Polite But Boring? Trade-offs Between Engagement and Psychological Reactance to Chatbot Feedback Styles

Paper Title

Polite But Boring? Trade-offs Between Engagement and Psychological Reactance to Chatbot Feedback Styles

Publication Info

• Topic area: Chatbot feedback styles in behavior-change interventions.
• Keywords: Chatbots, feedback styles, psychological reactance, politeness, verbal leakage, engagement, behavior change, conversational agents, user perceptions, HCI.

Background and Problem

• Problem / challenge: Designing chatbot feedback styles that minimize psychological reactance while maintaining user engagement is challenging. Prior work often focuses on politeness strategies but neglects trade-offs with engagement and novelty.
• Significance: Understanding these trade-offs is critical for designing effective conversational agents in domains like health, sustainability, and civic participation.
• Motivation and related work: Previous studies show that direct styles can provoke reactance, while politeness reduces it but risks disengagement. Novel styles like verbal leakage remain underexplored, offering potential to balance persuasion and engagement.

Solution

• Proposed approach: Comparison of three chatbot feedback styles—Direct, Politeness, and Verbal Leakage—across personally-affecting and societally-affecting scenarios.
• Novelty:
  1. Empirical comparison of Direct, Politeness, and Verbal Leakage styles in chatbot feedback.
  2. Exploration of trade-offs between minimizing reactance and sustaining engagement.
  3. Analysis of psychological distance (personal vs. societal scenarios) on user perceptions.
• Procedure and key techniques:
  • Conducted a 3 × 2 mixed factorial study (N = 158 participants) using hypothetical scenarios.
  • Feedback styles manipulated via GPT-4o-generated chatbot responses.
  • Measured emotional reactance (anger, guilt, surprise), perceived threats to freedom, and message effectiveness (processing, persuasiveness).
  • Qualitative analysis of open-ended user feedback.

Results

• Concrete findings:
  • Politeness evoked the least anger (M = 1.73) and threats to freedom (M = 2.00) but was rated as less surprising (M = 1.87).
  • Verbal Leakage elicited higher surprise (M = 2.52) and engagement but also higher guilt (M = 2.53) and threats to freedom (M = 2.83).
  • Direct style triggered the most anger (M = 2.45) and was least persuasive (M = 2.62).
• Advantage over baselines:
  • Politeness was significantly more persuasive than Direct (p <.0001) and Verbal Leakage (p = 0.0054).
  • Verbal Leakage was more engaging but risked reactance compared to Politeness.
• Experiments / evaluation:
  • Participants interacted with chatbots across two psychological distance scenarios (personal and societal).
  • Quantitative measures included Likert scales for emotional reactance and message effectiveness.
  • Qualitative coding assessed behavioral intention and sentiment.
• Limitations and future work:
  • Scenario-based design limits external validity; longitudinal studies needed.
  • Findings may not generalize across cultures or modalities (e.g., voice interfaces).
  • Future work should explore explanation types and user preferences for chatbot styles.

Summary

This study compared three chatbot feedback styles—Direct, Politeness, and Verbal Leakage—in behavior-change scenarios. Politeness minimized psychological reactance but was perceived as unengaging, while Verbal Leakage increased engagement and surprise but evoked higher reactance. Direct style was least effective, provoking anger and threats to freedom. Results highlight trade-offs between minimizing reactance and fostering engagement, suggesting that designers should adopt diverse feedback styles tailored to user preferences and context. Future work should investigate longitudinal effects, cultural differences, and personalization strategies.