User Perceptions of Extraversion in Chatbots after Repeated Use

Title of the Paper

User Perceptions of Extraversion in Chatbots after Repeated Use

Paper Information

• Subject Area: Human-Computer Interaction, specifically personalized design and user experience in text-based chatbots
• Keywords: Chatbot, conversational agent, extraversion, personalization, personality, user experience

Research Background and Problem Statement

• Identified Problems or Challenges:

  • While assigning personality traits to robots or voice assistants can enhance user experience, its application in text-based chatbots remains poorly understood.
  • Current commercial conversational agents often adopt a "one-size-fits-all" design approach, overlooking the potential benefits of personalization.
  • Most studies on personalized chatbots focus only on short-term interactions and fail to explore how user perceptions of personality change during long-term use.
  • There is a lack of systematic methods for conversational agents to convey specific personality traits (e.g., extraversion) solely through linguistic cues.

• Importance of the Research:

  • Chatbots are widely used in domains such as health support, and understanding how to shape personalized conversational agents is crucial for enhancing user engagement and therapeutic outcomes.
  • Personalized chatbots can foster greater trust, enjoyment, and deep engagement behaviors among users.

• Research Motivation and Related Work:

  • Based on psycholinguistic literature on the relationship between language and personality, the authors propose using specific linguistic markers to simulate varying levels of extraversion.
  • Most existing studies focus on personality shaping in robots or voice assistants through non-verbal behaviors (e.g., tone, gaze), lacking in-depth investigation into personality shaping through purely linguistic means.

Proposed Solution

• Method or Solution Proposed:

  • Design chatbots with three different levels of extraversion (extraverted, moderate, introverted) and test user feedback after prolonged use.
  • Adjust the linguistic output styles of the chatbots using psycholinguistic markers associated with extraversion (e.g., word frequency, emotional vocabulary, syntactic complexity) for each version.

• Innovative Aspects of the Solution:

  • Gradual manipulation of linguistic markers in text-based chatbots to achieve continuous rather than binary personality shaping.
  • Introduction of a systematic approach to evaluate changes in user perceptions of personality after long-term use through online experiments.

• Implementation Steps and Key Techniques:

  1. Design Phase:
     • Use psycholinguistic markers to design the linguistic style of the chatbots (e.g., extraverted bots use long sentences and emotional vocabulary; introverted bots use formal language and complex words).
     • Implement chatbot interactions via the Telegram platform, systematically debugging and testing each version.
  2. Experimental Phase:
     • Participant Recruitment and Grouping: Recruit 34 participants and group them based on self-assessed extraversion levels (extraverted, moderate, introverted).
     • Experiment Workflow Design: Participants interact with each chatbot version for 4 days (writing stress diaries for 3 days and generating feedback reports on the 4th day), for a total cycle of 12 days.
     • Data Collection: Gather participants' open-ended evaluations, interaction behaviors (e.g., word count), and chatbot extraversion ratings.
  3. Questionnaire and Data Analysis:
     • Use the Big Five personality questionnaire to assess chatbot extraversion.
     • Record user preferences and evaluations.
     • Develop linear mixed models to analyze data and explore the impact of user personality on chatbot preferences.

Research Findings

• Specific Findings:

  • Effectiveness of Extraversion Shaping:
    • Users clearly perceived the extraversion traits in chatbots designed with extraverted and moderately extraverted characteristics.
    • Despite being designed with introverted traits, linguistic markers alone failed to effectively convey introversion; users perceived introverted chatbots as more "formal" or "rigorous."
  • User Preferences:
    • Most users preferred extraverted chatbots, followed by introverted ones; moderately extraverted chatbots ranked lowest.
    • The open conversational style and sociability of extraverted chatbots were widely regarded as more friendly and human-like.
  • User Behavior:
    • Users tended to input more text to introverted chatbots, possibly due to the "professional" tone encouraging more thoughtful responses.
  • Impact of Agreeableness on Preferences:
    • All chatbots were perceived as agreeable by users, but this trait was more prominently associated with extraverted chatbots.
  • Personalization Adaptation:
    • No significant correlation was found between users' personal extraversion scores and their preferences for chatbots.

• Experimental or Evaluation Results:

  • No significant differences were observed in usability scores among the three chatbots, indicating that user preferences were based on personality rather than functional utility.
  • User experience analysis demonstrated that personalized chatbots designed with motivational traits could significantly enhance user engagement and satisfaction.

• Limitations and Future Directions:

  • The sample size was small and overly concentrated on native English speakers in the UK; future research should involve larger sample sizes and diverse cultural backgrounds.
  • Ineffectiveness in shaping introversion suggests the need to explore methods beyond linguistic markers (e.g., delayed responses, reduced interaction frequency).
  • Recommend developing new personality measurement tools tailored for the Human-Computer Interaction field to complement existing personality questionnaires.