Speaking Through Chatbots or Text: How Format Shapes Information Agreement, Reactance , Environmental Awareness, and Trust

Paper Title

Speaking Through Chatbots or Text: How Format Shapes Information Agreement, Reactance, Environmental Awareness, and Trust

Publication Info

• Topic area: Effects of conversational agents (CAs) versus static text on environmental communication.
• Keywords: Conversational agents, static text, environmental awareness, trust, reactance, agreement, valence, sustainability, longitudinal study, eFuels.

Background and Problem

• Problem / challenge: Human-Computer Interaction (HCI) lacks systematic empirical evidence on whether conversational agents (CAs) outperform static text in fostering environmental awareness, reducing reactance, and increasing trust and agreement.
• Significance: Understanding how information delivery format influences environmental awareness and attitudes is critical for designing effective sustainability interventions.
• Motivation and related work: Prior research suggests CAs can enhance persuasion and trust in various domains, but their specific role in contested sustainability topics like eFuels remains underexplored. Emotional framing (valence) and longitudinal effects also require further investigation.

Solution

• Proposed approach: A longitudinal study comparing the effects of CAs and static text on agreement, reactance, environmental awareness, and trust, with variations in information valence (positive, neutral, negative).
• Novelty:
  1. First longitudinal comparison of CA and static text for environmental communication.
  2. Examination of valence effects on trust and reactance over time.
  3. Focus on contested sustainability topics (eFuels) to explore emotional and attitudinal responses.
  4. Integration of trust as a dynamic, format-specific construct.
• Procedure and key techniques:
  • Participants (N=449) were exposed to weekly interventions over four weeks in one of six conditions (CA vs. text × positive, neutral, negative valence).
  • Agreement, reactance, environmental awareness, and trust (CA condition only) were measured longitudinally.
  • eFuels were used as the topic, with content tailored to valence and format.

Results

• Concrete findings:
  • Agreement was higher for CAs (M=4.74) than text (M=3.99), with neutral valence yielding the highest agreement.
  • Reactance was lower for CAs (M=2.00) than text (M=2.53), with negative valence reactance decreasing over time.
  • Environmental awareness increased modestly over time (week 1: M=3.86; week 6: M=3.97) but was unaffected by format or valence.
  • Trust in CAs increased over time for negative and neutral valence but remained stable for positive valence.
• Advantage over baselines: CAs consistently outperformed static text in reducing reactance and increasing agreement, regardless of valence or time.
• Experiments / evaluation:
  • Participants were randomly assigned to six conditions and exposed to weekly interventions on eFuels.
  • Measures included agreement, reactance, environmental awareness, and trust (CA only).
  • Longitudinal design captured changes over four weeks.
• Limitations and future work:
  • Lack of personalization in CA content.
  • Narrow focus on eFuels; broader environmental topics are needed.
  • Four-week duration may be insufficient for significant attitudinal change.
  • Self-reported trust measures lack behavioral validation.
  • Valence distinctions may need stronger pre-testing or machine learning validation.

Summary

This study demonstrates that conversational agents (CAs) outperform static text in increasing agreement and reducing reactance to environmental information, with these effects persisting across valence and time. Trust in CAs also increased over time, particularly for negative valence, but environmental awareness showed only modest, format-independent growth. These findings highlight the potential of CAs to foster engagement with sustainability topics while emphasizing the need for responsible design to mitigate risks of uncritical acceptance. Longer-term studies and broader environmental contexts are recommended to further explore CAs' impact on awareness and behavior.