WEEK 7: Project Proposal
THE PROBLEM:
Streaming platforms are really good at recommending music that it thinks you’ll like based on what you save and listen to, but it can be difficult to find brand new stuff if you don’t know what to look for, or how to find it.
THE PROJECT:
Music recommender not based on listening history, as a way to introduce people to new genres they might not have found otherwise.
REFERENCES:
1. This project I did in undergrad
![]()
![]()
![]()
2. Gnoosic: Recommends an artist based on initial input from the user of 3 artists that they like. Gives the option to listen to a song from that artist immidiately via Amazon Music
![]()
![]()
3. A since-discontinued Spotify playlist that updated every 2 hours for each user based on time of day and what that user typically listened to during that time of day/day of the week. They had silly names based on the types of songs in the playlist. I like this idea and it was cool to see how the time of day affected my listening habits. After a while though, the songs seemed to be the same every morning and every afternoon and evening, so it lost its charm.
![]()
![]()
![]()
Bonus: My undergrad thesis, Click, Skip, Share: How Streaming is Changing Our Relationship with Music. I have always been interested in how we experience music, how we catalog it, and how we share it, and how those processes change our feelings and relationships with the music itself. I interviewed many streaming users and a lot of people said that they feel overwhelmed by options when looking for something to listen to, and the platform-curated playlists sometimes feel repetitive. It’s hard to jump completely into a new genre when so much of the platform’s design is centered around personalization.
4. Every Noise At Once is an incredible map of genres, with the option to play a clip of a song from every genre.
![]()
NOTES:
I want it to give a different answer every time the user interacts with it - I think I can achieve this by including a variable like local weather data or what time it is to influence the answer given.
There doesn’t need to be a real link between the metrics and final recommendation.
Questions I have:
Is genre the best “result” to give? Or should it be an artist / album / etc ? The categorization and how platforms tag music with different genres is another thing I am interested in. If I go with genre, should I also include
API recommendations?
What other metrics should I use to determine what someone should listen to?
Any other ideas for how to achieve this goal with a different application / project?
3. A since-discontinued Spotify playlist that updated every 2 hours for each user based on time of day and what that user typically listened to during that time of day/day of the week. They had silly names based on the types of songs in the playlist. I like this idea and it was cool to see how the time of day affected my listening habits. After a while though, the songs seemed to be the same every morning and every afternoon and evening, so it lost its charm.
Bonus: My undergrad thesis, Click, Skip, Share: How Streaming is Changing Our Relationship with Music. I have always been interested in how we experience music, how we catalog it, and how we share it, and how those processes change our feelings and relationships with the music itself. I interviewed many streaming users and a lot of people said that they feel overwhelmed by options when looking for something to listen to, and the platform-curated playlists sometimes feel repetitive. It’s hard to jump completely into a new genre when so much of the platform’s design is centered around personalization.
Streaming platforms are really good at recommending music that it thinks you’ll like based on what you save and listen to, but it can be difficult to find brand new stuff if you don’t know what to look for, or how to find it.
THE PROJECT:
Music recommender not based on listening history, as a way to introduce people to new genres they might not have found otherwise.
REFERENCES:
1. This project I did in undergrad
2. Gnoosic: Recommends an artist based on initial input from the user of 3 artists that they like. Gives the option to listen to a song from that artist immidiately via Amazon Music
3. A since-discontinued Spotify playlist that updated every 2 hours for each user based on time of day and what that user typically listened to during that time of day/day of the week. They had silly names based on the types of songs in the playlist. I like this idea and it was cool to see how the time of day affected my listening habits. After a while though, the songs seemed to be the same every morning and every afternoon and evening, so it lost its charm.
Bonus: My undergrad thesis, Click, Skip, Share: How Streaming is Changing Our Relationship with Music. I have always been interested in how we experience music, how we catalog it, and how we share it, and how those processes change our feelings and relationships with the music itself. I interviewed many streaming users and a lot of people said that they feel overwhelmed by options when looking for something to listen to, and the platform-curated playlists sometimes feel repetitive. It’s hard to jump completely into a new genre when so much of the platform’s design is centered around personalization.
4. Every Noise At Once is an incredible map of genres, with the option to play a clip of a song from every genre.
NOTES:
I want it to give a different answer every time the user interacts with it - I think I can achieve this by including a variable like local weather data or what time it is to influence the answer given.
There doesn’t need to be a real link between the metrics and final recommendation.
Questions I have:
Is genre the best “result” to give? Or should it be an artist / album / etc ? The categorization and how platforms tag music with different genres is another thing I am interested in. If I go with genre, should I also include
API recommendations?
What other metrics should I use to determine what someone should listen to?
Any other ideas for how to achieve this goal with a different application / project?
- Ideas for a project title?
What Should I Listen to Right Now?
- Project description (max 100 words)
Instead of relying on an algorithm tailored to your music interests based on listening history and current activity, What Should I Listen to Right Now takes into account objective factors such as local weather, time of day, and a few lifestyle questions to recommend a genre of music to explore. The goal of this project is to expose listeners to a genre they might not have been familiar with or given much thought to, and to relieve the "paradox of choice" that streaming users are faced with.
- Who is the user? (the intended audience can also be you)
Anyone who has ever felt like they don’t know what they want to listen to, and are overwhelmed by options on streaming services.
- What is the user experience?
Anyone with an internet connection will be able to go to a website where the project is hosted and enter some information to determine location (zip code?), and they will be presented with a genre of music to explore. I’d like for a recommendation of artist / playlist / song to populate if I can make it work. I want to add a few other places of input from the user that have nothing to do with music taste as well.
- What technologies are required? Why?
- What do you need to learn/research/decide?
API integration - how complicated is it? What APIs will I need to use?
- What assets/content/code already exists? (your own sketches or other people's sketches)
- What assets/content/code needs to be created?
- What resources/books/sites/examples will you rely on?
- What is the end goal?
To introduce someone to a new genre of music that they likely wouldn’t have found through personalized recommendations on Spotify/Apple Music/etc
- How will you determine if this project is a success?
- Cite 3 sources of inspiration (How did you become interested in this idea? Quotes, photographs, products, projects, people, music, political events, social ills, etc.)
- This project I did in undergrad
2. Gnoosic: Recommends an artist based on initial input from the user of 3 artists that they like. Gives the option to listen to a song from that artist immidiately via Amazon Music
3. A since-discontinued Spotify playlist that updated every 2 hours for each user based on time of day and what that user typically listened to during that time of day/day of the week. They had silly names based on the types of songs in the playlist. I like this idea and it was cool to see how the time of day affected my listening habits. After a while though, the songs seemed to be the same every morning and every afternoon and evening, so it lost its charm.
Bonus: My undergrad thesis, Click, Skip, Share: How Streaming is Changing Our Relationship with Music. I have always been interested in how we experience music, how we catalog it, and how we share it, and how those processes change our feelings and relationships with the music itself. I interviewed many streaming users and a lot of people said that they feel overwhelmed by options when looking for something to listen to, and the platform-curated playlists sometimes feel repetitive. It’s hard to jump completely into a new genre when so much of the platform’s design is centered around personalization.
- Do you think all of this feasible in 7 weeks? Why?
- Questions for your classmates. What are you unsure of, conceptually and technically? On what aspect(s) would you like feedback?