Truace
Subscribe
70/100

Moderate evidence

Cite
More
Export CSVExport JSONEmailLinkedInX

Chatbots that ask questions and admit limits get more buy-in on eco advice, studies find

Study title: Talking green: how chatbot interactivity and transparency foster user receptivity to chatbot recommendations in environmental communication

Evidence type
Controlled or validated comparison
Sample
n = 2,294
Sources
1 sourcePeer-reviewed
Published
2 min read
Truvace record
TRV-2026-2683cite or check it

Study title: Talking green: how chatbot interactivity and transparency foster user receptivity to chatbot recommendations in environmental communication

Researchers Chi-Horng Liao and Hope Wilfred Banda ran three online experiments with Taiwanese participants to test how two chatbot design features shape willingness to accept sustainability advice. Participants imagined asking a language model how to be more environmentally friendly. Responsiveness (the chatbot asked about existing eco-habits first) and disclosure of limits (the chatbot said it offered general guidance and was prone to error) were varied between groups.

Read the study abstract

This study examines how chatbot interactivity and limitation transparency shape users’ receptiveness to chatbot recommendations in environmental communication. Drawing on the theory of interactive media effects, we conceptualize these two chatbot features as distinct design cues that influence message acceptance through different psychological mechanisms. Across three experimental studies with Taiwanese participants, the findings show that chatbot interactivity primarily increases receptiveness by strengthening…

The gain

In a 1,200-person experiment in Taiwan, a chatbot that both invited follow-up questions and admitted it was a language model with limitations produced higher receptiveness to its eco-friendly advice than chatbots lacking one or both features (β=0.626).

In brief

In Study 1 (600 participants), the more interactive chatbot drew higher receptiveness, and this ran through higher perceived integrity. In Study 2 (494 participants), the chatbot that admitted its limits also drew higher receptiveness, but through lower skepticism rather than integrity. In Study 3 (1,200 participants), the chatbot combining both features produced the strongest effect, working through both higher integrity and lower skepticism.

The results suggest that chatbots delivering sustainability messages may be better received when they pair conversational responsiveness with honest statements of what they cannot do. The evidence is limited to short, single-scenario vignettes and self-reported attitudes, so it does not show that people actually change their environmental behavior.

Main points

  1. Study 1 randomly assigned 600 Taiwanese participants to a low- or high-interactivity chatbot; high interactivity raised receptiveness (β=0.236), raised perceived integrity and lowered skepticism.
  2. In Study 1, perceived integrity mediated the effect of interactivity on receptiveness, while skepticism had no significant effect on receptiveness.
  3. Study 2 (494 participants) compared a chatbot giving confident advice with one that disclosed its epistemic limits; disclosure raised receptiveness (β=0.260) and lowered skepticism (β=-0.587).
  4. In Study 2, skepticism mediated the transparency effect, but perceived integrity did not; users rated the chatbot as more honest without becoming more receptive through that route.
  5. Study 3 used a 2x2 design with 1,200 participants; the high-interactivity, high-transparency condition produced the highest receptiveness (β=0.626), higher integrity (β=0.863) and lower skepticism (β=-0.256).
  6. Both indirect paths in Study 3 were significant: through integrity (β=0.064) and through skepticism (β=0.185).
  7. Each study ran a pretest (n=64, 66 and 78) confirming the manipulations were noticed, and each used a single standardized vignette so only the two design features varied.

The full story

All three studies used vignette scenarios in which participants imagined asking a language model how to be more environmentally friendly. In the low-interactivity version the chatbot gave a direct answer. In the high-interactivity version it first asked about eco-friendly habits the user already practiced, then tailored suggestions and invited further questions. Participants were recruited through an online platform and received convenience-store gift cards.

In the transparency conditions, the low version gave highly confident recommendations with no caveats. The high version stated it was a language model offering general guidance and was prone to limitations. Perceived integrity, skepticism and receptiveness were measured on 7-point Likert scales, and effects were analyzed with SPSS PROCESS models.

Study 1 participants were 57.7% female, with 44.2% aged 41-55. Study 2 had 494 participants, 87.0% of them employed. Study 3 had 1,200 participants, 54.1% female and 79.6% employed. The authors note that neither perceived integrity nor skepticism alone consistently raised receptiveness across studies, so both may be necessary but insufficient.

The authors conclude that chatbots for sustainability communication should combine engaging interaction with explicit acknowledgment of system limitations, since each cue works through a different psychological route.

Sources

1 on file
  1. Humanities and Social Sciences Communicationsdoi:10.1057/s41599-026-07965-8
    Peer-reviewed2026-06-23

Talk it over

Four questions to start onWho gets this benefit, and who was left out of the study or story?

Start with a question

Chatbots that ask questions and admit limits get more buy-in on eco advice, studies findhttps://truvace.com/item/talking-green-how-chatbot-interactivity-and-transparency-foster-user-receptivity

  1. Who gets this benefit, and who was left out of the study or story?

    The gain In a 1,200-person experiment in Taiwan, a chatbot that both invited follow-up questions and admitted it was a language model with limitations produced higher receptiveness to its eco-friendly advice than chatbots lacking one or both features (β=0.626).

    Answer in the thread

  2. How do we know? What can this kind of evidence show, and what can’t it?

    The evidence This rests on a controlled comparison with a sample of 2,294, from Humanities and Social Sciences Communications.

    Answer in the thread

  3. What does it leave open? What else would you want to know before acting on it?

    The record’s own caveat The experiments used a single standardized vignette per condition and no extended multi-turn dialogue, and they measured self-reported receptiveness among Taiwanese online participants rather than real behavior change.

    Answer in the thread

  4. Have you seen AI used this way where you study, work or live? Did it go the same way?

    Answer in the thread

The questions are the same on every record. The details under them are taken from this record as written.

What readers say

More from Climate