Georgina Born, Professor of Anthropology and Music, Dept of Anthropology and Institute of Advanced Studies, UCL
Fernando Diaz, Research Scientist, Google; Adjunct Professor, McGill University
Andres Ferraro, Postdoctoral Fellow, McGill University and Pandora
Gustavo Ferreira, Postdoctoral Fellow, McGill University
Music recommender systems (RS) are central to global streaming services like Spotify, Apple Music and Amazon and are the dominant means for many consumers to curate and discover music. By employing data on consumer behaviour and repeatedly influencing consumer choices, RS can shape a population’s consumption trends and tastes. RS introduce a type of ‘monitoring-based marketing’ (Andrejevic 2012) at a much larger scale and more rapidly, recursive, and intensively than earlier, non-computational market research methods. RS, therefore, represent a new stage in the automation of everyday music consumption (Bowker and Star 1999, Gillespie 2014).
Currently, these systems optimise for listener engagement and retention (Chen et al. 2019) following a logic of ‘similarity’ enacted by ML. However, this logic ignores system design’s social and cultural implications (Mehrotra et al. 2018). Issues like algorithmic bias – which listeners and creators are optimised for and which are not (Baym 2013, Morris 2015, Crawford 2016a, Baeza-Yates 2018) – are symptomatic of major problems in design (Danks & London 2017, Springer et al. 2018), due in part to the practice of turning social dimensions as indirect effects (Overdorf et al. 2018, Callon 1998). Consumer bias may be introduced by the amplification of dominant groups, with likely long-term effects in shaping musical taste and identities. For musicians, algorithmic distribution incentivises what type of music to produce, potentially privileging the generic over the different, posing long-term effects on creativity (Born, Diaz et al. 2021, Anderson 2014). Some of these problems are intrinsically linked to twin features of the existing recommender system logic – commercialisation and personalisation – and their ideas of music listening and what a human is (Prey 2018, Stark 2018, Seaver 2022).
To deal with these issues, four core, related elements defined our project. We propose, first, that recommender design moves beyond a solely commercial orientation towards goals of human musical flourishing and public good (Moe 2008, Andrejevic 2013, Hesmondhalgh 2013, Born 2018). Second, we propose that alongside personalisation, recommender design should develop ways of analysing and modifying aggregate and cumulative influence of RS to achieve the above goals. Third, addressing these two proposals, we have developed a new metric called ‘commonality’, which measures the degree to which recommendations familiarise a whole user population with previously chosen content categories (Ferraro, Ferreira, Diaz & Born 2022). Fourth, as a key case study in reflexive’ values in design’ (Knobel and Bowker 2011, Fish and Stark 2021), we designed this metric adopting normative principles guiding public service media systems (diversity, universality, cultural citizenship). The goal is to enhance the diversity of recommendations by boosting a whole user population’s familiarity with underrepresented music categories.
These four elements resulted from workshops between recommendation engineers (Ferraro and Diaz) and SSH scholars (Ferreira and Born), which generated mutual translation between CS and SSH, spanning high-level methodological, epistemological, and normative questions to concrete design issues. Such sustained interdisciplinarity has enabled insights from SSH to be integrated with those from CS and vice versa. The team is committed to innovating in socially and culturally responsible design in recommender systems and seeks to contribute to a growing body of scholarship developing public good rationales for digital media and machine learning systems.