UX Research

Identity Disclosure

in Service Chatbots

Trust · Customer Retention · UX Design

SubjectUser Experience Design

Duration6 minutes · 6 slides

DateJune 2026

Why Identity Disclosure Matters

Problem · Research question · Relevance

Thesis Presentation

Central Research Question

What role does disclosing a chatbot's non-human identity play in trust and customer retention?

Service chats: users increasingly interact with chatbots instead of humans
Trust and positive experiences are essential for churn prevention
AI hospitality market projected to reach $3.2B by 2025
Transparency about machine identity becomes a key UX design factor

Theoretical Background

Trust as a driver of customer retention

Thesis Presentation

Funke et al. (2023)

Theoretical foundation · our study builds on

Disclosed AIConcealed AIGood outcomeSolid trustTrust ↓Bad outcomeTrust ↓Lowest trust

Methodological Approach

Online vignette study · n = 37 · randomized order

Thesis Presentation

Prior work: disclosed vs. concealed AI — we split disclosure to test how explicit it should be.

Radically transparentAI stated explicitly in chat

Subtle disclosure“AI powered” label only · not in chat

Human disguise

Design derived from Funke et al. (2023): three disclosure levels operationalize open, subtle, and concealed AI · Measures: trust, fairness, retention, cancellation fairness, empathy, comfort (7-point Likert)

Results

Effects vary by dimension

Thesis Presentation

Online vignette study · n = 37 · randomized within-subjects order · paired t-tests · 7-point Likert scales

Paired t-tests

n = 35 complete cases

Significant: Empathy · A vs C · t(34) = −2.50 · p = .017*

Not significant (p > .05): Trust (A·B p = .834 · A·C p = .117 · B·C p = .208) · Retention (all n.s.) · Comfort (all n.s.) · Empathy A·B p = .285 · Empathy B·C p = .098 · exploratory · no Bonferroni correction

Key findingDescriptively: B leads on trust (4.2), C on empathy (3.7), A on comfort (4.1). Only empathy A vs. C is statistically significant — all other paired comparisons n.s.

Discussion

Descriptive patterns · paired t-test interpretation

Thesis Presentation

A — Radical

Comfort highest (4.1). Empathy significantly lower than C (p = .017*) — clearest t-test effect involving A. Open AI identity feels honest but less warm.

B — Subtle

Descriptive trust leader (4.2), but no significant difference vs. A (p = .834) or C (p = .208). Minimal “AI powered” cue — promising, not statistically proven.

C — Disguise

Significantly highest empathy vs. A (p = .017*). Lowest trust & comfort descriptively — not significant in t-tests. Concealment may boost warmth but not verified on retention.

Inferential takeaway: Of 12 paired comparisons, only empathy A vs. C reached significance (t = −2.50, p = .017*). Descriptive leaders on trust (B) and comfort (A) were not statistically confirmed.

Limitations

Student sample (n = 37; t-tests n = 35) · hypothetical vignettes · 12 pairwise tests without correction · small n limits power · no service-outcome manipulation

Conclusion & Outlook

Thesis Presentation

Recommendations

Match disclosure style to the primary UX goal — descriptive patterns differ by dimension
Do not fully conceal AI — C shows lowest trust and comfort
Use B when trust drives retention; use A for clarity — but only empathy A vs. C was statistically significant
Report inferential stats cautiously: small n, exploratory design

Outlook

Larger samples, Bonferroni or ANOVA follow-ups, real cancellation flows, and Funke-style service-outcome interactions.

Thank you!