An interview
We sat down with Björn Hommel during ECP20 to talk about his talk, titled “The Effectiveness of High-Dimensional Semantic Construct-Vectors in Measuring Personality Traits with Open-Ended Questions”. Björn is a research associate at the Department of Work and Organisational Psychology at Leipzig University.
Read on to learn more about Björn’s background and his other interests!
Q: Can you tell us a bit about yourself and what made you become interested in personality psychology?
During my time as an undergraduate student, I was very interested in forensic and criminal psychology. As a result, my thesis on spatial navigation and psychopathy was quite elaborate, and it involved conducting an experiment in a correction facility in Germany. There were lots of bureaucracy to go through and it was a long tedious process – there were certainly easier ways to get the degree!
There was one particular moment when I changed my mind about what I wanted to do in psychology. It was when a decision was made about the eligibility for parole in one individual case, as that decision was largely based on a specific psychometric assessment inventory. The inventory was still in its unrevised original form from the 70s, and information about the instrument’s quality criteria was measured only using internal consistency. There was nothing about validation. The supervising psychologist of the facility had no issue with the lack of validation and continued to administer it. I thought that was really appalling, as two years of someone’s life were determined by a questionable psychometric test. That was the turning point for me when I decided to study more about measurement. After receiving my Master's degree, I founded magnolia psychometrics – a psychological consulting agency to bring a better standard for testing and decision-making into hiring decisions.
Q: What do you like to do outside of work?
I'm trying to divide my life into more distinct times of intense work and then go travelling for longer periods of time and just not doing anything work-related. But since the pandemic, to be honest, this has been difficult. As I have the consulting agency as well as doing my PhD, there have been so many things going on, so many fires that we have to put out. So, in recent times I haven't even really found time for exercise. I guess I have to catch up with my personal life!
Q: Can you give us a summary of the talk you gave at ECP20?
I think we can agree that rating scales have established themselves as the top response format for instance when we assess personality traits. The rating scale (i.e., subjects indicate the extent to which they agree with a statement presented) has many advantages, especially when compared to unstandardized response formats, such as open-ended questions. One major disadvantage of open-ended questions compared to rating scales is that the former is notoriously difficult to score and the traditional way for psychologists to approach this is to have raters agree on the response to freely written texts. However, this often results in low objectivity, reliability, and validity. Another way to process open-ended responses is with natural language processing. By most accounts, the history of natural language processing dates back to the 50s. But there have been some important recent innovations in the field of natural language processing (NLP). In particular, the integration of the attention mechanism in the transformer model architecture proposed by Vaswani and colleagues, in a very influential paper really was the turning point for natural language processing. The authors show that transformer models excel at a wide range of NLP-tasks, such as generation, summarization, and translation of text, to name a few. My co-authors and I provide a more in-depth introduction to these types of models in our article “Transformer-Based Deep Neural Language Modeling for Construct-Specific Automatic Item Generation”, which was published in Psychometrika earlier this year. Transformer models convert words and sentences to semantically meaningful vectors that allow for mathematical operations.
In my current work, which I conduct with Philipp Schäpers and August Nilsson, we use these vectors – so-called “embeddings” – to score freely written responses against vectors that represent various personality traits. Initial results look very promising and show that we indeed were able to identify trait expressions in written text. Using our method, we found substantial correlations between participants' freely written responses to open-ended items (i.e., “When someone has badly wronged me, I cut most, if not all, contact with that person.”) and their scores on the HEXACO personality inventory (HEXACO-PI-R 60). Additionally, these responses show correlations to self-reported behavioural acts that we expected to be linked to specific personality traits (e.g., “How often have you complained about service in a restaurant in recent times?”).
Q: What advice would you give younger researchers?
I cannot give an expert opinion on that because I'm still doing my PhD of course! But I changed the topic of my dissertation twice in the last three years of the PhD. It was always in the vicinity of psychological measurements, but my previous research topics were more conventional as compared to the interdisciplinary work I do now. I found it really difficult to motivate myself, especially when confronted with the different obstacles that we have in psychological research, and started to question the meaningfulness of my own work. I think most PhD students have experienced these doubts.
Q: Ah yes, the PhD blues!
Yeah, the term seems accurate! I just threw away all of my previous work because there were too many methodological issues. Instead, I started to think about what I want to see happen in psychology in the near future. If somebody who is not from the field of psychology asks about my work, I want to feel that the contribution of my research is obvious to them. It’s much more fun, of course, to work on something that you find meaningful.
Thinking out of the box is sort of cliché, but it helped me to just take a step back from the state of affairs in psychology and psychological measurements and to try to find different directions. I find it really rewarding.
Q: Can you tell us a bit about the scholars and researchers that have inspired you and your work?
Back when I was an undergraduate student and interested in criminal psychology, there were a couple of scholars that I admired for their contributions to the field. Today, I feel more inspired by people who I was fortunate to learn through personal collaboration. Each of these researchers has different strengths that I appreciate and attempt to adopt.