kamya-alkesh-bhanushali profile photo

Kamya Alkesh Bhanushali

User Experience Design MSc

I create human-centered digital experiences that balance business value with user needs. Through a thoughtful blend of interaction design, accessibility, and usability testing, I aim to deliver solutions that are simple, scalable, and impactful.

This study explores how emotionally adaptive tone in AI chatbots influences user trust and satisfaction during e-commerce service failures, revealing that empathetic language can positively reshape digital service recovery experiences.

Problem Statement

AI-powered customer service chatbots are increasingly used in e-commerce to handle user queries, especially during order failures. However, many still sound robotic or impersonal, negatively impacting user trust and satisfaction. Without emotional intelligence, these bots can escalate frustration during already stressful situations. There is limited research comparing emotionally adaptive chatbot tone to neutral tone in real service-failure scenarios, leaving a gap in how digital empathy is implemented and evaluated.

Research Questions

Question 1:
How does emotionally adaptive tone in AI chatbots influence user trust and satisfaction during negative user experiences in customer service?

Question 2:
Do different demographic groups like age & tech literacy respond to emotionally adaptive chatbot interactions in negative scenarios?

Research Objectives

  • Design and test two chatbot prototypes with contrasting tones: neutral vs emotionally adaptive.
  • Evaluate how adaptive tone impacts user trust, satisfaction, and emotional perception.
  • Analyse variations in user response across age and tech literacy groups.
  • Use validated UX metrics to measure tone effectiveness through quantitative and thematic analysis.
  • Apply emotional design principles to build more inclusive and human-centred chatbot experiences.

Prototype Development

Research Plan

This project followed the Design Thinking process to create and evaluate two chatbot prototypes: one with a neutral tone and one with an emotionally adaptive tone.

  • Empathise: Explored user frustrations with robotic chatbot responses in service failure scenarios.
  • Define: Identified the need for emotional tone to improve trust and satisfaction.
  • Ideate: Developed response scripts using identical content but two distinct tones.
  • Prototype: Built chatbots using Botsonic with a consistent refund scenario.
  • Test: Conducted a within-subjects study using UX surveys and qualitative feedback.
  • Implement (Future): Findings support more empathetic chatbot design in real-world services.

Research Method

To explore how chatbot tone affects trust and satisfaction, I followed a Design Thinking approach grounded in user-centred methods.

A focus group helped identify pain points in existing chatbot interactions. Participants emphasised the importance of trustworthiness, personalised responses, and clear guidance in emotionally charged service failure scenarios. Feedback highlighted the need for bots to offer empathy, instant feedback, and clarity in tone and terminology.

Guided by these insights, I developed two chatbot prototypes (neutral and emotionally adaptive) for a controlled user study. Each participant interacted with both prototypes during a failed delivery scenario.

Data was collected through:

  • Validated UX metrics: System Usability Scale (SUS), Customer Satisfaction (CSAT), and Trust in Automation
  • Open-ended questions for qualitative insights

Quantitative data was analysed using t-tests and Wilcoxon tests, while qualitative data was thematically coded to extract key patterns in tone perception.

The method balanced both emotional design and statistical rigour, ensuring user needs informed every design decision.

User feedback was collected through a focus group and formative testing, guiding the refinement of both chatbot tone and study structure.

Participants shared key expectations:

Users valued quick acknowledgment and resolution-focused responses.

Chatbots should sound genuine and emotionally aware, not robotic.

Responses should be clear, structured, and free from complex terminology.

Design Adjustments Based on Feedback

  • Introduced a realistic task scenario (failed delivery + refund request)
  • Added a strict time frame for resolving the issue to simulate urgency
  • Structured the conversation to mimic real customer service expectations
  • Adjusted tone to sound professional but empathetic in critical moments

This ensured the prototype reflected real-world user frustrations and expectations, making the chatbot interaction more authentic, time-sensitive, and emotionally engaging.

Experiment Design

Study Design

This study followed a within-subjects experimental design, where each participant interacted with two chatbot prototypes: one with a neutral tone and one with an emotionally adaptive tone, across the same service failure scenario.

A total of 40 participants aged 18–35 were recruited voluntarily through online distribution. They were divided into demographic groups based on age and tech literacy. Each participant engaged with both chatbot prototypes, one neutral, one emotionally adaptive, in a simulated failed delivery scenario. After each session, they completed post-interaction surveys measuring satisfaction, trust, and usability from academically validated scales.

Design Analysis

  • Normality Test: Shapiro-Wilk for all key variables
  • Inferential Tests:
    • Satisfaction: Paired-sample t-test (Sat_B vs Sat_A)
    • Trust: Wilcoxon signed-rank test (non-parametric)
    • Usability (SUS): Paired-sample t-test
  • Demographic Comparisons:
    • Age group vs. Satisfaction/Trust → Independent t-tests
    • Tech literacy vs. Satisfaction/Trust → Independent t-tests or Mann–Whitney U
  • Effect Sizes:
    • Cohen’s d for t-tests
    • r for Wilcoxon test
  • Power Analysis: Calculated using G*Power to confirm sample sufficiency

Qualitative Analysis

Open-ended feedback was analysed using thematic analysis with an inductive approach, allowing themes to emerge naturally from participant responses.

Key Themes Identified

  • Emotional Understanding – Users appreciated when the chatbot acknowledged feelings
  • Stress Reduction – Adaptive tone helped reduce frustration
  • Genuineness of Tone – Authentic responses increased trust
  • Tone Clarity – Clear and structured replies were preferred
  • Prototype Preference – Most users favoured the emotionally adaptive chatbot
  • Past Experience Influence – Prior chatbot use shaped tone perception

These insights complemented the quantitative findings, highlighting the importance of emotional tone in chatbot design.

Design Building Plan

This section outlines how both the chatbot prototypes and application flow were designed using Botsonic and structured UX techniques.
Chatbot Design (Botsonic)
Designed two chatbot prototypes for a failed delivery scenario using Botsonic

Both followed the same logic but differed in tone:

  • Emotionally Adaptive Chatbot: Used empathetic, human-like language and positive reinforcement
  • Neutral Chatbot: Used direct, transactional responses with no emotional cues

Tone scripting was based on emotional design principles (Norman, 2004).
Custom conversation flows were created in Google Sheets and uploaded to Botsonic.
Fallback messages and escalation responses were added to simulate real service experience.
Each bot was tested for tone consistency and smooth conversational flow.

Application & Interaction Design

Scenario Framing: Participants completed a time-limited e-commerce task (e.g., failed delivery) to simulate real urgency.

Interface Presentation: Chat UI was shown via Figma or web embed with consistent font, layout, and timing.

User Flow: Consent → Scenario → Chatbot A → Survey → Short Break → Chatbot B → Survey → Final Feedback.

Testing Tools: SPSS (data analysis), Google Sheets (bot flow), JISC Online Surveys (questionnaires), and word cloud tools (qualitative themes).

Testing

A within-subjects experiment was conducted to evaluate two chatbot tones, neutral and emotionally adaptive, using a realistic e-commerce task: resolving a failed delivery.

Participants followed a structured user flow:

  • Received a push notification simulating a delivery failure
  • Accessed the Shopiq order page and navigated to the chatbot
  • Engaged in a time-limited interaction with the chatbot to resolve the issue
  • Completed post-interaction surveys after each prototype

Evaluation, Results & Discussion

The evaluation assessed how chatbot tone influenced user satisfaction, trust, and usability during a simulated e-commerce service failure.

Quantitative Results:

  • Satisfaction was significantly higher for the emotionally adaptive chatbot (p < .001, d = 0.61)
  • Trust also increased with the adaptive tone (Wilcoxon p = .005)
  • Usability (SUS) was slightly higher for the neutral chatbot, suggesting adaptive tone may increase cognitive load

Qualitative Insights:

Open-text responses revealed key themes:

  • Emotional Understanding and Tone of Genuineness were highly valued
  • Users felt more reassured and heard with emotionally adaptive responses
  • Prior chatbot experiences influenced how users interpreted tone
  • Most participants preferred Prototype B (emotionally adaptive) for its human-like interaction

Interpretation:

While both prototypes effectively completed the task, the emotionally adaptive chatbot enhanced the user experience by building emotional trust and reducing stress. However, the slight dip in usability suggests that designers must balance empathy with clarity and simplicity.

Key Findings

  • Adaptive tone improved trust and satisfaction significantly compared to neutral tone.
  • Users valued empathy during service failures, feeling more understood with the adaptive chatbot.
  • The neutral chatbot had slightly better usability, indicating a trade-off between emotion and simplicity.
  • Older and tech-savvy users preferred the adaptive tone more strongly.
  • Authenticity and clarity were key — overly scripted empathy reduced trust.
  • Past chatbot experiences influenced how users perceived tone and usefulness.
Modern promotion strategy. Keywords research, online advertising business, seo optimization. Internet promotion, digital marketing, information search.

Design Recommendations

  • Use adaptive tone in emotional moments: Employ empathy when users are frustrated, but avoid overuse to maintain credibility.
  • Balance tone with clarity: Keep messages concise and structured even when emotionally adaptive language is used.
  • Prioritise tone authenticity: Ensure the chatbot doesn’t sound overly scripted — natural, human-like responses increase trust.
  • Tailor tone to user profiles: Consider age and tech literacy in tone design — not all users respond to empathy in the same way.
  • Provide resolution-focused responses: Empathetic language is most effective when paired with clear solutions and actionable next steps.

Implications for Design Practice

  • Tone is a UX element, not just personality, it directly affects trust, satisfaction, and emotional connection.
  • Emotionally adaptive chatbots can improve digital service recovery experiences, especially in high-stress scenarios.
  • Over-personalisation may reduce usability, highlighting the need for thoughtful tone integration.
  • Findings support the value of emotion-aware AI design, especially in customer support, e-commerce, and conversational interfaces.

Conclusion

This project explored how emotionally adaptive chatbot tone can shape user experience during service failure scenarios. Through designing and testing two chatbot prototypes, the study revealed that empathetic tone significantly improves trust and satisfaction, especially for users seeking support in stressful situations.

However, findings also highlight the need to balance empathy with clarity. While emotionally rich responses strengthen connection, overly scripted tone or complex phrasing can affect usability. The research demonstrates that tone is more than a stylistic choice; it’s a functional UX tool that impacts how users feel, respond, and resolve issues.

As a designer, this project taught me that emotional design is not just about aesthetics, it’s about creating conversations that feel human, supportive, and contextually aware.

Next - Bonny Chen →