From Curiosity to Belonging

Concept Video

A product vision walkthrough of Adaptive UI across listener tiers — recognition earned by listening, not by broadcasting.

The Opportunity

Amazon Music sponsored a five-person Carnegie Mellon MHCI team to reimagine the artist–listener relationship. Streaming holds years of history between a listener and an artist — and reflects almost none of it back. A fan with 3,200 plays sees the same artist page as someone hearing the song for the first time.

Survey evidence reframed the brief: fans feel connected to moods, memories, songs, and artists far more than to social features. Fandom is felt privately long before it is expressed. The question became: how might an interface recognize loyal listeners without asking them to perform?

Approach

• Framing: PESTLE and competitive audit, 30+ hours of device testing, a vinyl-shop probe, 22 interviews, and a two-part survey (n=85)
• Choosing: 600+ synthesis nodes across affinity boards, a client workshop, four themes, and 17 assumptions ranked by risk
• Testing belief: Two staged experiments on social tolerance and context; five concepts tested in parallel, then converged
• Validating: High-fidelity prototype in Amazon’s design system, evaluated in moderated sessions against the live Amazon Music app

Key Insights

• Fans reject the stage, not the recognition. Private visibility was accepted by 80% of participants; public posting fell to 20%. Social features became inbound rather than broadcast.
• Trust tracks explanation, not data type. The same data felt intrusive without explanation but helpful when its purpose was clear — every adaptation had to explain itself and remain reversible.
• The wall is audience, not visibility. Being seen is fine. Being asked to broadcast is not.

Solution — Adaptive UI

Same artist, different page for every fan. The catalog stays identical; what changes is how much of that listening history the interface admits to knowing, across three tiers earned by listening alone:

• Tier 1 — New here: The deterministic artist record, unchanged. What any listener sees.
• Tier 2 — Returning: Recognition emerges — ambient signals that the listener is seen, not just a visitor.
• Tier 3 — Longtime fan: Community, shame-free return framing, and eras of their journey with the artist.

Across three core surfaces, the interface progressively recognizes loyalty through listening history, milestones, and contextual personalization while preserving familiar navigation.

Impact

• User: Showed how Amazon Music could recognize long-term loyalty without requiring public participation, ranking, or self-promotion
• Organizational: Turned debate about social features into measurable design criteria and eliminated weak directions before engineering
• Long-term: Delivered reusable principles, tiers, and test protocols for evaluating future personalization concepts — reframing personalization as recognition rather than recommendation

Team & Role

Team: Dave Song, Erica Flora Yu, Rabiat Sadiq, Brook Xiao, Celine Tseng (CMU MHCI · Amazon Music DERBC Team)
Role: Project Manager & Engineer · Sponsors Matthew Spangenberg and Se One Park