Sound Verdict

Streaming

What streaming recommendation does to what you hear

Recommendation systems do not merely reflect taste. They shape it, and the research on how is more critical than the industry framing.

Updated 2 min read 12 citations Evidence strength 3/5

The optimisation target problem

Recommendation systems optimise for engagement — plays, skips avoided, session length. That is a legitimate business objective and it is not the same as helping you find music you would love but do not yet know about. Something familiar-adjacent reliably beats something unfamiliar on next-play probability, so the system learns to narrow.

Research on engineered inequality in musical taxonomies and streaming recommendation [5] examines how classification and recommendation together advantage some artists structurally. Studies of streaming uniformity and language diversity [2] approach the same concern at the level of what a catalogue actually surfaces.

Recommendation effects on diversity

The counter-effort

Not everyone in the field is content with this. Work on promoting unfamiliar music through data science [4] treats discovery of the unknown as the explicit objective rather than as a side effect. It is a genuinely hard problem: the metrics that would reward it are long-horizon and hard to attribute.

Personalised streaming systems continue to develop architecturally [1], and service-quality measurement scales for streaming apps have been built [3] — useful because they let platform quality be assessed on something other than catalogue size.

Practical listening

  • Follow people, not algorithms. Critics, labels and DJs have taste with a direction. Recommenders have a gradient.
  • Listen to albums in order. Sequencing is a compositional decision that shuffle discards.
  • Deliberately seek the unfamiliar. Nothing in a personalised feed will make you uncomfortable, and discomfort is where taste expands.
  • Pay attention to language and region. Uniformity research [2] suggests catalogue surfacing is narrower than catalogue size implies.

Common questions

Is my taste being narrowed by algorithms?
Research on recommendation, diversity and streaming uniformity suggests the concern is well founded [2][5].
Are algorithmic playlists useless?
No — they are excellent at finding more of what you already like. That is a different job from discovery.
Does streaming pay artists fairly?
Work on engineered inequality [5] addresses structural advantage in surfacing, which is upstream of the payment question.
Does audio quality matter?
Above a moderate bitrate, differences are small in blind listening. Mastering and playback system matter more.

References

Every citation below links to the original peer-reviewed record on PubMed or via DOI. Nothing here is a substitute for medical advice.

  1. Aura Music: A Scalable Personalized Music Streaming System Using Time-Weighted Recommendation and Hybrid Filtering Vardhan T · International Journal for Research in Applied Science and Engineering Technology · 2026 · Journal article DOI
  2. Algorithmic neogeneralism in streaming platforms: recommender systems and content strategy Pescatore G · Frontiers in Communication · 2026 · Journal article DOI
  3. Streaming Uniformity and Language Diversity on Netflix and Prime Video Streaming Platforms Florea S · Historia y Comunicación Social · 2026 · Journal article DOI
  4. Promoting Unfamiliar Music Through Data Science: MARS, the Music Affect Recommender System for Digital Library Engagement Bainbridge D, Dean R · Leonardo · 2025 · Journal article DOI
  5. ENGINEERED INEQUALITY: MUSICAL TAXONOMIES AND STREAMING RECOMMENDER SYSTEMS Campos Valverde R · AoIR Selected Papers of Internet Research · 2025 · Journal article DOI
  6. Rethinking the filter bubble? Developing a research agenda for the protective filter bubble Erickson J · Big Data & Society · 2024 · Journal article DOI
  7. Diversity and Serendipity Preference-Aware Recommender System Yin K, Zhao J · Journal of Computational and Cognitive Engineering · 2024 · Journal article DOI
  8. Music Recommender System using Autorec Method for Implicit Feedback Irawan M, Baizal Z · JURNAL MEDIA INFORMATIKA BUDIDARMA · 2023 · Journal article DOI
  9. Music Recommender System Based on Play Count Using Singular Value Decomposition++ Ramadhan M, Wibowo A · JURNAL MEDIA INFORMATIKA BUDIDARMA · 2023 · Journal article DOI
  10. Music Recommender System Using ChatBot Sakore S · International Journal for Research in Applied Science and Engineering Technology · 2021 · Journal article DOI
  11. Critique Generation to Increase Diversity in Conversational Recipe Recommender System Abbas F, Najjar N, Wilson D · The International FLAIRS Conference Proceedings · 2021 · Journal article DOI
  12. Encouraging Attention and Exploration in a Hybrid Recommender System for Libraries of Unfamiliar Music Taylor J, Dean R · Music & Science · 2019 · Journal article DOI