Predicting Human Errors From Brain Activity

Hugh Pickens writes "Researchers report that brain activity can be used to predict the likelihood of someone making an error about six seconds in advance, with gradual changes starting as much as 30 seconds ahead of time. The team used an imaging machine to scan the brains of a group of volunteers who performed a task in the presence of distracting information. When performing correctly the volunteers' brains showed increased levels of activity in those parts associated with cognitive effort, as would be expected. However, these areas gradually became less active before errors were made and at the same time another set of regions in the brain became more active. These regions are part of a so-called "default mode network" and show increased use when people are resting or asleep [PDF]. While imaging machines are far too big and complex to be used in workplaces to monitor the brain activity of people engaged in important tasks, the team hopes to correlate errors to changes in electrical activity in the brain with electroencephalography (EEG), using electrodes placed on the scalp. If EEG features can be found that correspond to the change in brain activity, then a hat that gives warning of an imminent mistake might one day become reality. We've previously discussed similar studies of brain activity."

But can it predict

[Anonymous Coward]

when the editors will err and post a dupe [slashdot.org]?

Re:I agree

[Anonymous Coward]

I hate pdfs, so here is a screencap of the brain imaging picture from the pdf:

http://img105.imageshack.us/img105/1130/brainimagingwz8.jpg [imageshack.us]

Re:Seems obvious to me

[amRadioHed]

It sounds to me that the key finding is that they can detect when a person is distracted by monitoring their brain waves. A much more interesting finding than that distracted people make errors.

Re:Errors?

[amRadioHed]

The device actually predicts when you are not concentrating on your work, and that is what people would be told if they ever were to use such a thing in production. I think the notice that you are losing concentration would probably be enough to get you concentrating again on your work. Especially if the workers were penalized for time spent not concentrating.

Re:uh oh

[billcopc]

I can make an EEG monitor that IS a baseball cap. By carefully measuring the angle of the visor relative to the wearer's head, as well as a few environmental factors such as precipitation, temperature and altitude, I can identify half-bred imbeciles with shocking accuracy.

I can also tell with 100% certainty that a person wearing a ballcap will make a mistake. That mistake is asking me to fix their computer.

Re:I don't believe it

[venicebeach]

Basically this particular task (the "flankers task") is so easy that people only make errors when they stop paying attention. You see something like SSSHHSSS or HHHSSHHH and you have to respond to the central letters while filtering out the outside letters. So what is essentially happening is they are measuring when people's attention wanes and errors happen to be reliably associated with this.

Keep in mind that in this study errors are not actually predicted before they happen in real time. That's virtually impossible with fMRI. What's happening is that after the subject completes the task, the experimenters find data which comes from before the button press and reliably predicts subsequent errors. This happens in post-processing, so the experimenters do not actually know what is going to happen before it happens.

EEG, Prediciting and Probability

[DynaSoar]

"Prediction" is not accurate because that implies an absolute. The activity correlates with an increased probability of making a mistake. The study relies on statistics throughout, from the analysis of the fMRI data on, and so can only deal in probabilities.

An Israeli team found that an increase in degree of synchronization of midline frontal theta EEG varied inversely with the probability of making a mistake. Such theta synchronization occurs over spans of 10 to 30 seconds. They also found that when a response occurs during the rising or falling slope of the synchronized theta (as opposed to near a peak), the person was more likely to make a mistake. The latter probably is the source of the evoked potential called the Error Related Negativity; it is the brain preparing to notice the error. The former seems to indicate a lagging in attention, which is when errors are most likely to occur. The two are related, meaning the brain "knows" when it is starting to droop and is more likely to make a mistake, and tells itself to get ready to notice a mistake if it happens.