Summary: Researchers have demonstrated that AI may identify personality traits from written words and, crucially, have now learned how these types make decisions. The team discovered how precise words and language patterns influence predictions based on significant psychological frameworks by using explainable AI techniques like integrated gradients.
The research found that Big Five faculties are more consistently identified than MBTI forms, with the latter matching more favorably with verbal behavior indicators. These reassurances may help to create clear, moral character assessments in fields like psychology, HR, education, and online platforms.
Important Information
- Observable AI: The “black box” of AI decision-making was opened by integrating variations to find out which words had an impact on personality predictions.
- The Big Five vs. MBTI comparison: The Big Five type was more accurate and cognitively grounded than the MBTI for AI-based character study.
- Functional Application: The findings could improve the use of medical assessments, personalized education, HR policies, and adaptive AI assistants.
University of Barcelona
For the first time ever, a research team at the University of Barcelona has analyzed how artificial intelligence ( AI ) models can identify personality traits from written texts in depth.
These findings, which were published in the journal PLOS ONE, provide new insights into how to create more accurate and clear automated detection tools and how to interpret personality as it manifests in natural language.
The Individual Differences Lab Research Group ( IDLab ) of the Faculty of Psychology and the Institute of Neurosciences ( UBneuro ), as well as Daniel Ortiz Martnez, researcher at the Faculty of Mathematics and Computer Science, are the authors of the paper, which is signed by three UB experts: David Saeteros and David Gallardo-Pujol, researchers and directors, respectively.
Opening the “black box” of techniques
The Big Five personality trait system, which includes openness to experience, responsibility, extraversion, agreeableness, and emotional stability, as well as the Myers-Briggs Type Indicator ( MBTI), an instrument that categorizes people according to the traits extrovert-introvert, sensory-intuitive, thinking-feeling, and thinking-feeling, are how two advanced AI models, BERT and RoBERTa, process text data to identify personality characteristics.
The researchers describe these two mental frameworks as “in psychology, there is a common model of personality and other less well-validated models, which we use to comprehend and assess individual differences in behavior, emotions, and thinking.”
The texts used in the study’s analysis were derived from two databases that had previously been categorized based on the presence of indicators for the various personality types and forms that make up the designs ( Big Five and MBTI).
In the future, experts have used observable AI techniques to observe the Artificial models and examine how language patterns affect the recognition of character traits in these writings.
The authors note that “explainability techniques allow us to “open the dark package” of algorithms, which makes sure projections are based on psychologically important signals and not on objects in the data.”
They exclusively used a method known as , integrated variations, which enables them to precisely predict which words or phrases are involved in a particular personality trait.
According to them,” This strategy has allowed us to imagine and calculate the significance of various linguistic components in the woman’s predictions.” For instance, they have observed that words like , hate, which would ordinarily be associated with negative traits, can appear in settings that actually reflect kindness (” I hate to see others suffer” ).
We perhaps draw the wrong conclusions without understanding how the design understands these words in context, they say.
This method ensures that AI models ‘ performance is scientifically accurate, he adds,” by ensuring that they are based on language patterns that are truly related to the mental constructs they are intended to estimate” and provides a solid foundation for future improvement.
The MBTI model’s limits
The Big Five model, which provides a stronger foundation for both automatic character analysis and traditional diagnostic analysis, has its advantages over the MBTI model, which the study also highlighted.
The MBTI model has significant limitations for automated character evaluation, they claim, noting that its findings tend to heavily depend more on objects than on actual patterns, despite being extensively used in computer technology and some applied psychology fields.
applications of involuntary personality testing
The integration of AI models and automated personality detection techniques can have a significant influence on the field of character psychology.
” Psychologists will discover linguistic patterns that are related to various personality traits that, using conventional methods, might go undetected,” according to these methods. The experts note that this can lead to more accurate and less aggressive examination techniques, which are particularly useful for studying large populations.
The authors in the medical industry claim that they can assist with “initial analysis and follow-up of patients by focusing attention on changes in language or linguistic expression as indicators of significant emotional elements for therapy.”
They also point out that they can be significant in various fields, such as personnel selection, education personalization, social research, or the development of linguistic agents and online assistants, because it would enable more natural and adapted interactions.
To ensure ethical and transparent use, they add,” It is important to emphasize that all such applications should be based on scientifically sound models and incorporate the explainability techniques we have explored.”
Despite their potential, researchers believe that these models will complement traditional personality tests and provide an additional and more in-depth perspective rather than replace them in the near future.
We are seeing an evolution toward a multimodal approach, where traditional assessments are combined with natural language analysis, digital behavior, and other data sources to obtain a more accurate picture of personality, they say.
The researchers believe that this integrative approach will “build on the strengths of each methodology” and give a “richer and more nuanced view of the human personality.”
In this context, AI models may be “especially useful in situations where traditional data collection is challenging or when large amounts of information need to be analysed effectively,” they add.
Validating research in other settings
Expanding the analysis to other text types, platforms, languages, and cultures is the next step in this study to check whether the patterns found are accurate in all different contexts. The researchers want to look into the potential use of these methods to psychological constructs other than personality, such as emotional states or attitudes.
Researchers are also experimenting with using technologies like automatic audio transcription ( Whisper ) to combine text with other forms of expression, such as voice or non-verbal behavior, and to incorporate multimodal data into these analyses. ai ) as well as their application in everyday life.
The team wants to” collaborate with clinicians and human resources professionals to evaluate the efficacy of these tools in real-world settings, ensuring that they have a positive and ethical impact,” they write in their report.
About this information on personality research and AI
Author: Rosa Martnez
Source: University of Barcelona
Contact: Rosa Martínez – University of Barcelona
Image: The image is credited to Neuroscience News
Open access to original research
David Saeteros and others ‘” Natural language processing’s insights into personality.” PLOS One
Abstract
Natural language processing’s insights into personality
Advancements in natural language processing ( NLP ) have made it possible to evaluate people differently in recent years.
This article uses interdisciplinary research that uses interdisciplinary techniques, particularly Integrated Gradients, to examine the decision-making processes of NLP models in personality evaluation and verify their compatibility with established personality theories.
Using the Essays and MBTI datasets, we use typological ( MBTI ) and dimensional ( Big Five ) models to compare their effectiveness.
To classify personality traits based on textual data, we use log-odds ratio, Informative Dirichlet Prior ( IDP ), and fine-tuned transformer-based models ( BERT and RoBERTa ).
Our findings show that personality prediction accuracy ranges from moderate to high, with NLP models accurately identifying personality signals in text, in accordance with previous research.
Our findings highlight significant biases in the MBTI dataset that led to less robust results while revealing theory-coherent patterns in language use in relation to various personality traits.
The study emphasizes the value of further interdisciplinary research to fully realize the capabilities of these transparent technologies and the potential of NLP in improving personality psychology.
