Music recommendation systems can struggle to recognise traditions with fewer recordings or different musical structures, an opinion article in The National argues. The piece points to a practical challenge for AI research: better recognition may need musicians’ expertise, not just more data and cleverer algorithms.
Watch Desk analysis
What happened
The 10 November article describes how recommendation systems work well with extensively recorded music, but can struggle with forms such as Emirati Al Ayyala and Arabic maqam, whose melodic intervals do not always fit Western equal-tempered tuning. The author argues that systems trained mainly on Western music may be less able to identify other traditions.
At NYU Abu Dhabi, researchers are combining computational methods with musical knowledge. One example is work to distinguish 13 types of stroke played on the Carnatic mridangam drum. The article also describes research using computational analysis to explore rhythmic relationships across musical traditions, including work on Candomblé and West African music. Read the article in The National.
Why it matters
A recommendation system can only help listeners find what it can recognise. If its training material and assumptions favour some musical traditions, the catalogue may be large while parts of its cultural range remain hard to discover. That is a question about both technology and what gets heard.
The article’s examples suggest a way forward: bring musicians and researchers into the work, and build methods around the musical structures being studied rather than expecting one model of music to fit all. The piece makes a case, not a comparative evaluation of recommendation systems.
Our read
This is a welcome reminder that “more data” is not always the whole answer. Musical knowledge can help researchers ask better questions of the recordings, and the results could make overlooked traditions easier to find without treating them as curiosities on the edge of someone else’s playlist.
What to watch
- Whether the NYU Abu Dhabi research produces tools that improve recognition of regional musical forms.
- How the first Arabic music dataset described in the article is used and shared.
- Whether recommendation services make their coverage across musical traditions easier to assess. Activity teaser: Music recommendation systems can struggle to recognise traditions with fewer recordings or different musical structures. A National opinion article describes research that combines computing with musicians’ knowledge, including work on identifying 13 types of stroke on a Carnatic drum. It’s a useful challenge to the idea that algorithms simply need more of the same data. Should recommendation platforms be responsible for helping listeners discover traditions their systems currently struggle to recognise?
Discussion spark: Should recommendation platforms be responsible for helping listeners discover musical traditions their systems currently struggle to recognise, or is that a job for researchers and cultural institutions?
Sources and evidence
- Source update (10 November 2026, 07:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.