Exploring the Impacts of HEXACO Personality Traits on Text Composition and Transcription

Research Background and Issues

• Problem and Importance: The authors investigate how personality traits (HEXACO model) influence composition and transcription behaviors in text input. This topic is significant because, although text input activities (e.g., typing, writing) are frequent in modern life, the relationship between text input and personality has rarely been studied. Particularly in freewriting, the impact of personality may differ from guided tasks such as academic writing.
• Research Motivation:
  1. Using text input behavior as a non-invasive method to infer user personality can aid in dynamically personalizing computer systems.
  2. Personalized design tailored to personality traits can enhance user experience.
  3. Existing studies in academic writing have emphasized the link between personality and performance, but these are mostly limited to student contexts and fail to directly map to broader text generation behaviors.

Solution

• Methods and Innovations: The study employs the six-factor HEXACO personality model, which adds the Honesty-Humility dimension to the traditional Big Five model for a more comprehensive capture of personality traits. The experimental design differentiates text input behaviors into two types:

  1. Composition Task: Participants engage in freewriting, with metrics recorded such as writing time, word count, editing behaviors, pause frequency, and readability.
  2. Transcription Task: Participants transcribe standard short sentences, with input speed, accuracy, and error correction rates assessed.

  Innovations include analyzing tasks in a strictly controlled laboratory environment to eliminate potential device-related influences present in field experiments.

• Key Techniques:

  • HEXACO-PI-R scale for measuring personality traits.
  • Custom text input application to collect real-time behavior data during transcription and composition.
  • Dale-Chall formula for calculating text readability, alongside multiple linear regression and Spearman correlation analysis to evaluate relationships between variables.

Research Findings

Composition Task

• Key Findings:

  1. Honesty-Humility and Agreeableness are the strongest predictors, closely linked to writing time, text length, editing behaviors, and other metrics.
  2. Extraversion did not significantly influence composition as expected, which contrasts with findings in academic writing studies and may reflect the creative nature of freewriting.
  3. Personality traits had weaker effects on "effort-related metrics" (e.g., editing behaviors and pauses), and readability was not strongly predicted by any single trait.

• Quantitative Results:

  • Honesty-Humility showed significant positive correlations with multiple metrics, with correlation coefficients exceeding 0.50.
  • Openness to Experience provided moderate support for editing behaviors and pause frequency.

• Comparative Experimental Analysis: Composite models combining multiple traits in linear regression predicted certain composition behaviors more accurately, though individual traits (e.g., Honesty-Humility) often had sufficiently strong effects.

Transcription Task

• Key Findings:

  1. Openness to Experience had the strongest correlation with input speed.
  2. Compared to composition, most personality traits showed limited influence in transcription tasks, likely due to the mechanical nature of transcription reducing the impact of personality.
  3. Error rates and correction behaviors were difficult to predict, with traits like Emotionality showing only moderate correlations with specific metrics.

• Quantitative Results:

  • Openness to Experience had a statistically significant correlation coefficient of 0.51 with input speed (WPM).

• Patterns and Behavioral Insights: Tasks requiring more editing effort and creative planning (e.g., composition) reflect personality traits more strongly than purely mechanical tasks.

Limitations and Future Directions

1. Sample Limitations: Participants were primarily young university students, lacking a broader age range.
2. Insufficient Assessment of Experience: Keyboard usage years as an experience metric may be inadequate, as it does not account for usage frequency or motivation.
3. Prediction Challenges: While correlations exist, whether real-time text input behavior can accurately predict personality requires further validation, especially in real-world scenarios.

Practical Implications

• Personalized Design: The study reveals that personality traits can be used to dynamically adjust user interfaces, particularly for optimizing autocorrection, short text prediction, and generative language models.
• Future Adaptation Directions:
  • Tailoring text input systems (e.g., intensity of autocomplete, frequency of grammar checkers).
  • Extending applications to mobile environments and more fragmented contexts such as chatting/texting scenarios.