EERPD: Leveraging Emotion and Emotion Regulation for Improving Personality Detection

TL;DR

EERPD addresses the challenge of automatic personality detection by infusing emotion regulation knowledge into a retrieval-augmented framework. It splits input text into Emotion Sentences and Emotion Regulation Sentences, retrieves two similar examples with CoTs via a reference library, and uses an LLM with psychological prompts to predict personality traits. The method demonstrates substantial improvements over strong baselines on MBTI and Big Five datasets, highlighting the value of incorporating emotion regulation in NLP for personality inference. The work showcases how psychological insights can guide LLM reasoning, offering a practical approach to more robust, few-shot personality detection with broad implications for HCI and personalized applications.

Abstract

Personality is a fundamental construct in psychology, reflecting an individual's behavior, thinking, and emotional patterns. Previous researches have made some progress in personality detection, primarily by utilizing the whole text to predict personality. However, these studies generally tend to overlook psychological knowledge: they rarely apply the well-established correlations between emotion regulation and personality. Based on this, we propose a new personality detection method called EERPD. This method introduces the use of emotion regulation, a psychological concept highly correlated with personality, for personality prediction. By combining this feature with emotion features, it retrieves few-shot examples and provides process CoTs for inferring labels from text. This approach enhances the understanding of LLM for personality within text and improves the performance in personality detection. Experimental results demonstrate that EERPD significantly enhances the accuracy and robustness of personality detection, outperforming previous SOTA by 15.05/4.29 in average F1 on the two benchmark datasets.

Figures (9)

• Figure 1: Examples for emotion and emotion regulation sentences. Emotion sentences tend to contain words that are experienced in the short term, while Emotion Regulation sentences tend to contain features that are stable in the long term.
• Figure 2: An overall framework of EERPD. The sentences in input text is categorized into Emotion Sentences and Emotion Regulation Sentences, and then are vectorized and proportionally combined. Using the new vectors, we retrieve two examples and generate their corresponding CoT processes. These examples, along with the input text, are then fed into the LLM with psychological knowledge to obtain the final prediction.
• Figure 3: An overview of prompt in our method.
• Figure 4: Impact of Alpha on Kaggle and Essays.
• Figure 5: The Performance of Different Retrieval Models on Kaggle dataset.