1. Determine what kind of social media site you are creating. Then answer the following questions, explaining your decisions. (Remember the key here is to create an ethical algorithm model).
I am creating a new music streaming platform called UrMixtape, similar to Spotify or Apple Music. The difference is that it can be interactive, meaning that there will be communities based on genres where people will be able to discuss new music releases from favorite artists or just share their song of the day. People will be able to join as many genre communities as possible that they like, and this will be available internationally. There will also be a community for small artists, to give them the chance to be discovered.
2. What will you measure? (ex: likes, hashtags, how long someone stays on a page, etc) Be sure to explain all your decisions.
UrMixtape is a music platform therefore likes, shares, and listens will be very prevalent. How many times has a song been shared or listened to? Where in the world is the song being listened to the most? These factors will be measured for the purpose of diversifying music genres as well as pushing music out where it may not be listened to as much. Another measure will be the option to privately dislike the song, to cater specifically to what kind of music you want to have suggested to you. The like-to-dislike ratio will be hidden. These measures all taken into consideration to make users experience exceptional. UrMixtape is supposed to feel catered to the individual, while building communities with common interests.
3. How will you weight these factors? (rank factors from most important to least important)
First, the algorithm will consider what songs the individual likes and does not like, in order to create a more pleasant experience for them. Second, the amount of times a song is listened to will be most important because not every user will be as active as another. Some users like to be ghosts and do not like or share anything. Third, the amount of likes a song has, followed by the number of shares.
4. What factors will not be measured? (what will not be measured in your algorithm and why)
Things that will not be measured are things like race/ethnicity, gender, or sexual orientation of each individual user. The user will be able to have a profile picture of themselves or not. The reason UrMixtape will not take these factors into consideration is because they are irrelevant to music. Anyone can enjoy music no matter who they are. K-pop is not just for Korean listens. Country is not just for white Americans. Spanish rock is not just for Hispanics or Latinx people. Music should unite people and help them connect over common interest of genre. Adding these factors will only create more room for division among users.
5. How will the model learn or adapt to new information? (this can include changes in language, etc)
UrMixtape will work with Google Translate. Each song will have transcripts for whatever language needed. For the forum section where people will be able to add comments on songs, there will be a “translate” option button at the end of the comment but will be kept in its original form until pressed. After being pressed, the comment can be translated to selected language.
6. What biases will be built into the model? How will this change the algorithm’s outcome?
Other than language barriers, the algorithm will have certain words that cannot be used. These words include any form of discriminatory language. When a harmful comment is detected, it will go through revision. Users can also flag comments that seem insensitive. If a user is taking advantage of the “flagging” option is over-flagging, their profile will go under revision as well. Each time a comment is found insensitive or hurtful, a private message will be sent to the user. The messages will explain why they are flagged and give context to the word(s) used and why they are harmful. There will be a 3-strike system. After the third strike, the account will be terminated.
7. How will the model live out the (un)ethical practices we have discussed?
This model will allow users to be able to find common interests through music. It will distract from the idea that they might not come from the same place, or they identify as different gender. Something I would like to add is that deaf people can enjoy music as well! The transcript will also allow deaf users to enjoy the lyrics as well as adapt bass controls to be able to feel vibrations. The idea of this streaming platform and forum is to expel racism, sexism, ableist, and any other discriminatory practice. My hope is that people can unite through music.