Schneier on Security
A blog covering security and security technology.
« "The Behavioral Economics of Personal Information" |
| Me and the Christmas Underwear Bomber »
December 29, 2009
Interesting video demonstrating change blindness: the human brain's tendency to ignore major visual changes. The implications for security are pretty serious.
EDITED TO ADD (1/11): Two more videos. Daniel C. Dennett on the topic.
Posted on December 29, 2009 at 6:34 AM
• 35 Comments
To receive these entries once a month by e-mail, sign up for the Crypto-Gram Newsletter.
The only two shown as noticing the change were perhaps women who fancied the first clerk?
Implication - TSA should only employee women who like men with beards...
Or more sensibly, just people who naturally tend to make a lot of eye contact. The "failing" subjects seemed to make almost no eye contact.
a friend entirely focuses on the bust size of women (not the face, hair, hands, hight etc) - could have happened to him ...
eye witness accounts are proven to be very unreliable but yet we rely on them too much
This seems like classic misdirection. The subjects were focusing on the form, not the person presenting the form. Though the two greeters were clearly different, they were still quite similar. They were of similar height, weight, age, sex, race, and economic status. They were wearing similar clothing, almost a uniform for office functionaries. Had one been a girl, 100% of the subjects would have taken note of the switch.
Also, had they expected that they were already being observed as part of the experiment, I suggest that the recognition rate would have been much higher.
This is less startling than the video of the gorilla walking through the basketball players. Though, to be sure, still disturbing.
@Andy: in the Derren Brown video linked above, some folks actually did NOT notice when Derren changed places with a female.
Is it a case of they didn't notice, or just didn't care? Functionaries being switched out might be or such little importance to some people, they register it but don't overtly react. I believe in many cases, the person has taken in the changes, processed the information and the instinctual side of the brain, says "no threat" and moves on.
Major visual changes? A major change in minor details is not a major change to the overall visual. Had the race or gender changed I would have been impressed. Had one guy not had on a shirt or had a Hawaiian shirt, I would have been impressed, but changing one guy for someone that could be his brother is unimpressive.
Time, past time, for all good security professionals to add "magician" to their resume skill set.
@Eric Case: "Major visual changes? A major change in minor details is not a major change to the overall visual. Had the race or gender changed I would have been impressed. Had one guy not had on a shirt or had a Hawaiian shirt, I would have been impressed, but changing one guy for someone that could be his brother is unimpressive."
Good point. How many times do even intelligent people get seated at a restaurant, have their orders taken, then when they need something they don't remember which one is their waitress? (They never have to ask if they had a waiter or a waitress, nor do they mistake race.)
I could see where an intelligent person could even notice a difference from one person to the next, and chalk it up to "I must not have being paying attention before" and act like it's the same person.
Often when I'm working the cash register with a certain cashier on the next register, if one of us has a difficult customer causing problems we simply switch places to give us a chance to 'cool off' and deal with the other persons customer. Very few people notice despite the fact that I am a 25 year old man and the other cashier is a 43 year old woman. There usually too busy focusing on the fact that Patron is expensive or that the size they call a pint is actually a fifth.
I didn't notice that the researcher's shirt changed color, but I do remember thinking that they had filmed the researcher in two different locations. When I went back, the location never changed, just the shirt color.
Sounds like a perfect alibi. You only need an accomplice to switch places and people doing an exam or something similar.
There is also the embarrassment factor and a fear of being wrong.
Would you offer up something as absurd as "Well, it was a different fella at first, then he ducked down and this other fella popped up" unless you were 100% certain you weren't mistaken?
People don't want to be wrong or stupid.
Now Derrin Brown's video is striking. It's anecdotal, and I'm sure that there were people he didn't show that did react when the person asking for directions changed gender or race. Hopefully a majority, but clearly not 100%.
Brown's setup is pretty clever. Definitely higher stress, and set up so he could get away with much more.
I'd be interested to see the results of this experiment in a situation where a major change occurs that is of consequence to the test subject.
For instance, set up a guard desk where subjects are "temps" hired to check identities before allowing people to proceed. After a high number of normal entries, have actor 1 approach the desk, show his identification, drop something, bend over to pick it up, and swap with actor 2, who should stand up exuding the exact demeanor of someone who has just been granted entry.
Though I'd expect the failure rate to be much lower than 75%, I'm sure it would still be sufficiently high to raise eyebrows.
@Jason: "People don't want to be wrong or stupid."
That is true, and I think the argument probably applies to a number of people who were "tested" by Derrin Brown.
However, in the case of the experiment video posted by Bruce, people are being interviewed after the experiment, and some of them admit that they did not notice the change of person.
Most of the posters are silly, with mutterings about "anecdotes" and disjointed "reasons" for these results.
There's a simple heuristic of object permanence and continued identity of individuals -- of course, you pay attention to next to nothing on a day-to-day basis.
Why? Because paying attention to everything is expensive and error prone. You're much better off going by a few rules of thumbs -- like, individuals don't swap identity when they disappear for a second.
The situations here almost never happen to anyone; so of course we don't notice them. You notice switches that happen with high probability -- you spend your effort at extracting information, so your heuristics approximate an informational entropy measure.
Most of our lives are in our heads -- the "empiricist position" is fairly delusional.
I started looking around and found a few videos of this principle in action and I found that I only caught some of the changes. But what was interesting to me was that even when I caught a change, I couldn't say for certain what changed without rewinding to verify. I just knew something was different.
It seems to be a facet of focus. The most effect tricks were the ones that engaged the participant in an idea or object to the point they focused on it. After that the "noise" is filtered out. And then you can change an item that is "noise".
It's basically the same principle as sleight of hand. Get someone's attention then quickly and smoothly change that which they are not paying attention to.
The funny thing is I bet it wouldn't work as well on Schizophrenics. I recall reading that one of the symptoms commonly encountered by sufferers is the inability to filter out what most people would ignore. (Failure to habituate.)
I would guess that "passing" this test is actually a bad sign.
Reading the wikipedia article made me think of my old Magic Club days. There's a trick that you can easily find on the Internet. Here's the basic premise (obviously, in practice it is dressed up with drama, misdirection and humor):
Present five cards to someone. Ask them to mentally select one. Withdraw the cards and remove one. Show them the four remaining cards. Ta-da, the one they selected is gone! How? The four cards you showed them are all different from the original five cards. People are so focused on "their card" that they nearly never notice the other four are also gone.
Do you notice that when the researchers asked the subjects whether they noticed anything strange, everyone assumed it must have been some content on the form? No one even mentioned the guy behind the counter.
As a self-defense instructor, it makes me think this has safety implications as well.
People notice what's important and don't notice what's unimportant. I would expect this and don't consider it significant nor blindness. It's a natural part of the subconscious filtering of information to the conscious mind.
However Bruce's point that there are "implications for security" and it could be exploited are nonetheless true. Magicians exploit this effect all the time.
The brown video had some people who noticed - and one instance where the marks companion noticed.
The distraction involved exploits this effect - used by stage magicians and pick pockets. In one instance the magician changes sex, height, clothes, and race.
Adding to this: you don't need a safe to secure your wallet - just draw a yellow line around it...
(Though - passers by may have figured that the wallet was being watched - and they were right.)
I partially agree with the preceding commenters who were unimpressed.
In fact change blindness researchers have done some remarkable experiments on just how much you can get away with while (most) people don't notice, and it does have security implications -- but this particular example was a little lame. The two test actors were very similar and unremarkable in appearance; very similarly dressed in weak pastel shades which both looked grey in the dim office lighting; and more importantly, the actors were in a situation where most of the subjects were simply not paying any attention to them.
For example, watch the eyes of the girl in the blue t-shirt. She looks at the huge distracting sign above the actor's head, starts to scan the room (picking up such minute details as the potting medium in the planter) but then focuses on the papers thrust under her nose, and does not make eye contact with either actor until saying "thank-you" to actor no. 2 a moment before leaving. She literally doesn't look at actor no. 1's face. The girl in the brown t-shirt glances at actor no. 1's face very briefly when asking him the date, but not otherwise. (And since she doesn't know what day it is, she may be a little distracted otherwise!) Two of the men who didn't notice the change emphasised that they were in a rush; one of them is nervously glancing down the hall (where he wants to go) and in every direction except at the actor, whilst incidentally also adjusting his spectacles which are slipping down his nose: yes, this "change blind" subject is also short-sighted ...
[And as a cute aside, whilst reviewing the video to see exactly what each subject did, I finally noticed that Professor Simon's shirt also changes colour between the first and second scenes of his interview .. very cute gag!!]
The more remarkable security conclusion about this experiment is not that people noticed nothing but gender about a person when they didn't look at his face. The remarkable thing is that so many apparently educated people signed -- without reading -- a piece of paper handed to them by an unknown stranger whom they didn't even look at !!
Near the end, Prof. Simons remarks that is is not known whether particular people have a different likelihood of noticing the change, or if they were just lucky on the day. This would be very interesting information and should be fairly easy to test by simply exposing each test subject to more than one scene change before telling them what was going on. It would also be interesting to quantify the percentage against different types and sizes of changes.
This thread and the links make me wonder why Movie people are so focused on continuity and how do people catch continuity errors in movies.
I also had a look at the Derren Brown video -- the one where an actor asks a stranger for directions and is swapped mid-conversation (after being interrupted by a large painting being carried between the actor and the test subject.)
I have to say that on close examination, that video is also less than impressive. Of the subjects that Brown's editors chose to leave in the show, only one -- a highly extroverted person -- outright challenges the actor about the substitution. However in nearly all cases, the subject shows some kind of facial indication of consternation, but is simply baffled as to how to react.
For example the second subject -- the man with the central European accent -- doesn't openly challenge the substitution, but as he walks away, at 1:03 he subtly checks the contents of his bag. It might be a coincidence, but it looks to me as if he is worried he is being scammed and is checking he hasn't been pickpocketed. Subject 3 doesn't remark on the substitution, but stops assisting (she very politely suggests the actor take his enquiry elsewhere, by nominating a person she has obviously picked at random.) Subject 4 is the most highly extroverted and helpful, and openly voices his confusion at the substitution.
Subject 5 superficially appears the most extreme example, as a tall, close cropped white man is substituted by a much shorter Asian woman with a dramatic ponytail, and the subject seems not to notice. However, the subject DOES react: after a moment of speaking to the Asian actor, she looks around and then starts scowling and her tone of voice changes. It seems to me that she thinks the Asian woman is the first actor's wife or girlfriend, and that she is still speaking to both of them; when she realises the male actor has walked off mid-sentence, she becomes annoyed at his rudeness. I admit this explanation is speculative, but at any rate it is certainly not the case that she doesn't react.
This video would have been far more impressive if each subject was stopped again slightly further down the street and asked what they thought had just happened. As it stands, all that can be said is that when these people were accosted on the street in a big city, *if* they noticed something hinky they mostly chose not to verbalise it. But that is neither surprising nor interesting.
To add to what Roger said, I'd say that asking someone for *directions* is also a particularly poor test, which will focus someone's attention on their own memory / mental map / physical surroundings.
I'm asked for directions pretty frequently, and after I've understood the person's question, I'm certainly not paying any attention to them. I'll answer the question, get an acknowledgment, and get back to walking to wherever it was I was going.
There's nothing particularly shocking or novel about simple misdirection of uninterested people. The gorilla/basketball test is a *vastly* better example in every way of whatever point can be made about missing major changes.
To those people who complain about the test setup (lack of attention on the actors, poor setup etc.): the missing focus on the changed subject/object is exactly what "change blindness" is about.
The remarkable thing about these experiments is how tiny the focus is and how much filtering the brain does, unknown to most of us in our daily lives.
> To those people who complain about the test setup ... the missing focus on the changed subject/object is exactly what "change blindness" is about.
I don't believe that's quite true. The phenomenon of change blindness occurs when someone is *viewing* the scene that changes and doesn't notice it. We can get all semantic-ky on what we mean by "viewing", but it is obviously trivially true that someone will not notice a change in a scene that they are not viewing, so the "phenomenon" becomes trite if it is defined too broadly. It must be defined so that it is in some sense surprising that the change was not noticed.
For example, the dancing gorilla is surprising because we expect that an observer who is closely monitoring the actors would notice a gorilla in their midst. Most people are surprised that the mere act of concentrating on a subset of actors could cause one to not notice a (simulated) large, aggressive animal walking right through your field of view.
On the other hand, the experiment described here is unremarkable because we expect that at least some people pay no attention to -- in fact, do not even look at -- a minor functionary who hands them a piece of paper. If they don't even look at him, it is unsurprising that they do not notice his face.
Watch how a typical 1 to 3 year old approaches the world. What are they learning? To a great extent, they're testing the world to find patterns that can be ignored later. Knock things into motion 50 times and you learn that they fall down, scatter, etc, so for most of the rest of your life you don't pay attention to how things fail. In other words, it seems to me that change blindness is something we spend a great amount of our toddler years learning to do. It's a feature, not a flaw, that helps us cope with the huge amount of information in the world.
Schneier.com is a personal website. Opinions expressed are not necessarily those of BT.