Digital footprints of personality: Large-scale analysis reveals quantitative links between MBTI traits and emotional expression on Reddit

Liu Xingcai(Institute of Psychology of the Chinese Academy of Sciences)

Abstract

The intersection of personality psychology and computational social science offers unprecedented opportunities to understand human behaviour at scale. However, the precise link between stable personality traits and an individual’s online emotional expression remains unquantified across large, naturalistic datasets, with research often bifurcating between complex, "black-box" predictive models and small-scale qualitative studies. Here, we bridge this divide by applying a computationally efficient, lexicon-based framework to a massive dataset of 2,005 Reddit users who publicly self-declared their Myers-Briggs Type Indicator (MBTI) personality type. We analysed the aggregated emotional content of their comment histories, controlling for user activity. We find robust, statistically significant associations between personality dimensions and distinct emotional footprints. The Thinking (T) versus Feeling (F) dimension emerged as the most potent predictor of average sentiment polarity, with ’Thinking’ types exhibiting significantly lower average sentiment polarity (β=-0.061, p<1.5e-19). We also find that ’Thinking’ types display significantly lower emotional volatility, measured by the standard deviation of their sentiment scores (β=-0.008, p=0.038). Contrary to our initial pro-social hypothesis, Extroversion (E) was not significantly linked to specific’joy’or ’trust’ frequencies in the full model, but did correspond to a small but significant increase in overall positive polarity (β=0.016, p=0.032). Furthermore, the Judging (J) dimension unexpectedly emerged as a strong, positive predictor of both overall sentiment (β=0.023, p<0.001) and ’joy’ expression (β=0.079, p=0.003). These findings demonstrate that foundational personality traits leave distinct, quantifiable imprints in digital language and that trait-based emotional expression is a multi-dimensional phenomenon where decision-making (T/F) and planning (J/P) traits can be as, or more, influential than social (I/E) traits.

Keywords

MBTI personality traits，Online emotional expression，Digital footprints，Text analysis，Lexicon-based framework

References

A, A., et al. (2023). Deep emotion recognition in textual conversations: a survey. Artificial Intelligence Review, 56:11333–11406.

A, B., et al. (2021). Ai knows you: Deep learning model for prediction of extroversion personality trait. IEEE Access, 9:118464–118475.

B, R., et al. (2019). Personality prediction with social behavior by analyzing social media data – a survey. International Journal of Advanced Computer Science and Applications, 10(6):324–332.

C, L., et al. (2019). The persuasive power of emoticons in electronic word-of-mouth communication in social networking. Journal of Business Research, 101:50–65.

C, S., et al. (2021). Big data analytics on social networks for real-time depression detection. Journal of Big Data, 8:111.

H, A. A., et al. (2021). Beyond sentiment analysis: A review of recent trends in text based sentiment analysis and emotion detection. Journal of Big Data, 8:83.

H, L., et al. (2023). Large language models can infer psychological dispositions of social media users. Nature Communications, 14:5678.

I, W. W., et al. (2022). Text based personality prediction from multiple social media data sources using pre-trained language model and model averaging. Journal of Big Data, 9:109.

L, F., et al. (2017). Emotion expression on social networking sites: A study of young persons’ use of facebook and twitter in the uk. Journal of Psychological Research and Behaviour Management, 10:247–255.

M, A., et al. (2020). A hybrid deep learning technique for personality trait classification from text. IEEE Access, 8:175960–175971.

Musk, D. R., & K., A. (2023). Mental health analysis in social media posts: A survey. ACM Computing Surveys, 55(1):1–38.

N, A. A., et al. (2021). Big five personality prediction based in indonesian tweets using machine learning methods. PeerJ Computer Science, 7:e521.

DOI: http://dx.doi.org/10.26549/mmpp.v7i1.36167