Wouldn’t it be great to read others’ minds? Glimpse into their brains and see what they think of when they scroll your website, taste the latest dish on your menu, or listen to your latest piece of music?
Electroencephalography (EEG) is a biometric method that can give you fairly detailed information on ongoing brain activity associated with perception, cognition, and emotion.
However, simply attaching electrodes to someone’s head and recording data won’t be sufficient to get the desired outputs. Because EEG signals as recorded from the surface of the scalp originate inside the brain, data analysis and interpretation can get quite complex, requiring expert knowledge of brain structures and function.
However, you don’t need a PhD in Neuroscience to extract meaningful information from EEG. While EEG headsets record raw voltages, computer algorithms exist that transform these voltages instantaneously into derived cognitive metrics such as workload, motivation, or drowsiness, allowing you to evaluate the efficiency of your stimuli. But how is this fascinating transformation accomplished?
As we start with raw EEG data, all we have is a time series of electric potentials picked up from each electrode site. On screen, EEG data actually looks quite similar to audio recordings. Audio recordings can be messed up if the microphone is moved or by surrounding noise (e.g., the nearby airport, a roommate sneezing), and EEG data is comparably corrupted by electromagnetic and motion artifacts.
For example, line noise from power outlets introduces noise to the EEG data, swaying arms or other extremities is not recommended, and EEG cap or electrode movements can cause serious artifacts in the data.
In order to attenuate these artifacts and to amplify the “meaningful” signals, the raw EEG data has to be processed in several steps. Decontaminated EEG data is raw data that has had filters applied to attenuate the artifacts. Filters can be applied in time to make the signal time course look smoother (e.g., to get rid of high-frequency spikes or low-frequency data drifts).
Additionally, spatial filters can be applied (e.g, spline filters to generate the Current Source Density of the data). Other filters attenuate specific artifacts generated by blinking or vertical eye movements.
Derived EEG metrics are built from decontaminated data, using specific bandwidths (signal types – see this blog post) in specific electrode regions that are associated with certain higher level cognitive processes or mental states.
The “frontal asymmetry index” is one of the most prominent states (see here for more info), but several companies offer further out-of-the-box metrics (for example Advanced Brain Monitoring or Emotiv), each differing slightly dependent on the electrode positions involved as well as the computer algorithms and procedures used to extract the metrics. The most common and frequently used live metrics are:
Workload increases with increasing working memory load, for example if we’re told to count backwards from 101 in steps of seven. Also, workload increases during problem-solving, mental arithmetic, integration of information, and analytical reasoning.
Workload reflects a sub-category of executive functions necessary to keep us going and focus on important activities. EEG-workload levels are significantly correlated with both objective performance and subjective workload ratings in tasks with varying levels of difficulty including forward and backward digit span, mental arithmetic and n-back working memory tests (Berka, 2006).
Workload is the key metric for any usability study where it is used to assess user interface design.
The engagement index is related to processes involving information-gathering, visual scanning and sustained attention. Engagement is increased as respondents have to allocate their attention resources to the encoding and processing of auditory, visual or haptic stimuli.
This metric is an excellent candidate for measuring stimulus effectiveness and to answer questions such as, does your advertising work and do people like your packaging?
Distraction / Drowsiness
Distraction is a notion of the subject’s being involved somewhere other than the cognitive tasks of interest. This has involved instances of frustration, boredom or confusion.
One component may be distraction caused by background environmental noises. Another component may be the distraction associated with a subject making a mistake. This is often followed by a period of increased engagement. Sleep states and near sleep states have been investigated using EEG for sleep apnea and other sleep disorders.
Monitoring drowsiness is useful for any system where missing the detection of rare events can be very cost intense (e.g., controllers preventing midair collisions).