Yu, S., Xiong, J., & Shen, H. (2024). The rise of chatbots: The effect of using chatbot agents on consumers' responses to request rejection. Journal of Consumer Psychology, 34(1), 35-48. by Bonnie Chan

Topic & Focus

bonnieccchen — 2025-09-16 17:27:21 UTC

Main Theme: The rise of chatbots in service interactions.
Focus: How do consumers respond differently to chatbots vs. human agents, especially when their service requests are rejected?

Research Question

bonnieccchen — 2025-09-16 17:42:33 UTC

Do consumers react more negatively when their requests are rejected by chatbots compared to human agents, or vice versa?
Under what conditions (e.g., service rejection vs. acceptance, emotional vs. non-emotional apologies) do chatbots or human agents perform better?
What role does perceived flexibility play in shaping consumer evaluations?

Theoretical Background

bonnieccchen — 2025-09-16 18:01:27 UTC

Algorithm Aversion: Prior research shows consumers often distrust algorithms and prefer humans, especially when tasks require flexibility, empathy, or subjective judgment (Dietvorst et al., 2015; Longoni et al., 2019).
Attribution Theory: Consumers explain service outcomes by attributing them to causes (e.g., effort, flexibility, rules). Expectations about the type of agent (human vs. chatbot) shape these attributions.
Perceptions of Robots vs. Humans:
- Humans = emotional, flexible, adaptive.
- Robots = rule-based, rigid, emotionless.
Key Tension: Rigidity can be a disadvantage in successful service, but may buffer negative reactions in failed service.

Hypotheses

bonnieccchen — 2025-09-16 18:03:46 UTC

H1a: When consumers receive a rejection of their service request, they evaluate the service less negatively if the service is handled by a chatbot agent (vs. a human agent).
H1b: The effect of chatbots (vs. human agents) on service evaluation is driven by consumers' perception that robot agents are less flexible.
H2a: Consumers predict the service from a human to be more flexible and therefore better than that from a robot if they have not experienced it.
H2b: Consumers react to the service from a human more favorably than that from a robot if their request is accepted.
H3: When an apology message for a failed service delivery does not involve emotions, consumers might respond to the service more favorably if it is handled by a chatbot agent versus a human agent. However, this effect is reversed if the service agent conveys emotions to apologize for the failed service delivery.

Experimental Design Summary（in Chinese）

bonnieccchen — 2025-09-16 18:09:46 UTC

Experimental Design Summary

bonnieccchen — 2025-09-16 18:17:27 UTC

Results (Key Findings from Six Studies)

bonnieccchen — 2025-09-16 18:20:59 UTC

Study 1–3 (Rejection scenarios):
- Consumers rated chatbot agents more positively than human agents when requests were rejected.
- Mediated by lower expectations of flexibility for chatbots.
- Effects replicated in both online and real-life field settings.
Study 4A (Before service):
- Consumers predicted human agents > chatbots (algorithm aversion).
Study 4B (Outcome moderates):
- Failure → Chatbots > Humans.
- Success → Humans > Chatbots.
Study 5 (Apology moderates):
- Non-emotional apology → Chatbots > Humans.
- Emotional apology → Humans > Chatbots (perceived as more sincere).

General Discussion

bonnieccchen — 2025-09-16 18:21:52 UTC

Consumers evaluate chatbots more positively in rejection cases because they expect less flexibility.
When services succeed, or when emotional apologies are used, humans outperform chatbots.
Results show that the same trait (rigidity) can be a liability (in success) but an advantage (in failure).
Provides evidence that agent type shapes attribution:
- Rejection by human → attributed to “lack of effort.”
- Rejection by chatbot → attributed to “rule constraints.”

Implications 1

bonnieccchen — 2025-09-16 18:24:28 UTC

Theoretical Implications

Extends literature on algorithm aversion by showing contexts where consumers actually prefer robots.
Highlights the dual role of rigidity: disadvantage in success, advantage in failure.
Broadens attribution theory: agent type shapes expectations and attributions.
Adds nuance to apology research: emotional apologies work better for humans, not for chatbots.

Please leave your questions here

bonnieccchen — 2025-09-16 18:29:33 UTC

请在此模块留言您的问题

bonnieccchen — 2025-09-16 18:41:02 UTC

Application & Broader Reflection

If you were a manager, in what situations would you prefer to use chatbots, and in what situations would you prefer humans?
Can the methods used here be applied to other domains (education, healthcare, tourism)? How?
How does this paper inspire your own research ideas about AI, service, or consumer psychology?

Method-Focused Discussion （1）

bonnieccchen — 2025-09-16 18:45:20 UTC

The authors used both scenario-based experiments (e.g., imagining canceling an iPhone order) and field experiments (e.g., gift redemption, hand cream shortage) to collect data.

Do you think scenario-based experiments can truly capture real consumer behavior?
Do field experiments provide stronger evidence?
What are the strengths and weaknesses of each approach in this study?

作者在研究中既使用了情境模拟实验（例如想象取消 iPhone 订单），也使用了田野实验（例如礼品兑换、护手霜缺货）来收集数据。

你认为情境模拟实验能真实反映消费者的行为吗？
田野实验是否能提供更有力的证据？
在这项研究中，这两种方法各自的优点和缺点是什么？

Method-Focused Discussion（2）

bonnieccchen — 2025-09-16 18:48:11 UTC

Measurement: The authors used both self-reported measures (Likert scales) and behavioral outcomes (e.g., actual gift redemption).

What are the strengths and weaknesses of these two approaches in terms of validity and realism?
Do attitudes from surveys always match real consumer behavior? Why or why not?
How could combining them — or adding new techniques (e.g., eye-tracking, digital trace data) — improve the study?

作者同时使用了 自陈量表（李克特量表）和 行为数据（如实际礼品兑换）。

在 效度与真实性 方面，这两种方法各自的优缺点是什么？
自陈态度是否总能与真实消费者行为一致？为什么？
结合两种数据，或加入新的方法（如眼动追踪、数字行为数据），能如何改进研究？

Method-Focused Discussion （3）

bonnieccchen — 2025-09-16 18:51:15 UTC

Cross-cultural samples: The authors used Western online panels (Prolific, MTurk) and Chinese university students.

How might cultural differences influence the results?
Does this diversity of samples strengthen the study by improving generalizability, or weaken it due to potential comparability issues?

跨文化样本： 作者既使用了西方的线上样本（Prolific、MTurk），也使用了中国大学生样本。

文化差异可能会如何影响研究结果？
这种样本多样性是增强了研究的普遍性，还是因可比性问题而削弱了研究？

Critical Thinking 1

bonnieccchen — 2025-09-16 18:53:42 UTC

Do you agree with the authors’ conclusion that “chatbots may soften the negative impact of rejection”? Why or why not?
你是否同意作者的结论：“聊天机器人可能会减轻拒绝带来的负面影响”？为什么同意或不同意？

Critical Thinking 2

bonnieccchen — 2025-09-16 18:54:33 UTC

What are the biggest strengths and weaknesses of the research design?
这项研究设计的最大优点和缺点是什么？

Critical Thinking 3

bonnieccchen — 2025-09-16 18:56:00 UTC

Based on this article, what future research directions do you see?What unanswered questions remain?

基于这篇文章，你认为未来的研究可以有哪些方向？哪些问题仍未得到解答？

Implications 2

bonnieccchen — 2025-09-16 19:00:24 UTC

Managerial Implications

Firms can use chatbots to deliver rejections—they soften consumer dissatisfaction.
Be cautious with personalization and flexibility promises for human agents → higher expectations can backfire if requests are rejected.
Do not program chatbots to deliver emotional apologies—consumers see them as insincere.
Use human agents in success scenarios or when authentic emotion is required.

bonnieccchen — 2025-09-18 17:58:20 UTC

In this article, the authors discuss both boundary conditions and attribution theory.

What is the difference between these two concepts?
How do they work together in explaining consumer reactions to chatbots vs. human agents?

在这篇文章中，作者既讨论了边界条件，也提到了归因理论。

你觉得这两个概念有什么区别？
它们是如何结合起来解释消费者对聊天机器人和人工客服反应的？