Social influence on can do more than speed up hits. Under the right conditions it can decouple what people listen to from what they initially preferred. The paper distinguishes two kinds of influence: narrow influence, where people are exposed only to content aligned with their preferences, and wide influence that exposes people to content outside their usual repertoires. To produce collective outcomes that are unpredictable from initial preferences—and bring about cultural change—mere strength of influence is not enough.
Over a million users
Looking at more than a million Spotify users, the study shows that when wide influence is operative people are exposed and susceptible to taste-transcending choices from peers. Then surprising collective outcomes can emerge.
For this to happen, the study shows that partial taste overlap between the senders and the recipients of social influence is crucial. Too little similarity in music taste means low susceptibility; too much similarity no novelty. Partial overlap helps balance exposure to novelty and willingness to adopt it. The empirical analysis shows this trade-off: exposures from more similar peers are more influential while less similar peers tend to expose listeners to more novel artists.
Peer-to-peer influence
The research combines three strands of analysis on Spotify data containing publicly shared user playlists, follower ties, and adoption events: First, the authors employ topic modelling, a widely used technique from natural language processing, to map artists to musical genres and measure users’ musical tastes. Based on these measures, second, they conduct high-dimensional statistical matching to estimate peer-to-peer influence while controlling for taste and selection effects. This is no easy task, as in observational data influence and homophily, humans’ tendency to connect to and follow others with similar interests and tastes, is intertwined. Finally, the study uses agent-based computer simulations to examine how the micro-level influence mechanism found empirically scales up to macro outcomes. Simulated macro dynamics demonstrate that when networks contain partially overlapping taste connections, unexpected and unpredictable market outcomes emerge. Networks with only identical or only dissimilar ties do not produce the same decoupling.
Diversity
The results refine our understanding of cultural diffusion: success and inequality in cultural markets are not solely driven by intrinsic quality or by simple popularity cascades. Instead, network structure and the width of influence—how peers expose and make novel options acceptable—also shape whether previously unlikely items can become major hits. This has implications for industry strategy and theories of social contagion more broadly. Knowledge about social-influence mechanisms can also inform the design of algorithmic newsfeeds on social media platforms that suffer from one-sided views and the spreading of misinformation: exposing users to content that is different from what they initially believed in, but maintains partial ideological overlap—sufficiently close to ensure trust and sufficiently distant to broaden exposure—can help spread more politically diverse, and hopefully, less biased news.
