A new study shows machine learning can be used to predict aspects of a person’s personality from their electrical brain rhythms by Hayley Jach, Dr Daniel Feuerriegel and Associate Professor Luke Smillie, of University of Melbourne, Australia (first published in Pursuit).
As we think, fear, love and dream – 86 billion neurons are firing electrical and chemical signals at one another in a complex cascade of information exchange.
This brain activity must somehow represent all parts of a person’s psychology, including their personality. And our new study published in Science Direct (peer reviewed journal) shows that we can predict some personality traits from the electrical rhythms generated by our brains.
What are personality traits?
Put simply, personality traits are descriptions of what we are like. Are you generally quick to anger or even-tempered? Are you shy or more outgoing?
Of course, people will behave differently in different situations — even the most buzzing extravert enjoys a little quiet time now and then. Our thoughts, emotions, and behaviours are shifting constantly throughout time, just like the temperature shifts from day to day.
Although Melbourne gets heat waves and Brisbane can (sometimes) get chilly, it’s still reasonable to say, “Brisbane is a warmer place than Melbourne”.
A similar consistency can be found in personality.
For example, people who say they are polite will not be polite at all times in every situation, but on average, they will be a little more polite across more situations than someone who says they are impolite.
Because related traits cluster together, people who are polite also tend to be compassionate and social. Taken together, these traits describe how agreeable someone is. Agreeableness is one of the ‘Big Five’ personality traits that represent the main ways people differ from one another.
The other four are neuroticism (describing anxiety, depression and distress), extraversion (enthusiasm and assertiveness), conscientiousness (achievement-orientation and neatness), and openness to experience (curiosity, creativity and intellect).
How do we measure brain rhythms?
We can measure rhythmic brain activity using electroencephalography, or EEG. This technique uses an array of electrodes placed on the scalp to record the transient electrical activity generated by neurons (brain cells). Imagine surrounding your head with dozens of tiny microphones that ‘listen’ to the brain.
Any process over time will have a rhythm. Think of your daily routine, which starts and ends in bed and follows a relatively similar pattern each new day. The rhythm of your activity echoes the rhythm of the sun rising and coffee brewing. Brain activity works in the same way, with various repeating cycles.
With EEG, we can measure how frequently the cycles reappear over the period of, say, one second, and this is called the frequency of the rhythm. For instance, alpha waves describe frequencies of between eight to 12 cycles per second.
Another important concept is the energy in a frequency band. For example, two people may brew coffee at the same time each day, but one of them may use four scoops of coffee and the other person only one.
The first coffee has more energy than the second coffee. In a similar way, the alpha waves of one person may have more energy than the alpha waves of another person.
So, EEG reveals how much energy there is across many different frequencies in the activity of the brain.
How did we decode personality?
As part of our research, we initially asked people to rate their personality traits using a well-validated questionnaire. Then, we measured their brain rhythms with EEG while they sat quietly.
We then trained statistical models to predict personality traits based on the EEG data. We fed the model an EEG ‘data soup’ of the patterns of energy across multiple frequencies and trained the model to associate each person’s data soup with their personality scores.
The model was then fed some EEG data without personality score data, and asked the model to predict their personality scores.
What did we find? Well, out of the five personality traits, we could consistently predict how agreeable and neurotic people were.
What does this mean?
Why can machine learning decode a person’s agreeableness and neuroticism? Clearly, personality is not simply written in our brain waves, like a list of ingredients on a tub of hummus.
In general, the brain rhythms that we are decoding could be representing two things.
Firstly, we could be decoding a momentary process, such as a particular state of mind or pattern of behaviour. For example, agreeable people may have felt happy to help with our research and contribute to scientific progress. These momentary positive emotions and thoughts may have been picked up in the EEG recordings.
Secondly, we could be decoding some longer-lasting process; perhaps agreeable people have different patterns of brain rhythms from non-agreeable people in a stable way that isn’t tied to their momentary emotions or thoughts.
Figuring out which of these is correct has important theoretical and practical applications.
For instance, if we can find evidence that brain activity is linked to momentary processes, but not long-lasting processes, then this might suggest that agreeableness and neuroticism ‘emerge’ out of many small interactions between the brain and the environment.
And this may help us develop novel, dynamic theories about the neural foundations of personality.
It is also important to note that neuroticism is a risk factor for developing depression, anxiety, and many other clinical disorders. If we find that we are decoding stable processes for neuroticism, then this could help to establish ‘biomarkers’ to assist in the diagnosis and treatment of these disorders.
Of course, future research is needed to shed light on these possibilities.
Our study is one of the first to show that we can predict personality from brain rhythms. As exciting as this is, it is still an early step in our mission to identify the processes in the brain that make us the way we are.