A team of neuroscientists have developed a system that allows them to read a person’s intentions before they even act upon them. It sounds amazing; but their results are not too impressive.

The report came from the journal, Current Biology, by Professor John-Dylan Haynes from the Max Planck Institute for Human Cognitive; working with other word-leading neuroscientists, Hayes claims the system can recognize a person’s intentions with 70 percent accuracy based solely on their brain activity.

The team used functional magnetic resonance imaging (FMRI) along with a computer programmed to recognize specific patterns in the brain that usually occur in conjunction with specific thoughts. It’s a neat concept, but there are too many problems and way too much hype over this small “breakthrough.”

“Using the scanner, we could look around the brain for this information and read out something that from the outside there’s no way you could possibly tell is in there,” said Hayes, “you need to observe people thinking a certain number of different kinds of thoughts before you can determine what patterns of brain activity are linked with different thoughts.”

Test subjects were brought into the lab and hooked to these devices. They were asked to choose between 2 different mathematical tasks, adding or subtracting, in which they were to keep a secret; after focusing on their intention to add or subtract for nearly 10 seconds, they were shown a screen with a pair of numbers on it.

After the pair was added or subtracted in the subject’s head, they were then shown a screen with 4 numbers on it; one being the correct answer for addition, one being the correct answer for subtraction, and the other 2 being wrong answers. I would be a mind reader if I knew why they displayed 2 wrong answers; especially if they were trying to receive an accurate result.

In any case, the test subjects were to select the right answer on the screen; this process continued for 40 minutes before they switched the system over from “learning” the intentions to “repeating” the intentions.

Hayes claimed that after 40 minutes of learning each individual’s intentions, the computer was able to guess their intention to either add or subtract 2 numbers before the numbers were even displayed on the screen within 70% accuracy. I usually try to be optimistic about things, but there is just not enough “wow” in this experiment to excite me; I’ll tell you why.

First, I could make a program to give you this same result with 50% accuracy; there are 2 options: add or subtract, yes or no, on or off. Probability tells me that in 2 cases, given enough guesses, I could read your mind half of the time. 70% is not that far off, and the study does not mention how many guesses it took the computer to achieve this 70%; it’s important.

Second, every brain is a snowflake; no 2 brains are exactly alike. It is for this reason that they had to study each test subject for 40 minutes to enable them to perform this calculation based on that specific person. It would be unreasonable to assume that a computer could read any person’s intentions without learning their specific brain’s style. Had the person been dishonest during the initial “learning,” the computer may have shown inaccurate less than 50% of the time.

“It’s important to see if we can further increase the accuracy,” Haynes said, adding that it may have been more accurate if the computer was trained for a longer period of time. This may be the case; but for this technology to be used to help the disabled control robotic prosthetic limbs, it would need to predict hundreds, if not thousands, of variables.

Professor Colin Blakemore, director of the Medical Research Council, took a more realistic approach noting that, “We shouldn’t go overboard about the power of these technologies at the moment.”

Related:

Entry was posted on Friday, February 9th, 2007 at 4:36 pm under Science and technology.

Read more entries by George Gardner.
Stumble It!