Reached Via a Mind-Reading Device, Deeply Paralyzed Patients Say They Want to Live (technologyreview.com)

Neuroscientists have designed a brain-reading device to hold simple conversations with "locked-in" patients that promises to transform the lives of people who are too disabled to communicate. Details of four patients who were able to communicate using what is being touted as a groundbreaking system were made public this week. From a report on MIT Technology Review: Now researchers in Europe say they've found out the answer after using a brain-computer interface to communicate with four people completely locked in after losing all voluntary movement due to Lou Gehrig's disease, or amyotrophic lateral sclerosis. In response to the statement "I love to live" three of the four replied yes. They also said yes when asked "Are you happy?" Designed by neuroscientist Niels Birbaumer, now at the Wyss Center for Bio and Neuroengineering in Geneva, the brain-computer interface fits on a person's head like a swimming cap and measures changes in electrical waves emanating from the brain and also blood flow using a technique known as near-infrared spectroscopy. To verify the four could communicate, Birbaumer's team asked patients, over the course of about 10 days of testing, to respond yes or no to statements such as "You were born in Berlin" or "Paris is the capital of Germany" by modulating their thoughts and altering the blood-flow pattern. The answers relayed through the system were consistent about 70 percent of the time, substantially better than chance.

Comments

Harsh

[Anonymous Coward]

Doctor: Paris is the capital of Germany
American patient: Yes
Doctor: Okay, this one can die

Re:Harsh

[Anonymous Coward]

Eh, RepublicanCare ("The World's Greatest Health Plan") will go more like this:

Doctor: Do you have money to pay us for keeping you alive?
American patient: Yes
Doctor: Okay, this one can live. Follow up in one week for repeat testing.

Re:Harsh

[davester666]

A week is WAY too generous. Try prepay. "The week you paid for is up. How much money can you give us now? $X That buys you an additional 3 days here. See you in 3 days."

Re: Harsh

[micahraleigh]

Step 1) Promise government services via taxation

Step 2) Claim those who use said services are a drain

Step 3) Squander the tax revenue on government programs that create exotic waste (like hyperloop, brains in vats, colonizing the inside of volcanos, etc).

Re:Harsh

[davester666]

Yes, it could eventually be an awesome breakthrough, that will likely only help incredibly wealthy individuals.

Seventy Percent

[Anonymous Coward]

Sorry to be the skeptic, but 70% of the time isn't really that high. it means that the machine can read your thought 40% of the time and then the remaining 60% is a coin flip.

And this assumes no biases introduced into the process. We've seen in the past that ways of reading a paralyzed persons thought have turned out to either be scams, or well-intentioned people unconsciously affecting the results of the readings.

Now, I really hope we find ways to accomplish what this researchers are claiming, but I am skeptical.

Re:Seventy Percent

[toonces33]

True, but even 40% is better than nothing. If all of this is true of course.

Re:Seventy Percent

[omnichad]

it means that the machine can read your thought 40% of the time and then the remaining 60% is a coin flip.

That's not how probabilities work. It's true that it's better by 20% than a coin flip, but the rest of your conclusions are completely wrong.

Re:Seventy Percent

[Gr8Apes]

40% would be worse than guessing, with a yes/no possibility.

Re:Seventy Percent

[MBGMorden]

No because the 40% was already eliminating the guessing and was reducing the actual result of 70% (the whole 40% thing was confusing and unnecessary).

The real % of the time it's right is 70%, which is significantly better than random chance (which would be 50%).

Re:Seventy Percent

[ShanghaiBill]

True, but even 40% is better than nothing.

The 40% is likely not random. The response may be stronger when it is something the patient cares deeply about. So the response to "Is your shoe size nine?" may be weaker than "Do you want us to kill you?".

Also, like many bio-feedback systems, the accuracy could be improved with training. If the patient practices, they may get much better at giving the intended response. It is not like they have something better to do with their time.

Re:Seventy Percent

[AvitarX]

Can non locked in people use the device?

What percentage?

Also, 70% should allow sentences to be written (frequency order the letter using a language library for next (similar to Dash), slowly show them, have the patient think YES fromnwhrb they see the one they want until "yes" is registered, and then start with the next letter.

At 70% one should be able to get close enough to desirable the results.

Re:Seventy Percent

[Anonymous Coward]

The response may be stronger when it is something the patient cares deeply about. So the response to "Is your shoe size nine?" may be weaker than "Do you want us to kill you?".

Then again, certain answers may be more nuanced than a simple yes or no. I imagine that my answer to the "Do you want us to kill you?" question would be along the lines of "Do whatever is best for my immediate family." Maybe my wife and daughter would like having me around even if I was "locked in" or maybe I would just be a burden - both financial and otherwise.

Not all questions have a yes or no answer.

Re:Seventy Percent

[AthanasiusKircher]

That's not how probabilities work

Exactly. It COULD be that this is what's happening (i.e., 40% of time reading accurately, the rest of the time just chance). But it could also be all sorts of other more complex relationships with data.

That division depends on a strict delineation between results which are KNOWN to be accurate vs. those subject to chance. But it could also be that measurements of brain activity only his a certain threshold in a subset of measurements.

For example, think of rolling a 10-sided die, and if you get 1 through 7, that's "success." 8 or above is "beyond the threshold." That would produce a 70% "success" rate. But it would be inaccurate to say that rolling the die had a 40% accuracy "within threshold" but 60% of it was "a coin flip." Which rolls of the die would be the "coin flip" ones exactly, and which ones the "guaranteed below threshold"? Now, if we had some way to measure the outcomes of a few "dice rolls" with 100% accuracy ahead of time (say, all rolls of 2, 3, 6, and 7), while you couldn't differentiate the rest (1, 4, 5, 8, 9, 10 -- all completely indistinguishable in advance), the GP's statement could be accurate. But that's not necessarily how this measurement system works in such cases.

Not saying this experiment worked like a 10-sided die either. I'm just saying there are all sorts of possible mechanisms here and ways to measure, and most of them can't just reduce a 70% success rate to a 40% "actual" success rate and a 60% "coin flip."

Re:Seventy Percent

[mutantSushi]

Exactly... Both the tech itself can be improved in accuracy, and normal bio-feedback training be used to increase the accuracy once baseline communication is established. The thing is, and this is also applicable to "questions that aren't just Yes/No", is that multiple questions can be used to follow up: "Is Paris the capital of Germany?" "OK, is that right, you just told me "NO"?" ...etc. This also addresses cases where certain topics may be less reliable or complicated.

Re:Seventy Percent

[rahvin112]

You can only claim better than Random in a situation like this if the questions and responses aren't conditioned or influenced by the questioner. Count me a skeptic. Time will tell if the methodologies are sound, without consensus and duplication the individual results should be interesting but not confirmed.

Re:Seventy Percent

[Immerman]

Indeed. This should be subjected to the same rigorous testing that any other claim of mind reading gets in a scientific setting. Ditch the subjective statement and "consistency" measures - that's just a playground for wishful thinking.

Go through a deck of black and white flashcards, "For each card, tell me if it's black". Or warm and cold objects held against the skin. Ideally a whole range of different binary stimuli to make sure it's detecting thought and not just the recognition of a particular stimulus. If they can get 70% accuracy on objectively verifiable randomized tests, where the people applying the test and interpreting results don't know the correct answer until after everything's been commited, THEN I'll take them seriously. Until then it sounds like nothing more than a way to guilt caretakers out of more money to maintain people who've been imprisoned in their own bodies.

Re:Seventy Percent

[rtb61]

Sorry but to claim that you must use massive data sets. So for example toss a coin three times and get three heads in a row, still highly probable just less than a fifty percent chance. To validate, you must toss that coin at least 10 times and preferably 100 to actually get anywhere near fifty percent heads and fifty percent tales, the fewer the tosses the far more likely 80% heads and twenty percent tales etc. is quite readily possible. So the test seems to be more about making people more comfortable with generating profits from people trapped in a vegetative state by them (so the ones that say no, are they still trapped and feeding profits with their suffering).

Re:Seventy Percent

[Immerman]

True, but frequency ordering would get horribly tedious fast. Something more like a binary tree would be much more efficient - you can identify any letter in the alphabet with only 5 yes/no questions: Does the next letter come before N? - yes. Before H? -no Before K? -yes Before I?-yes Your letter is H. Next letter...

Of course with a binary search you *always* need to answer log2(N) questions (or one less, if you for a non-power-of-two number of options), and we could do much better using an unbalanced tree where the most frequently used letters required answering fewer questions.

Probably you'd want to navigate something like a static Huffman coding tree. As I recall (can anyone confirm?) an optimal coding of US English averages around 2.5 bits per letter, so on average you'd only need about 15 answers per six-letter word. For added convenience you could also add one or more "escape characters" to allow fast access to a palette of common phrases.

It never ceases to amaze me how crude many such communication devices are. I could understand doctors ignoring decades of information theory as outside their field of expertise, but the programmers who actually write the software have no excuse. I mean yes, there's going to be a learning curve as you learn your way around your Huffman tree, but if you're stuck communicating through a computer for the rest of your life, I'd think it would be worth the effort. Though I could understand picking something a little suboptimal just to avoid running into those 13+bit letters in the middle of a sentence.

And of course, if your equipment can distinguish between more than just two states, most such encoding trees can be trivially extended to 3- or more-way trees to drastically reduce the average number of nodes you need to traverse

Re:Seventy Percent

[Immerman]

Same way you would with any other mind reading claims - have the patient identify a long random sequence of flash cards (or other completely non-subjective input) that the person interpreting the instruments can't see. This was settled methodology almost a century ago - that these people are offering such shaky claims about "consistency" on completely subjective statements, without offering objective proof of their claims, suggests to me that this is nothing more than a scam.

Re:Seventy Percent

[AvitarX]

I was working on the assumption there were three states though "no answer, yes, no" and that no answer would have a much higher success rate.

I would suspect that with frequency (based on history, not entire language) you could get pretty close, for example after prett "y" would be next, after "gue" first S then R (I just typed gue into google, I don't know the reality).

If you are working by elimination, I feel the failure mode will become very difficult to decipher, but if you can say OK, it was this letter or the 4 before it, someone could decipher it pretty easy (this is of course assuming there's a resting state that is identifiable, and it isn't always reading yes or no).

Another solution is use your method, but ask each grouping 4 times, and if it is split, ask 6, for increments of 2 (or some such nonsense), this should get the groupings pretty accurate, and allow a focus on nonsense to fill in a wildcard (like an X or some such, getting the odds even better).

As far as these results go (without RTFA, because I'm not a n00b) it seems meaningless to me. I see little evidence they can even comprehend a question well enough to answer it.

Of course, as soon as they can communicate, I suspect the desire to live would go up too.

Re:Seventy Percent

[b783719]

70% of the time means if you ask the patient 100 times if they like green eggs with ham, the machine should give you either 70 no's with 30 yes's or 70 yes's with 30 no's.

Also

the machine can read your thought 40% of the time and then the remaining 60% is a coin flip

is not a great way to look at probability. We could say the same thing like this,
the machine can read your thought 70% of the time and then the remaining 30% will surely fail. Or
the machine can read your thought 0% of the time and the remaining 100% is a chance of 7 out of 10 die roll.

In the end, you've just confused yourself.

The easiest way is to do this." 70% of the time" means EACH time they pick a decision, there is 70% chance that the machine return the same as the choice they picked.

For example, you have a 6-sided die (dice for plural) and you wanted number 1-4. you roll it once. "66% of the time" you will get 1-4 from that roll. If the die is fair and after you roll it 100 times, it will STILL be "66% of the time" that you get 1-4. In that case, "66% of the time" the die will return a number you want.

Re:Seventy Percent

[Black Parrot]

Sorry to be the skeptic, but 70% of the time isn't really that high

That would be a pathetic result in machine learning.

I wonder if it was double-blind.

I wonder if 50% of the test questions were 'yes' and 50% 'no'.

I wonder whether the test questions were presented in random order.

I wonder if it requires a human interpretor.

I wonder if other researchers can duplicate the results.

Sounds like someone is trying to apply a lie detector to unconscious people.

Re:Seventy Percent

[fph il quozientatore]

"Success percentage" means nothing, as everyone who has studied a little bit of statistics knows well. There are well-established methods to tell if a result is meaningful or not. In other words: P-values or GTFO.

Re:Seventy Percent

[Immerman]

True - assuming it could actually also detect a third "null" state would improve things dramatically. A linear scanning system though would basically be ignoring the "no" state, so you'd be wasting around 1/3rd of your agonizingly narrow input channel.

Consider too, if you're averaging substantially less than three nodes per letter then you can scan MUCH slower than you would linearly, which would presumably increase signaling accuracy and decrease user stress, while still increasing speed. Instead of "they picked one of the last four letters" you ca For 44% coverage you only need two steps in a tree, versus 5 linearly. For 85% coverage that increases to 3 in a tree versus 13 linearly. (As pulled from this trinary Huffman tree I built out of curiosity: letters by number of choices needed to reach them, with combined probabilities )
2: [space] e t a o (~44%)
3: c u i n s h r d l (~41%)
4: b v m w f g y p ( ~14%)
5: k j
6: x q z

Of course there's no reason you couldn't combine it with predictive text (partial?)completion and autocorrection, just add a "use one of the guesses" character, perhaps have [space] invoke a special "inter-word mode" with options to pause, edit, etc. as well. One extra choice to "continue" on each word,

Tree navigation would also make the mental patterns somewhat like writing - left-down for E, down-right-right-left for G, etc. I suspect that with practice it could become quite fast and comfortable.

In fact, I think I would ditch the scanning altogether, that seems like a ruthlessly mechanistic compromise for those who can only send a single signal. If you can send two signals (plus null) then I suspect the gains in comfort of being able to "write at your own speed" would outweigh any speed benefits from an extra input state. It might even end up being faster - the ability to organically improve your speed might well outweigh the benefits of a shallower tree. Besides, if you can manage to reliably detect two brain signals, a third probably isn't far behind. though beyond that you probably hit rapidly diminishing returns.

Hmm, now I'm wondering if anyone has experimented with using something like a modified Morse code using the timing between signals for people who can only generate a single signal. I mean three states shouldn't bee too hard to generate - short, long, really-long(null). So long as the three are clearly distinct, you would have a lot of flexibility in pacing. You could even analyze the current pace and adjust timing thresholds on the fly.

Re:Seventy Percent

[Immerman]

Indeed, but try compressing text files using any tool you want and you'll find that a factor of 3 is actually nothing special. Huffman coding actually has severe limitations due to being unable to allocate bits accurately enough to reflect actual probabilities (you can't allocate only part of a bit to a character). ASCII is horribly inefficient for plain text encoding - the top half the character space wasn't even part of the original 7-bit ASCII standard, and even in that there's an awful lot of control codes and other rare special characters included.

Just storing the alphabet plus space, with no capitalization or special encoding, only requires log2(27) = 4.7 bits, and vast majority of the characters, by use count, come from a small subset of that, and the more common a character, the less actual information it's appearance contains - in general the information content (in bits) = -log2(probability). So consider the 6 most common characters in english text:
[space] 13% = 2.9 bits
e 13% = 2.9
t 9% = 3.5
a 8% = 3.6
o 7.5%= 3.7
i 7% = 3.8 ... all of which are bigger than my claimed number, so I must be misremembering. Probably I'm thinking of context sensitive Huffman coding, which can push the probabilities much higher by looking at the last character(s) and "guessing" at what comes next. For example if the previous character was a Q, the odds that the next character is a U is extremely high, and so you can encode it with only 1 bit to essential say "yep, you guessed right", while in another context where U is very unlikely it might take 13 bits to encode.

Re:Seventy Percent

[peawormsworth]

Maybe the computer reads your mind 100% the first day and then spends the next day just sitting around flipping coins.

Or maybe it was right 100% of the first 20% and then only 90%, 80%, 70%... down to 0% on various ranges of percentages. And the total being correct 70% of the time.

Did you know a coin flip is not really 50/50, but rather it is always heads 25% of the time and only has a 1 in 3 chance of being heads in the remaining flips.

Or maybe your example and all of mine are wrong. (gamblers fallacy)

Re: Seventy Percent

[rpstrong]

Hey, I'm a sexist pig - you insensitive clod.

Interesting, but I'm not sure I trust it

[scubamage]

This is neat research, but I want it much more strongly vetted. It reminds way too much of the facilitated communication mess we encountered several years ago.

Re:Interesting, but I'm not sure I trust it

[Anonymous Coward]

As they hooked up the device and turned it on, I felt an immense sense of relief. Finally, after all these years of staring at the bleak beige ceiling tiles.

"Do you want to live?" the doctor asked.

In my mind, I shouted, "No!" No! A million times no! Release me from this dead body so that I can move on to the summerland! No! Let me die!

The doctor turned to my wife and told her, "He says that he wants to live! We should be able to keep him alive for another good 30 years."

I can't move my mouth, and I must scream.

Re:Interesting, but I'm not sure I trust it

[The-Ixian]

This is neat research, but I want it much more strongly vetted.

Nope. We have reached the top of the mountain. There is no further we can go. All research into this will halt now.

Sorry, this is as good as it gets.

Re:Interesting, but I'm not sure I trust it

[MBGMorden]

You say that as someone who isn't enduring it.

To many people anything less than what they have now is completely unbearable. To the rich they think if they were poor they'd kill themselves - despite many poor people living happy lives. To the young many think that they'd rather die than grow old - even though "old" is a moving target that keeps getting a little farther out as you age.

The will to live is strong - those without it don't pass on their genes as readily. Don't presume to know whether they'd want to live unless you're in their situation.

Re:Interesting, but I'm not sure I trust it

[Anonymous Coward]

Personally I'd rather I get results from a scientific study than creative writing . . .

Re:Interesting, but I'm not sure I trust it

[scubamage]

That's some fine science work there, Lou!

Re:Interesting, but I'm not sure I trust it

[Tablizer]

Please don't use the phrase "strongly vetted". You-know-who ruined it, now meaning, "Reject 'em if they give me heebie-jeebies".

Re:Interesting, but I'm not sure I trust it

[hey!]

People underestimate their ability to adapt.

Probably the most traumatic thing imaginable is the death of a child; it's hard to imagine wanting to go on. But for the most part they do go on. It doesn't make the experience less horrible and traumatic.

People adapt to whatever level of pain or comfort they experience. You can win the lottery and you'll be happy for a while, but very quickly, within a year or two your happiness level returns to what it had been. Happiness as an emotion exists to motivate us; it's not for having it's for chasing.

Re:Interesting, but I'm not sure I trust it

[IcyWolfy]

I read once in the local paper an interview with a patient who had some sort of paralysis, (maybe coma?) that they unexpectedly recovered from.

The only thing I remember about the article was her description that:

It was just like dreaming when I was asleep. But when I woke up, I was originally scared, and afraid, and once I accepted it, my imagination started to take over. After a while, I was able to picture myself doing everything I wanted. Everything felt like a dream. And then one day, when I wasn't being busy, I was just relaxing, doing nothing, when I heard a faint ringing sound. And I just started to focus on my inner self. And then I realised that the body I identified with, wasn't my physical body, and that everything I was experiencing was inside my body, expanding and contracting, and like when you realise you're dreaming and you wake up; what I saw and felt around me started to dissolve, as I could feel parts of the world around me, my hearing waking up from it's own generations to taking the sounds of the world outside. And then, I slowly woke up. And like a dream, many bits I could remember simply started to disappear. And then I realised that that body I was used to the past few months, was in a lot better state than how this one's become.

From that account, it really pushed into me the notion that in such a state, like those good at visualizatoin, or meditation -- You basically just learn to re-create your own full reality. Hyper-real lucid dreaming. Better than reality, and more connected, as you realise that your perception of reality is all self-generated in your head, taking into advisement the pulses of electromagic radiation hitting our senses, make it into a pulse, and we create our own personal reality.

Re:Interesting, but I'm not sure I trust it

[nsre]

The "only thing" you remember is a word-for-word paragraph-long account of waking up?

Re:Interesting, but I'm not sure I trust it

[Quirkz]

Also, if they're frequently in a dream state, it may be relatively pleasant. I know that there have been lots of mornings where, when the alarm clock goes off, I find myself wishing I could go back to the dream I was having, even though I'm awake enough to know it wasn't real.

Re:Interesting, but I'm not sure I trust it

[Hognoxious]

Are you interpreting the message as "don't drink and drive, or we'll kill one of these disabled people"?

Re:Interesting, but I'm not sure I trust it

[Hognoxious]

Yeah I can't imagine any patient in that state wanting anything other than death.

I agree. Recovery my ass.

Re:Interesting, but I'm not sure I trust it

[Trogre]

Damn good point there. I'd mod you up if I could.

Re:Interesting, but I'm not sure I trust it

[Immerman]

>Probably the most traumatic thing imaginable is the death of a child
Then your imagination is woefully limited - infant mortality may be traumatic, but it's a trauma evolution has equipped us to deal with - after all for most of human history most children died within their first few years. If you can't deal, you stop breeding and your fragile genes are erased from species. There's a reason so many cultures don't give their children "real" names until they're many years old.

Re:Interesting, but I'm not sure I trust it

[irrational_design]

You haven't read much if you think that is probably the most traumatic thing imaginable.

Re:Interesting, but I'm not sure I trust it

[JonnyCalcutta]

...in America

Re:Interesting, but I'm not sure I trust it

[micahraleigh]

First time I read your post I thought you were being sarcastic.

Success rate

[arth1]

70% doesn't seem high enough to make any decisions.

And how was this controlled for confirmation bias, like has been discredited for other techniques where the person that reads the results also knows the answers, like e.g. dog training and lie detectors?

Without doing double blinds, 70% seems like a horribly bad result, and no more than what would be expected from confirmation bias.

Re:Success rate

[omnichad]

And how was this controlled for confirmation bias, like has been discredited for other techniques where the person that reads the results also knows the answers

The computer could be double-blinding it to an extent. It depends on if they're looking at the signal and making a human determination or if the computer is doing statistical analysis on the input and making the 'yes/no' answer for the researcher.

Re:Success rate

[DickBreath]

It is probably good enough if you have three responses instead of two.

Instead of two responses: Yes and No

have three responses: Down Arrow, Right Arrow, and ENTER

Now I bet the reliability of your Yes / No responses is much higher.

Re:Success rate

[Fragnet]

Needs to be replicated a few more times, especially after the various neuroscience debacles we've had recently.

Re:Success rate

[AthanasiusKircher]

70% doesn't seem high enough to make any decisions.

To make legal decisions on care or something? Probably not. But it MIGHT be evidence that we're on the right track to communication.

And how was this controlled for confirmation bias, like has been discredited for other techniques where the person that reads the results also knows the answers, like e.g. dog training and lie detectors?

This is a valid question. I skimmed the actual study, but I don't have time right now to dig through the jargon and see how much these results are likely to be due to confirmation bias.

Here's the actual study. Does someone who knows more about these sorts of measurements want to sort out whether or not there were adequate procedural constraints to prevent confirmation bias?

Without doing double blinds, 70% seems like a horribly bad result, and no more than what would be expected from confirmation bias.

That's just nonsense. You can't tell whether confirmation bias is present by the level of success! That's not how stats work. In some cases, confirmation bias could easily produce a 95% or even 100% success rate. In other cases, it would be barely better than chance. You can only tell confirmation bias by looking at procedure and data analysis techniques.

And in any case, I'm surprised at the statistical ignorance shown by many posts in this thread. 70% success where 50% is chance may or may not be a significant finding -- if you do it with 10 questions (or coin flips or whatever), it's probably not significant. But if you ask a million questions or flip a coin a MILLION times and see 70% heads or whatever, it's pretty strong evidence of a pattern. (Would you place a 1:1 wager and gamble against heads on a coin after a million flips like that?)

But again, whether the result shows strong statistical significance from data analysis is a different question from whether confirmation bias could be present in the procedure.

Re:Success rate

[arth1]

That's just nonsense. You can't tell whether confirmation bias is present by the level of success!

I wasn't attempting to. I attempted to say that a 70% success is well within what's often seen with confirmation bias affecting the results for other types of tests.

Unless the person interpreting the resulting answer never heard the question, nor could deduce it from others present, I question the validity. Even as a preliminary. We should not get our hopes up. This could turn into the next lie detector hoax, and be abused the same way.
In which case it would be worse than nothing.

Re:Success rate

[rahvin112]

It's possible, even with yes/no questions, to manipulate the question, the phrasing and the body language to indicate a desired result when the person is actually capable of physically responding, I can't even begin to imagine the confirmation bias that could be imparted when the questionee isn't capable of responding in a recognized way.

Wait to see if the study can be confirmed by someone else without the same motives of this researcher. This result of "I want to live" reeks of someone that's trying to prevent euthanasia and shut down the debate and caution should be taken with any claimed results without repetition and review by experienced researchers.

Re:Success rate

[omnichad]

This result of "I want to live" reeks of someone that's trying to prevent euthanasia

I wasn't sure if it was that or something with a high likelihood of a strong emotional response. Though I don't think the test method depends on the emotional response, but something more deliberate.

Re:Success rate

[gonz]

This is a valid question. I skimmed the actual study, but I don't have time right now to dig through the jargon and see how much these results are likely to be due to confirmation bias.

How would that happen exactly? The results are calculated using an SVM classifier algorithm, not a human "interpreting" the results. Basically they train the classifier on 50 sessions, and then test in on maybe 7 sessions. Each session involves asking the person 20 questions.

Here's the actual study. Does someone who knows more about these sorts of measurements want to sort out whether or not there were adequate procedural constraints to prevent confirmation bias?

The most likely bias in this scenario would be sampling bias, not conformation bias. It would include mistakes like this: - Testing a bunch of different subjects, but only counting the favorable ones as your sample set (i.e. "he didn't have ESP") - Doing a bunch of trials, but eliminating the ones with unfavorable outcomes (i.e. "the equipment wasn't working right that time") - Choosing to stop the trials at the point where a favorable result is obtained (i.e. after the random walk went where you wanted) The Discussion session seems to address these concerns:

Four patients in CLIS communicated with frontocentral cortical oxygenation-based BCI with an above-chance-level correct response rate over 70% during a period of several weeks. The performance of the binary SVM classifier across all the patients, except a few training sessions of patient B, was above chance level. None of the sessions were eliminated in the analysis, and only very few sessions had to be interrupted because of life-saving measures such as sucking saliva; thus, no bias for selecting âoesuccessfulâ sessions incriminates the results.

So the main question is really just about the sample size itself. 20 questions x 4 people x around 5 sessions = 400 coin tosses. Is that enough to get excited about 70%? In pure mathematics, yes, but experiments can be corrupted in all sorts of ways. (What if the research assistant simply talked louder when he was saying a "true" question?) It would be great at least to see the results replicated by other groups.

Re:Success rate

[micahraleigh]

The only thing horrible is your murderous attitude.

70% of slashdot belongs behind bars.

Bad Questions

[Spy Handler]

"Paris is the capital of Germany" -- some people will legitly not know the answer. And besides France and Germany are both white-man countries in Western Europe that are very close to each other; someone not fully alert could easily confuse the two.

"You were born in Berlin" -- people with varying degrees of amnesia or repression forget their personal details, but still retain general knowledge of the world. For instance, "Do Birds Fly?" or "Is ice hotter than the Sun?" are questions that even full-on amnesiacs can answer correctly.

Also, 70% seems like a pretty horrible accuracy rate. For yes/no answers to such super-simple questions, the success rate should be 100% or close to it.

Re:Bad Questions

[Oswald McWeany]

For instance, "Do Birds Fly?" or "Is ice hotter than the Sun?" are questions that even full-on amnesiacs can answer correctly.

Also, 70% seems like a pretty horrible accuracy rate. For yes/no answers to such super-simple questions, the success rate should be 100% or close to it.

Yes, but if you ask someone if birds fly- they might think about penguins or emu, or ostrich, or dodo, or one of the other flightless birds.

Re:Bad Questions

[Lumpy]

You cant count americans in that. The american education is so bad that a large swath think that europe is a country.

Re:Bad Questions

[DickBreath]

Are you saying it is not?

Re:Bad Questions

[Wycliffe]

Yes, but if you ask someone if birds fly- they might think about penguins or emu, or ostrich, or dodo, or one of the other flightless birds.

So keep it simple and don't rely on their knowledge or the amount of amnesia or brain damage. What's wrong with questions like "Is 3+3=7" and "Does the word Frog start with a G?" If all you're trying to do is detect Yes and No then there are plenty of black and white questions that are much easier to answer.

Re:Bad Questions

[Oswald McWeany]

I've had Americans tell me that my English is amazing, I sound almost fluent. (I'm English). I've also had an American say "I love your accent, are you from Korea" (I'm not asian).

Re:Bad Questions

[Anonymous Coward]

The subjects are in Europe, in Geneva. I seriously doubt if you could find anyone over the age of 10 in Geneva who doesn't know the answer to those questions.

Re:Bad Questions

[penandpaper]

Yes, because Paris doesn't exist in Texas. Moscow doesn't exist in Idaho. London doesn't exist in Kentucky.

I used to think the same as you do i.e. American education is dumb because Americans don't know where Paris is! Until I realized that major city names are common across the US. When you say "I am from Paris" and the response being "Where is that".. It doesn't mean they don't know about Paris France but are wondering what local city in the state or neighboring state you are talking about.

Re:Bad Questions

[slew]

You cant count americans in that. The american education is so bad that a large swath think that europe is a country.

Of course the USA wanted Europe to become a country like the US after WWII and embraced this age old idea of "The United States of Europe". After all us Americans expect the whole world to copy us eventually ;^)

Re:Bad Questions

[Quirkz]

Well, I think most people should know that even if there's a tiny village in their area named Paris, if they're older than about age 8, they should know there's a world-renowned Paris out there that it's named after. I personally grew up within a short drive of the cities of Versailles and Milan (pronounced ver-SALES and MY-lun, by the way), and if there was a time when I might have confused them against the European cities, it's so long ago I can't remember it. Also, if you're talking in the context of the capital of a country, you're even less likely to mix that up with your local crossroads.

Re:Bad Questions

[thinkwaitfast]

Don't forget Perris, CA. Near LA. (Los Angeles, not Louisiana)

Re:Bad Questions

[rickyslashdot]

Don't forget Moscow, Texas - https://en.wikipedia.org/wiki/...

Re:Bad Questions

[JonnyCalcutta]

I stay not too far from Moscow - https://en.wikipedia.org/wiki/...

Capt. Pike

[Tablizer]

Slashdot didn't use the Captain Pike keyword under the story. Dammit, I'm gonna revoke their Geek Card!

Re:Capt. Pike

[slew]

Slashdot didn't use the Captain Pike keyword under the story. Dammit, I'm gonna revoke their Geek Card!

Didn't happen in the new Kelvin Timeline, he was killed by Khan. Keep up with your temporal mechanics or I'll have to revoke *your* Geek Card ;^)

Re:Capt. Pike

[Tablizer]

You get a Dweeb Card.

For the record

[admin7087]

I just want to make clear here and now that I do not want to live in case I should ever by completely paralyzed.

On a side note, if you share my sentiment you should check out your local lawmakers' provisions for such cases. Since my parents live in Germany, I know that at least in Germany you can make what's called a Patientenverfuegung at a notary. Emergency services and doctors will respect this document if you have it in your pocket and will switch off the machines. You can also determine a next of kin who will be responsible for this decision if reachable. The document needs to be official, though.

Re:For the record

[Tablizer]

I just want to make clear here and now that I do not want to live in case I should ever by completely paralyzed.

But there's always the possibility remote sensing will soon allow you to cruise around in a robot body, visit trade shows, and grab p.... uh, packet analyzers.

Experimental direct-brain hookups already look promising.

Being paralyzed itself is not what scares me; it's being bored if I cannot interact with anyone or read books or rant about stupid web GUI (non) standards on slashdot, etc. The interaction being remote is good enough. After all, my mom's basement wasn't so bad.

Re:For the record

[Oswald McWeany]

On the plus side, if you're completely paralyzed bad web GUI on Slashdot won't bother you.

Re:For the record

[Tablizer]

Paralyzed people use Lynx? You surveyed them already? (I wasn't referring to Slashdot's own UI, by the way, although now that you bring that up...

Re:For the record

[Khashishi]

Being bored isn't so bad. I find I can occupy my brain with day dreaming or full dreaming.

Re:For the record

[david_thornley]

If you are completely paralyzed, you might change your mind. It happens. In that case, it would be good to be able to confirm that's what you want before pulling the plug. In the meantime, see what you need as an advance healthcare directive or living will or whatever.

Am I the only one wondering?

[negrace]

Why use disabled people for testing? Use healthy ones, this is how you will know if it is working right or not.

Re:Am I the only one wondering?

[omnichad]

They probably did. But since ALS damages brain cells, you likely have to re-calibrate. Looks like they were doing this in 2010 with less severely disabled people. I'm too lazy to look further back.

Re:Am I the only one wondering?

[Solandri]

The device probably works better with certain people, not so well with others. So the average success rate in controlled studies may not be relevant to a specific individual.

Still, it seems like it'd be trivial for family members to come up with questions that only the patient would know the answer to. Write them on a note, doctor takes it into the room and asks the questions, writes down what the machine says are the answers, and brings it back out to the family for review. (Can't have the family in the room when the calibration questions are asked, lest the doctor takes queues from the family and guesses the answer.)

Re:Am I the only one wondering?

[houghi]

If the doctor can guess the answer, it is not a very good test. He could guess wrong even without the influence of the family.

And perhaps the patient has such a stupid sense of humor that he answers wrong just for the heck of it. I know I would and I know my dad would as well. Paris the capital of Germany? Sure. Ask stupid questions, get stupid answers.
Then there is my mom who would answer yes if she would not be sure. And she would be confused by such a question. Even the fact that it is such a stupid question would throw here of.
I might say yes as well if you ask me the question and I am not in a coma yet. Too early and not a weekend yet.

Binary communication

[SeaFox]

If this new mind-reading device can deduce between two (three?) specific states of the brain, this opens the door to much more once the efficiency improves.
You could communicate in Morse code, since you can now have signals that can be interpreted as "dots" or "dashes".

Re:Binary communication

[Black Parrot]

Jean-Dominique Bauby wrote his memoir by blinking one eye.

Re:ROUGH FUCKING VEGETABLES

[colinrichardday]

They can't consent, either, so it is rape.

A lot of Europeans seem to think so. Is USA?

[raymorris]

Just curious, do you think the United States is a country?
Decide and we'll get back to that in a moment.

> The american education is so bad that a large swath think that europe is a country.

Ignoring for the moment that your education apparently didn't teach you what proper nouns are, I've noticed that a lot of Europeans seem to think Europe is a country these days. Ask them where they are from and they say "Europe." They aren't completely wrong, Europe (EU) has a parliament that makes law and a high court that interprets law. Because the (former?) countries that make up the EU are no longer completely sovereign, they are, to some measure, no longer countries.

You may say "each member of the EU *consented* to delegating only *some* powers to the EU. Because only some power is delegated, and that by consent, each remains a country." Very well then, have a look at the enumerated powers clause of the US Constitution. It's a list of all the powers that the US federal (federation) government has, limited powers delegated by the consent of the individual states. Very much like the EU. If the US is a country because it has a federation government, then so is the EU - it too has a federation government, common currency, EU-wide elections, etc.

Re:A lot of Europeans seem to think so. Is USA?

[tepples]

One key difference as far as I'm aware is that the several states of the United States ceded their power to make treaties with other countries or to lay punitive import duties. (U.S. Const., article I, section 10) I was under the impression that the countries in the European Union retained this power.

Re:A lot of Europeans seem to think so. Is USA?

[sid crimson]

>>Just curious, do you think the United States is a country?

Yes, and "America" is one of three continents.
To describe someone as an "American" is kind of like calling someone from Libya an "African." BTW, if the USA isn't an country, then I'm a Californian.

Re:A lot of Europeans seem to think so. Is USA?

[raymorris]

> America" is one of three continents. ...
> I'm a Californian.

Indeed. California public schools, it seems.

Re:A lot of Europeans seem to think so. Is USA?

[Cederic]

I've noticed that a lot of Europeans seem to think Europe is a country these days. Ask them where they are from and they say "Europe."

That's a non-sequitur. I am from Europe, but Europe is not a country. I am not from the UK but the UK is a country. I am not from Scotland but Scotland is a country.

I am British and I am European.

This causes me no confusion at all, or those around me.

Re:A lot of Europeans seem to think so. Is USA?

[sid crimson]

Would you care to explain your criticism?

Re:A lot of Europeans seem to think so. Is USA?

[raymorris]

Reading my post again, I'm a bit ashamed because that was rude. Of course people are rude on forums all the time, but that doesn't mean I should be. I apologize.

To my understanding, Central America is not a continent. Geographers recognize five to seven continents (Eurasia and the Americas are sometimes considered one continent each). The first two sentences from the Wikipedia entry "Continent" are well written, so I'll quote them:

  A continent is one of several very large landmasses on Earth. Generally identified by convention rather than any strict criteria, up to seven regions are commonly regarded as continents. Ordered from largest in size to smallest, they are: Asia, Africa, North America, South America, Antarctica, Europe, and Australia

Wiki quote this definition of continent:

"continents are understood to be large, continuous, discrete masses of land, ideally separated by expanses of water."

Re:A lot of Europeans seem to think so. Is USA?

[sid crimson]

Thank you for the apology. My mis-worded response didn't warrant it, but my point still stands. "American" is a misnomer commonly used to describe us folks who live in the USA. I travel a lot, and on a particular trip an Egyptian I had met to work will called me "an American". I politely explained I'm from the USA, and no more "American" than he is "African". We've been close friends ever since that trip.

I feel happy. I feel happy.

[Thud457]

"I don't want to go on the cart. "

3 of 4 said Yes, and the 4th reply

[Provocateur]

Q. "Do you want to live?"
1st paralyzed man: "Yes."
2nd paralyzed man: "Yes"
3rd paralyzed man: "Yes"
4th man: F*** yeah! What the f*** kind of question is that!

Doctor: "I'll put that down as an affirmative"

I don't want to die.

[Anonymous Coward]

This bullshit about how we all have to grow old and die has got to fucking stop. Humanity need to quit dragging its feet and fucking cure aging. Our species has waited long enough. It is time for the dying to stop.

For god's sake, get on it!

Re:I don't want to die.

[Anonymous Coward]

Meanwhile, Steve Bannon drools over the prospect of hundreds of millions of heathens dead in the next world war.

Re:I don't want to die.

[micahraleigh]

You wouldn't believe how much opposition there is to this. Especially on echo chambers like \.

Newsweek's "The case for killing grandma", etc.

Re:I don't want to die.

[michael_wojcik]

Especially on echo chambers like \.

Huh? I looked at backslashdot.org, and didn't see anything about it.

Unless the EU decides to set the tariffs

[raymorris]

That's interesting. The EU seems to be under impression that they've negotiated many trade treaties limiting tariffs, so member states no longer have the freedom to set their own:

http://ec.europa.eu/trade/poli...

And of course there are about a hundred EU rules about what kind of tariffs states may and may not have.

Not that there are NO differences between the EU and the US, of course, but at their core they are essentially the same type of thing - the US federal government has just been around longer, so it's parliament (Congress) has had more time to make more laws.

Re:Unless the EU decides to set the tariffs

[tepples]

Then why does each EU member have a separate seat at the UN, while the USA doesn't? Is it only that the USA predates the UN and the EU doesn't? Perhaps they're waiting for Brexit so that the veto power vested in France's permanent seat on the Security Council can be passed to the EU.

no commas

[nobuddy]

Doctor: Do you want to live?

Patient: No, kill me
Machine: (NO KILL ME)

Doctor: This one wants to live!

interesting point

[raymorris]

That's an interesting point. If the EU continues to act more and more like a single country, it will be fair to ask why they have so many seats. The US doesn't have one for each state.

Nope

[kuzb]

If you're asking me a very final question like "do you want to live?" the machine better be right 100% of the time.

70% is not good enough.

Re:Nope

[Black Parrot]

70% is not good enough.

Probably good enough to serve for snake oil for grieving relatives.

(Which is presumably what it was designed for.)

Re:This is Impossible.

[micahraleigh]

Who are you to decide who contributes to society?

Shall we put the lives of people in the hands an anonymous coward?

Re:Really?

[micahraleigh]

It's hard for people on slashdot who live for pirated episodes of GOTR, philandering with their peers, and fantasizing about building moon bases to understand that life may in fact have something to offer to them personally.

If we're going to put anyone out of their misery it should be the murderers who want to euthanize anything that moves.

To Blave

[hoggoth]

"Doctor I have a reading"
"What does it say?"
"He distinctly said 'To Blave'."

What about that 4th patient?

[martinfb]

So, then, will that 4th, unhappy patient wishing to die - get euthanized?