Navigation and service

Revealing Brain Scans: AI Can Predict Personality

Jülich, 8 August 2018 – MRI image data can reveal information on a person’s personality traits. This was shown by scientists from Forschungszentrum Jülich and the University of Düsseldorf who published their findings in the June issue of the journal Brain Structure and Function. In their study, the scientists identified networks in the human brain that were particularly active during different tasks. They then trained machine learning software to assign this activity to particular personality traits. As the researchers showed in a different study, this process makes it possible to use MRI data to recognize whether a person has schizophrenia or Parkinson’s, for example.

Functional magnetic resonance imaging (fMRI) makes changes in the oxygen saturation in brain blood vessels visible within seconds. Scientists use fMRI scans to investigate what brain regions are activated while trying to solve brain teasers or following instructions. However, a person’s brain can also be scanned with this method while they let their thoughts run free. It was these scans that Jülich and Düsseldorf scientists used to estimate the personalities of test subjects – for example how emotionally unstable or conscientious someone is.

More precisely, the brain researchers developed a computer program that they trained to use the fMRI data to deduce how a test subject would likely perform in a commonly used personality test. This NEO-FFI test consists of 60 statements along the lines of "I try to be friendly to everyone." The test subject then rates how strongly they agree or disagree with the statement. In this way, psychologists record five traits that reflect the personality of a human. These "big five" are: openness to experience, conscientiousness, extraversion, agreeableness, and neuroticism.

Activity patterns in functional networks

Several steps were necessary until the computer was highly likely to make correct predictions: "We first conducted what is called a meta-analysis, in which we analysed many thousands of previously published fMRI studies where test subjects were given tasks," explains Prof. Simon Eickhoff, head of Jülich’s Institute of Neuroscience and Medicine – Brain and Behaviour (INM-7). He and his team identified a total of nine different functional networks in the brain. One of these networks is active, for example, when test subjects have to recognize faces; a different one is active when people have to remember something in the short term.

In the second step, the researchers used the anonymous data on roughly 700 test subjects who had undergone both a brain scan and a personality test. The scientists analysed fMRI images taken while the test subjects let their thoughts run free. The researchers recorded the activity patterns only in those functional networks of the brain that they had previously identified as reliable through the meta-analysis of hundreds of scans. Reviewing the literature therefore provided the scientists with the knowledge of the organization of the brain that they needed for their assessment.

Software trained to predict personality

The subsequent machine learning step involved the analysis of the results of 90 % of the test subjects. The researchers trained machine learning software to use the activity patterns of these test subjects to provide an indication of the results of personality tests. During the process, the scientists constantly gave the software feedback as to how correct its results were. The machine learning software then adapted its mathematical model. "If you keep repeating this process many times, the model will become better and better," says Eickhoff. After the software was trained in this way, it ultimately predicted the personalities of the 10 % of test subjects whose NEO-FFI test results it did not know.

This revealed that the activity of individual, specific networks is directly linked with how pronounced certain personality traits are. Two of the nine functional networks proved to be suitable for predicting a person’s NEO-FFI results with respect to their openness to new experiences. One functional network each allowed a person’s agreeableness and neuroticism to be predicted.

Opportunities and limitations of technologies

Interestingly, other networks can be used to predict further personality traits – when the data on men and women are separated and inputted into the machine learning software separately in order to train it. "The patterns of connectivity strengths in the functional networks are mostly gender-specific," concludes Eickhoff.

These findings of the Jülich and Düsseldorf scientists are significant for basic research. "We not only want to understand the basic structure of the human brain, but also how brains differ from one person to the next," explains Eickhoff. "Of course, there are always concerns when it comes to this kind of research: from the danger of a transparent society to economic consequences, or even the potential abuse of patient data. This is why it is especially important to discuss the opportunities and limitations of such technologies in a transparent manner."

Prof. Simon EickhoffProf. Simon Eickhoff
Copyright: Forschungszentrum Jülich / Sascha Kreklau

fMRI data can also indicate mental illness

The brain researchers headed by Eickhoff together with other scientists had previously shown in a different publication that the combination of meta-analysis and machine learning also permits conclusions to be drawn on whether a person has schizophrenia or Parkinson’s disease, or is mentally healthy – using only fMRI images taken while test subjects let their thoughts run free. Using their methodology, the scientists had identified specific functional networks that are impaired by the respective illness. These results were published in the journal Human Brain Mapping. In addition to brain researchers from Jülich and Düsseldorf, scientists from JARA-BRAIN (Jülich-Aachen Research Alliance), RWTH Aachen University, and the universities of Duisburg-Essen, Heidelberg, and Cologne were also involved in this study.

Original publication: "Predicting personality from network-based resting-state functional connectivity", by Alessandra D. Nostro, Veronika I. Müller, Deepthi P. Varikuti, Rachel N. Pläschke, Felix Hoffstaedter, Robert Langner, Kaustubh R. Patil, Simon B. Eickhoff, Brain Structure and Function (2018),
DOI: 10.1007/s00429-018-1651-z

Further Information:

Interview with Prof. Simon Eickhoff "Brain Scan for Individual Prognosis"
The team headed by Prof. Simon Eickhoff analyses MRI and fMRI brain scans of sometimes hundreds of people in a very specific way: they train computers to determine from image data the activity patterns in the functional networks of the brain. Eickhoff hopes to use this information to predict the further development of illnesses in people with depression, schizophrenia, or Parkinson’s disease based on this information.

Institute of Neuroscience and Medicine - Brain and Behaviour (INM-7)

Twitter account INM-7


Prof. Dr. Simon Eickhoff
Institute of Neuroscience and Medicine - Brain and Behaviour (INM-7)
Forschungszentrum Jülich
Tel.: 02461 61-1791

Press contact:

Dr. Regine Panknin
Corporate communications
Forschungszentrum Jülich
Tel.: 02461 61-9054







Dear visitor,

To make our website suit your needs even more and to give it a more appealing design, we would like you to answer a few short questions.

Answering these questions will take approx. 10 min.

Start now Close window

Thank you for your support!


In case you have already taken part in our survey or in case you have no time to take part now, you can simply close the window by clicking "close".

If you have any questions on the survey, please do not hesitate to contact:


Your Team at Forschungszentrum Jülich


Note: Forschungszentrum Jülich works with the market research institute SKOPOS to anonymously conduct and analyze the survey. SKOPOS complies with the statutory requirements on data protection as well as with the regulations of ADM (Arbeitskreis Deutscher Markt- und Sozialforschungsinstitute e.V.) and ESOMAR (Europäische Gesellschaft für Meinungs- und Marketingforschung). Your data will not be forwarded to third parties.