When a music app recommends a song, it can feel as if the platform knows your taste. In reality, recommendation systems use data and patterns to make predictions about what you may want next. You will model simple recommendation ideas, investigate what real platforms disclose and evaluate how personalisation can influence discovery. By the […]
Recommendation systems do not need to ‘know you’ like a friend does. They can learn patterns from what people play, skip, save, search and repeat.
A recommender can look for people with similar behaviour or for songs that share useful characteristics. Real systems can combine many signals.
Recommendation systems are complex and change over time. Your job is to find what is documented rather than invent a secret formula.
Personalisation can help people discover music they love. It can also repeatedly reinforce familiar tastes, amplify popular content or make platform choices feel neutral when they are not.
Your finished piece should show data → patterns → recommendation → feedback and include one benefit, one limitation and one uncertainty.
Recommendations are useful because they reduce choice overload — but that means platform design can shape what becomes visible.
Which part of recommendation systems felt less ‘magical’ once you modelled it?
How could you deliberately widen what a recommendation system learns about your interests?
How might generative AI change music discovery and recommendation over the next few years?
Choose Your Tools
Use the tools available to you. Strong thinking and evidence matter more than a particular app.
Use Spotify or another music platform only if permitted. You can complete the challenge without a personal account by using screenshots, public explanations or a fictional listener profile.
Use credible explanations of recommender systems, personalisation and music discovery. Verify claims about how a specific platform works.
Use a flow diagram, slides, one-page explainer, audio or video.
