By Allegra Papera and Gianna Sanchez
AstroSalon is an astrology based application that takes astrological readings from Co-Star (an astrology app) and provides media recommendations (movies, readings, music) that complement the current state of the stars. The target users for this application are people in Generation Z. Gen Z-ers as a whole are interested in astrology, and they grew up consuming media regularly as they were born into the age of the internet. The AI features used in this application are recommendations. The user is first asked to answer questions about what type of media they prefer, which helps the application algorithm decide what type of media to suggest. The algorithm also uses keywords from Co-Star that trigger certain emotion/mood types with media types that complement them.
The main work we compared this to was Co-Star. We used Co-Star because it is the most popular astrology application amongst Gen-Z. They use it to see what their day, week, and birth chart looks like. The application says what will be positive and negative each week, but doesn’t provide solutions on how someone could improve their mood or what could match their mood. That’s where AstroSalon comes in. Eventually, if anybody were to fully develop this application, the hope is that the astrological details (birth charts, planetary positions, etc) found in Co-Star would be included in and expanded upon within AstroSalon, rather than connecting to Co-Star, so AstroSalon would just be the go-to app for all of this information.
The application opens with a login/sign up page. The user is able to connect to their Co-Star so the astrology information is imported into the app. Then the user can identify which media types they are interested in receiving recommendations from. From this point, AstroSalon identifies what moods Co-Star is projecting for the week and recommends media based on the corresponding mood. If a user picks music, the application suggests albums that complement the various moods. If movies or books are chosen, a list will be recommended. The list was made by us, and we collaborated on each selection. Because it’s recommendation-based, the user will be able to rate the suggestions so that the algorithm can adjust and later recommend better media based on their preferences. As a starting point for the application, we created four suggestions per mood with eight moods in total per media type. Below are the current types of media listed with their associated recommendations based on mood. As the app continues, a team of people will continue to make recommendations for the associated moods.
To test out these recommendations, users will be able to rate the media within the app. For now, we tested this data on random users. We created a survey and sent out to 18 people. We had each person rank from 1–5 whether they thought the recommendation was good for the mood. This data would be collected to help make the AI algorithm. This was a collection of our results, focusing on the mean ranking of each recommendation. It was difficult to do without using the application itself. The point of the recommendations are to show people something they may not know. This is hard to test on users who may not be willing to try the various types of media we recommended in our survey. In the future, this should be tested on a larger scale in order to get accurate information and further develop the application. If the testing were to be done through the app instead of a survey, we would have been given a better data set.
Most of our recommendations had less than ideal scores for the its mood pairing. We believe that personal biases as well unfamiliarity with certain media contributed to these low rankings. We saw particularly low rankings for our book recommendations, but we also only saw one vote for books as preferred media type. This leads us to believe that the low rankings in books had to do with unfamiliarity with the suggestions.
Besides the books, most scores were lower than ideal, and we would hope for our users to be more satisfied with our recommendations for their moods than what we saw here in this survey. In order to get better results, we might benefit from providing subjects with film trailers, book synopses, and audio samples to better enhance the survey-taking experience to allow them to give more informed answers.
