been drawn to a new technology proposed to fight crime and terrorism. "Brainwave fingerprinting," [which we call CKA] so dubbed by its main proponent, Dr. Larry Farwell of Brainwavescience.com, is meant as a tool which can be used to determine whether individuals have seen certain information which may indicate they are guilty of a crime. I have seen numerous articles on the popular media appear recently, some containing severe distortions of the facts, and have also seen some excellent questions about the technology which I would like to answer.
This essay is divided into four parts:
1) An overview of how the technology works, written for the intelligent layperson with no scientific background 2) Commentary on what might or might not constitute appropriate use of this technology 3) Information about the author and a couple relevant disclaimers 4) Replies to FAQ's
1) About the technology
To describe how brainwave fingerprinting works, I first must say a few things about how the human EEG (encephalogram, or brainwaves) works. The brain is essentially an electrical system. Anything you think, imagine, sense, feel, or otherwise experience produces changes in your brain's electrical activity. Some of these changes can be reliably detected by an electrode placed on the scalp. An electrode cap is just an elastic cap with several electrodes on specific regions of the head.
One way to study the EEG is to present a subject with a specific event, such as a tone or image, and study the brain's response to that event. This response is called the event related potential or ERP. ERPs are very well studied, and we have a pretty good picture of how the brain will respond to (for example) a picture. Many aspects of the ERP are sensitive to factors such as age, attention, or memory. For example, if you present someone with an image, and she is paying attention to that image, you will see a different ERP than if she were ignoring that image.
Another well studied observation is that at least one component of the ERP reflects memory. It looks different for familiar images vs. unfamiliar images. If you present a picture of the inside of an aircraft engine to me (I know nothing about aircraft engines), my brain would process it as an unfamiliar image, and that would be visible in my brainwaves. If you presented the same picture to an aircraft mechanic, his brain would recognize it as a familiar image.
The "brainwave fingerprinting" approach simply consists of showing people several images to see if their brain classifies them as familiar or unfamiliar. In a court environment, it can be used to determine whether someone has "guilty knowledge" that only the guilty party would have. For example, if someone is accused of robbing a bank, and he insists that he has never been inside that bank, he could be shown several images. These images would include pictures of the inside of the bank, the teller who was robbed, etc. If the man's brainwaves suggest that those pictures are unfamiliar, then he clearly has not been to the crime scene. On the other hand, if he recognizes the inside of the bank despite his claim he has never been there, this information should also be considered in determining his guilt or innocence.
In the case of counterterrorism, the images presented might be those of the inside of a terrorist training camp or a timing device on a bomb. People who have never been inside a terrorist base would process such an image as unfamiliar. People who have been inside one cannot help but recognize the picture as familiar. Such individuals could then be investigated further.
It must be emphasized that this technology is not a means of determining guilt. I would be vehemently opposed to a computerized system to judge guilt or innocence, and imagine most people would agree. This system merely provides information to investigators which they can use to judge someone's guilt or innocence. It is analogous to the use of regular fingerprints. If fingerprints were found at the aforementioned bank, and they matched the accused man who insisted he had never been inside the bank, this evidence would certainly be of interest to those judging his case. It does not, by itself, prove the man is guilty or innocent.
It must also be emphasized that we cannot detect anything and everything going on in your brain or mind with the EEG. The observation that it reliably reflects familiarity with images has been very, very well documented over decades in numerous different labs. However, the substantial majority of what goes on in your brain and mind is invisible to the EEG, for a variety of reasons. We cannot determine if you are thinking about one concept or another without presenting an image. We could put an electrode cap on you, and ask you to think about a tree or a car, and we could have no clue which one it was. We cannot determine emotion. We can determine if you have schizophrenia, but otherwise can learn almost nothing about your personality. We cannot determine what word you have in mind. We cannot tell who you are thinking about. We cannot determine what you are thinking about doing, your plans, hopes, or dreams. Perhaps we will be able to do so someday, probably with the use of technologies other than the EEG, but people have been trying for a long time with no luck, and in my opinion none of these constraints (which is only a partial list) are likely to change in the near future.
2) About how the technology could be used
I have been misinterpreted as believing that any use of brain scanning technology is good, that I have not thought about the ethical and practical issues of using it, or that I am confident that government and big business can and will use this responsibly with proper regard for individual civil liberties. I wish to make very clear that none of these are true, and I challenge anyone to present any of my writings or statements which state otherwise. On the contrary, I am very concerned about the possible misuse of brainwave fingerprinting and related technologies. I am quite aware that the classic freedom vs. security debate is often clouded by naïve proponents of technology who do not consider how it could be abused. It is my opinion that advances in science and engineering have provided us with options never before seen, and that the best way for us as a society to ensure that technology is not misused is for us to engage in informed debate and discussion on the subject. Part of what motivated me to write this is the observation that many authors and their readers are expressing either support for or fear of this technology based on incorrect assumptions.
Here are some straightforward statements about my beliefs:
· The use of brainwave fingerprinting should be an option. People should be
free to choose an alternate approach, which may or may not be easier, faster, or
more invasive.
SK: Unless we have an equally reliable method of screening passengers on the
plane that has the same accuracy as this technique, I do not think this should
be an option and I would be very suspicious of traveling on a plane where this
were not required. If you want to create "safe" planes (100%
passengers are tested) and "unsafe" planes (traditional screening)
that's fine with me. I'll ride in the "safe" planes. Know anyone who
wants to ride with the terrorists?
· If brainwaves are acquired for the explicit purpose of using them to determine whether someone has seen an image, they should not be used for any other purpose. This should constitute an illegal search. If you are told that someone is going to see if you have seen the inside of an Al Quada camp, and after they test you are told, "Ha! According to this, you had marijuana recently, and you're under arrest!" the case should be thrown out and the investigator punished.
· Any brainwaves acquired to detect deception should remain anonymous to the maximum extent possible.
· People should never be expected to take this test on trivial grounds. It should not be used for employee screening or monitoring, for example, with some very specific exceptions in high stakes circumstances with specific personnel which I outline below in the FAQ section.
· This system should be considered a supplementary means of finding the truth. It should not be used as the sole means of ensuring airline safety, nor should anyone be convicted or exonerated of a crime based only on brainwave fingerprinting evidence.
· It is my opinion that further research and development is necessary before
the system is appropriate for widespread implementation as an antiterrorist
measure. It could be ready in the very near future. It is, at this time,
appropriate for forensic deception detection under the specific circumstances
for which it was designed.
SK: A prototype to detect terrorists can be put together in 90 days. A full
system might take 2 years or a bit longer.
3) About the author
This essay is written by Brendan Allison, a Ph. D. student in the Cognitive Science department at UC San Diego. I earned my BS in Cognitive Science from UC San Diego with high honors in 1994, worked for Nextel for a year, and then went back to grad school in the same department where I remain. I earned my MS in Cognitive Science in 1997. I work in the Cognitive Neuroscience Laboratory, where I am director of the brain computer interface (BCI) research group. I have always been fascinated by the notion that people could communicate using only the EEG, and have two different systems which do exactly that in my lab. I became interested in this research because I have a very bright cousin with cerebral palsy who can benefit greatly from such a system, and also because I always wanted to play a computer game with my brainwaves. So far, we have some significant success to report. Our system has allowed about 30 different subjects to spell using only their EEG - much more slowly than typing, about 8 characters a minute, but far better than nothing. We have brainwave pong in the lab already and have discussed interfacing our system with Diablo II, a bestselling computer game from Blizzard Entertainment. We are funded by grants from the Natl. Institutes of Health to explore communication systems for the severely disabled. I am very proud to be involved in a research field which has already empowered people with severe injuries, Lou Gehrig's disease, and other disabilities which prevent them from communicating via conventional approaches such as typing, writing, or talking. More information on my work and links to many related scientists' labs is available from my website at bci.ucsd.edu. The views expressed here are mine alone and do not necessarily reflect the views of the lab, department, university, or scientific community. I also emphasize that I do not work with Dr. Farwell or Brainwave Science. I am not affiliated with any business or commercial entity involved in deception detection. I do not stand to profit financially from the acceptance or rejection of brainwave fingerprinting. I am motivated by my interest in seeing this technology discussed intelligently and by my concerns about ensuring it is used in a safe, productive, and noninvasive manner.
4) Frequently asked questions
This seems pretty far out to me. How do we know you are not just making this up? Has this technique been verified experimentally? Has it been tested in different labs? Has it been reviewed by objective scientists and others?
(This is the most FAQ, and rarely phrased so politely.) The technique has been verified experimentally in dozens of published studies from numerous different labs. The studies were published in numerous reputable, peer reviewed journals. Peer reviewed means that the paper which you submit is reviewed by a panel of scientists with no personal stake in you or your work. The number of scientists who have reviewed this approach numbers in the hundreds; more importantly, I don't know anyone who has produced a paper stating that the ERP components which Dr. Farwell uses do not reflect memory. In addition to the scientific community, Dr. Farwell's approach has been accepted by the FBI, CIA, and the legal system. Dr. Farwell has refined his approach over several years with the first two organizations and published a paper on it with an FBI agent this year. Evidence from his brainwave fingerprinting system was ruled admissible in court. I am not a legal expert, but I am sure that a new type of court evidence would not be accepted in a murder case without a substantial and very conservative review.
Is this the same as conventional polygraphy? Can people fool it by just acting nervous in response to every question?
No. This system is very different from conventional polygraphy. It is not based on classic physiological signs of nervousness such as increased heartbeat or sweating. Someone cannot fool the system by being more or less nervous, or by placing themselves in any emotional state.
Has this been tested on intelligent people? What about people trying to fool the system?
The system was tested extensively on FBI agents who were themselves trying to fool the system and failed. Dr. Farwell published an article in 1991 with Dr. Donchin which very clearly showed that his approach works even if subjects are trying to fool his system.
Which brainwave is involved?
The system uses an event related potential (ERP) called the P300. The P300 is one of the most thoroughly studied brainwaves in the literature. ERPs are named according to whether they are a positive or negative change in voltage (either P or N), followed by the delay between when the stimulus is presented and the peak of the resulting brainwave. Thus, the P300 is a positive shift in voltage which typically peaks 300 ms after an image is presented. (In fact, the P300 usually peaks after 300 ms, but people still refer to it as the P300.)
An excellent review paper on the P300 is: Polich, J. P300 clinical utility and control of variability. Journal of Clinical Neurophysiology 1998 Jan, 15(1):14-33. In fact, Dr. Polich has many articles on clinical factors which can affect the P300. Just do a search through any good library for articles he published and you will find them.
One reader asked a very good question: This is an interesting idea. I am curious if the machines during profiling, could determine the difference between having read material on a given subject, and having actually performed the task. For instance, in the wake of the WTC tragedy, many of us have immersed ourselves with information ranging from self defense, to the various chemical agents that could be used in a bio attack. Many have felt threatened by unknown future attacks, and have perhaps bought a gun, gone to a shooting range, read articles on how to create simple weapons out of common household things, the list could go on and on. How would this computer screening know the difference?
I replied: The brainwaves involved (called the P300 complex) reflect your familiarity with specific objects more than general categories. Another aspect of them which has not been as widely presented is that they also reflect the object's importance to you. (The jargon used among scientists is that they index stimulus novelty and saliency.) Thus, if you have a pistol or hunting rifle at home, and are shown a picture of an AK-47, you may get a slightly different response than someone who has never seen any gun, but you will certainly not get the same response as an individual who is very familiar with that weapon. True, someone who is a big fan of AK-47s and looks at pictures of them every day but has never actually used one may produce a false positive in the system. Here is how you could separate those people from individuals who are trained with that weapon. In addition to showing a picture of an AK-47 fully assembled, you could also show it in various states of disassembly which are meaningless to a casual observer. For example, you could show the weapon which has been disassembled for cleaning, or show it being unjammed. You can then mix in some bogus pictures of the AK in these two states, such as a disassembled weapon with some pieces missing, or extra parts, or the weapon in some weird configuration. Only someone who knows the weapon and how to use it would show a distinct brainwave response to the correct picture of it being cleaned or unjammed. Similarly, while many people have seen pictures of shooting ranges, few would legitimately be familiar with the shooting range at a terrorist training camp. Many people have heard of concerns with crop dusters, and have seen pictures of them, but fewer people would recognize the inside of one or pages from the inside of a training manual. I would assume that the pictures which people would see would reflect a variety of different images only a terrorist would consider familiar. It's true that there are people who are experts with AK-47s and are not terrorists, but such people (if innocent) would not also recognize the inside of a training camp as well as pictures of senior members of a terrorist network as well as pictures of the Khyber Pass taken from specific locations inside Afghanistan held by the Taliban.
It seems that this system only tests a very specific type of memory. Is there any research suggesting it could be used for anything other than presenting images?
We all have many different memory systems which use different areas of the brain. For example, we know that declarative memory- memory for objects, for example- is different from procedural memory- memory of how to do a task. The P300 complex reflects declarative memory. A brainwave fingerprinting system could be enhanced by utilizing multiple memory systems. People who have not only seen an AK but have also used it extensively will not only remember what it looks like, they will remember what it feels like and what their fingers have to do while using it. Thus, such people would show different activity in their primary motor cortex and primary somatosensory (touch) cortex. This would beautifully distinguish those who see pictures of an AK from those who have been trained with it. Yet another form of memory is priming. If I say "king" to you, your brain shows a certain response. However, if I say "king" to you after saying a related word, such as "queen" or "throne" or "crown," you will show a different response. This is because your language system cannot help but think of words related to a certain word, even if you aren't conscious of it. Here is how this could be incorporated into a brainwave fingerprinting type of approach. Let's say that the director of a terrorist training camp is named Bob Smith (not a viable pseudonym in Afghanistan, but let's pretend). If I show an innocent person a picture of Bob Smith, and then a picture of bin Laden, s/he would recognize only the latter. A guilty person would recognize both. So far, this has nothing to do with priming; it's using the P300 complex I mentioned above for object familiarity. However, a guilty person would also have been "primed" for the bin Laden picture, since he would know that Bob Smith was related to bin Laden. So the guilty person would be different in 2 ways. First, he would recognize Bob Smith, and second, he would recognize that Bob Smith and Osama are related. You are thus tapping into 2 different systems which could back each other up. Do EEGs require injections or other painful procedures? No. A brainwave scan merely requires that you place two or more electrodes on your head. You don't need a special magnetic field, needles, drugs, etc. An EEG collection system can weigh less than a pound, fit in your pocket, and cost under $100. We have an electrode headband which my advisor built here in the lab meeting those characteristics.
Why use the EEG, and not any other means of imaging the brain? The EEG, at this time, is the only system which is very portable, requires little training to set up and none to use, and does not require expensive equipment, invasive procedures, or drugs. A PET scan, for example, requires millions of dollars of equipment, highly trained technicians, and an injection of radioactive material. A functional MRI scan, which is a very powerful imaging technique, also requires very expensive equipment and trained technicians plus an extremely strong magnetic field. (The weaker MRIs use a magnetic field of 15,000 Gauss, or 1.5 Tesla.) Dialysis and single unit recording all require penetrating the skull. It is my opinion that the EEG will be the only safe, portable, easy, noninvasive means of imaging the brain for at least five years. It also has the advantage of being a very old technique and thus very well studied. The alpha rhythm was first published by Berger in 1929; the P300 was first published by Sutton in 1965.
So you have a system which is scanning lots of peoples brainwaves to see if they have certain "guilty knowledge." Are there any other positive side effects which could arise from the widespread implementation of such a system?
I said above that I believe that it would be unethical to ask someone to undergo a brain scan to detect deception, then use that data for something else. The sole exception I would consider acceptable is that, if the subject asks it to be done, the data could be reviewed for irregularities in the EEG which might indicate a medical problem. The EEG can reveal a wide variety of disorders, including various types of brain damage, Alzheimer's disease, schizophrenia from very early stages, and attentional disorders. While I am very concerned about people analyzing brainwaves for anything other than deception detection, it would be a shame if a potential positive side effect - providing needed medical advice to people which could prevent a problem from worsening- were ignored. I would still feel that any data acquired could not be used against the subject in any way, and the subject should be told of any medical concerns anonymously. This should not be relied upon as an ironclad means of ensuring one's health. That is, people should be told very clearly that their brainwave scan does not replace a regular visit to a doctor, and that any concerns raised by the examiner should be checked with a doctor.
So you have a system which is scanning lots of peoples brainwaves to see if they have certain "guilty knowledge." What is to stop the government from misusing this information?
An informed populace. As with any new technology, there is always a wave of disbelief and those who either wave it off or express views on it based on incorrect assumptions. Those who believe that we should not have an informed discussion about the ethics of brain computer interfaces are like those who don't want to talk about bioethics because they find cloning scary and confusing. Those who naively endorse the technology without any restrictions are as bad as those who naively insist it be avoided completely. Of course, it is fine to take either position, or anywhere in the middle, as long as people learn a bit more about the technology and avoid basing their arguments on incorrect conclusions. I wrote this essay to minimize misinformation, and I encourage those who want to learn more to read Dr. Farwell's web page, which includes a thorough list of relevant references, check out some other sources in my web page, or visit any good university or medical library if possible. I do not believe that this sort of technology necessarily leads to abuse. To borrow an example from Steve Kirsch from Propel Software, El Al has been grilling airline passengers for decades as a counterterrorism measure. This approach has been quite effective. A skeptic might have said, "Wait a second - if you give the government this kind of power, the next thing you know they'll be asking you about your sex life and other topic irrelevant to terrorism." They don't.
Should the government, or any other entity, require all Americans to take a brainwave scan and maintain a database of people's brainwaves?
No. I see no legitimate justification for this.
Could brainwave scanning be used to screen or monitor employees? Dr. Farwell's system could not. However, it is possible to learn certain information from the EEG, such as drug and alcohol consumption and whether or not someone is paying attention to a task, which could be of interest to big business. I would strongly oppose any such use of the technology. There is always a trade-off between freedom and security; in almost all cases, the invasiveness of such a system outweighs the benefits. There are a few exceptions which I would consider acceptable. For example, the US Naval Health Research Center and other government entities have funded the development of a system to monitor alertness levels in sonar technicians, nuclear plant operators, and other personnel in which a lapse in attention could have disastrous consequences. I think this would be an appropriate use of BCI technology, since these personnel are already subjected to an extensive security screening and monitoring process, and there are tremendous risks associated with a lapse in alertness. I also think that it would be acceptable to monitor these people's brainwaves for drug or alcohol abuse. Of course these people are free to drink on their own time, but I would certainly like to know if a nuclear plant operator is drunk on the job or has a substance abuse problem. Note that these people already are screened for these anyway. A brainwave system is not adding any new information, it is just an alternate approach to (for example) a blood or urine test.
There are other situations in which I think the ethics of a BCI are murky. I wish to emphasize that these are NOT things you could do with brainwave fingerprinting, but are feasible with other brainwave based technologies. I present them as points to ponder, and because I hope that an informed dialog about them will help clarify whether they are appropriate.
Situation #1) Would it be appropriate to require someone convicted of drunk driving to wear such a system to ensure that s/he is not drunk when operating a vehicle? Right now, some people with DUI convictions are only allowed to drive if their vehicle has a govt. approved breathalyzer attached to the ignition. They must blow into it in order to start their car. It seems no more invasive to instead require them to put on a cap or headband which is plugged in to the ignition. The brainwave based approach has a few advantages over a breath based approach. It can monitor someone continuously, whereas the breath based system only requires the person to pass the test once. The driver might get drunk after starting the vehicle with a breath based system. Also, a brainwave system could not be fooled by a friend breathing into it, since brainwaves are unique. Third, since the brainwave system would keep a running record of the driver's EEG, this could be used to help convict or free that driver as it could be reviewed later in court to confirm or refute his claim of sobriety. I certainly don't think that any driver should be required to wear one of these, but it is appropriate for those who have already been found guilty of a DUI. When you drive drunk, or commit any other serious crime, you have to accept that you may be subjected to additional scrutiny to ensure you don't do it again. An old friend of mine from college was killed last year by a drunk driver. The driver had repeat convictions for a DUI and (as the accident clearly showed) should not have been allowed to drive again. However, if people with repeat DUIs are allowed to drive, they should be subjected to substantially increased scrutiny. If you don't like it, don't drive drunk.
Situation #2) If someone has a prior conviction for drug offenses, and is released on parole, would it be acceptable to require him or her to undergo a brainwave test to ensure s/he has not started taking drugs again? Many people convicted of drug use are indeed required to undergo some kind of test to make sure they are not taking drugs again, such as a urine or blood test. The use of a brainwave based system is simply another option for how to administer the test.
Situation #3) Should key government officials, such as congressmen, be required to take a lie detector test to ensure they are acting in the best interests of their constituencies and are not overly influenced by special interests? I am not sure this is technologically possible. If it were, I think it would be tremendous fun to publicly challenge politicians to take the brainwave challenge, but it'll never happen. I would probably publicly oppose such use anyway as I don't want to be found the next day with 535 bullets in me.
Situation #4) Do parents have the option of requiring their children (under 18) to undergo a brainwave scan to look for possible medical problems? Yes, because this is just the same as a parent forcing an unwilling child to undergo a conventional exam by an MD. The EEG can reveal a variety of disorders and could be of tremendous benefit. I do not think that anyone should be required to undergo a brain scan to look for medical problems in any case other than the one I just stated, since parents are already allowed to make medical decisions for their children.
A note from Larry Farwell about CKA
Dear Steve:
Your points are excellent. On the issue of whether CKA is well established science or not: The judge in the Harrington case ruled that CKA was sufficiently reliable and well established science to be admissible in court. In order for a judge to rule that CKA is admissible in court, we had to prove four things regarding the science on which it is based: 1) It has been scientifically tested; 2) It has been peer reviewed and published; 3) It is accurate; 4) It is well established and accepted in the scientific community. (The relevant scientific community, in this case, is psychophysiologists, and not cognitive scientists, who are in a totally separate realm.)
The evidence considered in this decision was the testimony of three of the leading scientists in the field (amounting to several hundred pages of transcripts), along with over a hundred pages of additional documentation, which included peer-reviewed and published scientific research and referenced and summarized thousands of pages of additional scientific literature.
Anyone is entitled to an opinion, and people who lack expertise in the field and who choose to form an opinion without first becoming at least well informed will reach unpredictable and widely varying opinions. I do not think it is possible, however, for any intelligent person to actually consider the evidence in an open-minded manner and reach any other conclusion other than the one reached by the judge in the Harrington case: that CKA meets the standard for scientific evidence established in our criminal justice system.
Technologies can be useful even without meeting this high standard. The two-question screen currently in place, bomb-sniffing dogs, and the subjective evaluation of the x-ray scans of baggage by low-paid security personnel with no scientific training do not by any stretch of the imagination meet any kind of standard for scientific validity. Nevertheless, they seem to be perceived as having sufficient value that they remain in place, and no one seems to be advocating dropping them.
The scientific value and validity of the science involved in CKA is well established, as anyone who actually examines the evidence with an open mind will see. There can be debate, of course, on the sociological, political, ethical, and logistical issues involved in applying CKA in the counter-terrorism setting. This is a healthy thing. In the end, I believe that in an open society such as ours truth and reason will prevail. Regards, Larry