Meaningful Music Rabbit Holes
Timeline
June - October 2024
Role
Interaction Designer
Team
Solo
Overview
This case study summarises my MSc Human-Computer Interaction dissertation. It explores the concept of 'meaningful music experiences' within Spotify's personalisation framework through user research, prototyping, and evaluation. Key findings informed the design of a prototype that takes a new approach to music recommendations. User testing showed the prototype to be promising, with participants comparing it favourably to Spotify's existing features and expressing interest in future use. (Note: Spotify was not an official partner.)
Challenge
Active Music Streamers find algorithmic recommendations inconsistent and confusing. They want more clarity and control over their personalised experience, however obscure and underperforming algorithms instead evoke feelings of distrust.
How might we best engender meaningful music experiences through personalisation for music streaming users within the Spotify platform?
Solution
A music recommendation feature focused solely on users discovering new music. Users select a song as a starting point, and recommendations are displayed in a node-based UI. The selected song is the central node, with recommendations branching outwards. Users can refine results by choosing new nodes or applying filters, ensuring a dynamic and customisable experience.
Tap images to expand
The project was conducted in three stages aligned with the research objectives (RO) outlined below:
RO1 & RO2.
Initial observations and interviews
RO3.
Design of a novel prototype
RO4.
Formative evaluation of prototype
User Research - Interviews & Observations (RO1 & RO2)
Overview of 'RO1 & RO2 initial observations and interviews' process
13 participants were observed and interviewed.
The initial phase aimed to understand how and why users interact with Spotify’s recommendation and personalisation feeds, identifying factors that contribute to meaningful music experiences. Insights were gathered through semi-structured interviews and observations, incorporating think-alouds. Sessions were conducted online via Zoom to create a naturalistic environment.
Sessions were conducted online via zoom as shown above. This helped promote a naturalistic environment in which to observe participant behaviour.
Conceptual Design - Data Analysis and Wireframes
Overview of 'RO3 - Design of a novel prototype' process.
Data was analysed using Braun & Clarke’s thematic analysis. Transcripts were coded using Dovetail, with relevant insights grouped into six functional requirements:
Dovetail codes related to participants Meaningful Music Experiences can be seen on the left. On the right, an example of insights grouped within a spreadsheet can be seen.
Key Findings:
Active users enjoy discovering and curating music themselves, feeling accomplishment and enjoyment in the process.
Partial automation is favoured, where users can work in symbiosis with algorithms
Passive listening is common among all users, however less frequently leads to meaningful experiences.
Users desire more clarity about how features function and why recommendations are made.
Users want more control over the scope of recommendations
Anecdotal meaningful music experiences often occur outside the platform. A design should aim to engender meaningful music experiences within the platform.
“How might we best engender meaningful music experiences through personalisation for music streaming users within the platform”.
Nine initial design ideas were generated using ideation techniques such as "How Might We" and brainstorming. Viewing each of the nine design ideas side by side allowed the researcher to draw links and uncover which ideas were complimentary. A single proposal was generated:
Final design proposal
Summary: A node-based recommendation feature for discovering new music. Users prescribe a song as the starting point, with recommendations displayed in a circular, branching UI. Selecting a node generates new recommendations. Filters allow users to customise their experience, and design elements such as node size and proximity provide clarity on recommendation logic.
Meaningful music experience codes: The feeling of finding a new song, Pride related to creating playlists.
Embedded Functional Requirements: F1. Active Users don’t want everything done for them, F2. Active user working in symbiosis with algorithm, F4. Users desire more clarity, F5. Users desire more control, F6. Aim to engender meaningful music experiences within the platform.
Inspiration for both the conceptual design and visual UI was drawn from commercial and academic sources including: Dork et al.'s 'PivotPaths', Daniel Kuntz 'Better music exploration UI', tunebat.com & music-map.com, as well as many others.
Daniel Kuntz 'Better music exploration UI' can be seen on the left. Tunebat.com & music-map.com can be seen on the right.
Initial Sketched Wireframes
Low-fidelity Digital Wireframes
Detailed Design - Prototype
The prototype capitalised on research findings that users derive enjoyment from music discovery. The node-based design provides an engaging and interactive way to explore new music. Users prescribe a song as the central node, with eight surrounding nodes representing algorithmically generated recommendations. Selecting a node regenerates the recommendations, offering seemingly endless exploration paths. Additional controls such as filters for region, popularity, and similarity allow for greater personalisation.
Prototype Evaluation
Overview of 'RO4 - Formative evaluation of the prototype' process.
A formative evaluation was conducted to observe user interaction with the prototype and gather feedback on its usefulness and potential to support meaningful music experiences. Six participants from the UK, Australia, and NZ, all active Spotify users, were recruited for the study. The evaluation included a single interaction task followed by ten feedback questions.
A sample clip from a Prototype Evaluation.
Results
Understanding
All six (6/6) participants understood the core concept, as well as the logic behind node size and proximity. However, advanced settings like the Similarity Slider caused confusion.
Usefulness
All (6/6) participants agreed the feature supported music discovery.
Future Use
Five (5/6) participants expressed a strong likelihood of future use.
Clarity and Control
All (6/6) participants felt the feature provided more control than Spotify’s existing options.
Meaningful Experiences
Five (5/6) participants believed the feature could foster meaningful music experiences.
Areas for Improvement
Common usability issues included confusion with the hover interaction, difficulty viewing song details without playback, and the central node doubling as a "go back" button. Participants suggested adding play/pause, “add to library”, and the ability to save exploration paths.
