Come i sistemi di raccomandazione influenzano la cultura
Recommendation systems have become the primary gatekeepers of cultural experience for billions of people worldwide — a shift with consequences so profound that it has begun to reshape not only what culture people consume but what culture gets made.
Annunci
When 67% of music listeners discover new artists through streaming services according to the IFPI Global Music Report 2024, with algorithm-generated playlists cited as the primary discovery mechanism across all age groups, the algorithm has effectively displaced the radio programmer, the record store clerk, the film critic, and the cultural editor as the dominant force shaping cultural taste at scale.
The substitution is not merely administrative — it represents a fundamental change in the logic governing which cultural works reach audiences, because algorithmic selection optimizes for engagement metrics that are not identical to, and often conflict with, the values that human cultural curation has historically attempted to serve.
Research by Hesmondhalgh and colleagues published in 2023 found that recommendation algorithms tend to reduce cultural diversity by reinforcing existing preferences, a pattern confirmed by genre diversity data from Spotify’s Viral 50 Global chart between 2018 and 2024, which showed a progressive decline in genre variety and increasing domination by pop-oriented sounds that marginalizes experimental and regional genres.
A quantitative study by Spotify’s own Research Scientists confirmed that algorithmically generated recommendations significantly reduce consumption diversity, a finding that aligns with what Eli Pariser called “filter bubbles” — echo chambers reinforced by algorithms that isolate users not only informationally but culturally.
Annunci
What is at stake in this dynamic is not merely consumer preference but the ecology of cultural production itself, because the culture that algorithms reward is the culture that increasingly gets made.
The Architecture of Recommendation: How Systems Choose What You See
Recommendation systems are not neutral tools that match people with content they would independently prefer — they are optimization engines designed to maximize specific metrics on behalf of specific commercial interests, and understanding their architecture is essential to understanding their cultural effects.
Collaborative filtering — the dominant approach in most major recommendation systems — suggests content based on the behavior of users deemed similar to the current user, creating feedback loops in which popularity amplifies itself and obscurity becomes self-reinforcing, regardless of the actual quality or distinctiveness of the underlying content.
Spotify’s recommendation system, powered by convolutional neural networks and natural language processing, curates playlists that favor tracks with what industry insiders call “algorithmic streamability” — short intros, repetitive structures, immediate hooks, and mainstream genre positioning — characteristics that maximize the engagement metrics the algorithm is optimized for rather than characteristics associated with artistic innovation or cultural significance.
Netflix’s recommendation logic operates through a similar commercial framework: research on its system’s algorithmic logic found that it constructs cultural taste through opacity rather than transparency, engineering self-generative cultural processes that undermine what theorists of cultural democracy describe as “direct and spontaneous engagement with cultural goods.”
The opacity dimension is critical to understanding the algorithm’s cultural power: users experience recommendations as personalized suggestions that reflect their own preferences when they are in fact the output of commercial decisions about which content partners receive favorable positioning, which genres the platform is investing in, and which engagement behaviors the system is optimizing for.
Major labels negotiate favorable playlist positions, independent artists compete on uneven terrain regardless of musical quality, and the 2024 MIDiA Research report confirmed the outcome of these asymmetries: independent artists constitute 80% of Spotify’s catalog but receive only 20% of recommendation-driven streams.
++ Il ruolo degli hobby nella costruzione dell'identità
The Homogenization Effect: What Gets Lost When Algorithms Choose
The most documented cultural consequence of recommendation system dominance is homogenization — the progressive narrowing of the cultural diversity that reaches audiences as algorithms converge on the formats, structures, and stylistic characteristics that generate the most engagement at the greatest scale.
The “15-second rule” that has restructured songwriting for TikTok compatibility illustrates this effect with unusual clarity: data analysis of viral TikTok songs from 2021 to 2024 shows that intros have shortened dramatically, hooks appear almost immediately, and repetition dominates to ensure memorability within the attention window that the platform’s recommendation logic rewards.
A song like Adele’s “Someone Like You,” whose chorus arrives after nearly a minute of emotional development, would struggle for algorithmic traction in the current short-form recommendation environment regardless of its artistic quality — a structural incompatibility between algorithmic optimization and certain forms of artistic ambition that affects every creative decision made by artists aware of how the system works.
In markets like Brazil, genres like samba and forró are overshadowed by global pop and reggaeton through algorithmic momentum that rewards what already has the most global engagement data — a dynamic that erodes local musical heritage not through explicit censorship but through the structural disadvantage of competing against content whose global scale gives it an algorithmic advantage that local specificity cannot match.
| Cultural Domain | Homogenization Effect | What Gets Lost |
|---|---|---|
| Musica | Pop dominance, short-hook structure | Regional genres, experimental forms |
| Film/TV | Sequel and franchise preference | Auteur cinema, local storytelling |
| Notizia | Engagement-driven outrage content | Investigative, contextual journalism |
| Books | Bestseller self-reinforcement | Literary midlist, translation literature |
| Visual art | Aesthetic convergence on high-engagement styles | Difficult, challenging work |
The pattern across cultural domains reveals a consistent structural logic: recommendation systems reward content that generates immediate engagement from the largest possible audience, which systematically disadvantages content whose value depends on prior knowledge, sustained attention, tolerance for difficulty, or cultural specificity that limits its immediately addressable audience.
The Creator Response: When Art Adapts to the Algorithm
The cultural influence of recommendation systems extends beyond consumption patterns to production decisions, as creators across all cultural domains increasingly make artistic choices based on algorithmic compatibility rather than purely on creative judgment.
Musicians producing “Spotify-friendly” tracks — formulaic, immediately accessible, structurally optimized for the streaming engagement metrics that determine algorithmic promotion — represent the most visible version of this adaptation, but the same logic operates in film, television, publishing, and visual art wherever algorithmic distribution determines economic viability.
Filmmakers aware that Netflix’s recommendation system favors content with high completion rates — a metric that rewards familiar genres and conventional narrative structures — face commercial pressure to produce work that the algorithm can readily recommend rather than work that challenges or surprises audiences in ways that might reduce completion rates even while increasing cultural impact.
UNESCO has documented the threat that algorithmic cultural distribution poses to cultural diversity as a public good, noting in its frameworks on cultural expression that the commercial optimization logic of recommendation systems systematically disadvantages the cultural forms most dependent on public investment and public protection — precisely those whose value is understood across generations rather than optimized for quarterly engagement metrics.
The phenomenon researchers describe as “algorithmic precariousness” — the instability of cultural work produced by dependence on algorithmic favor that can shift without notice — represents a structural transformation of the conditions of cultural production, one whose long-term consequences for the range and ambition of human cultural expression are only beginning to be understood.
++ Come gli algoritmi influenzano ciò che vediamo e ciò in cui crediamo
The Paradox of Personalization: More Choice, Less Discovery
The marketing narrative of recommendation systems centers on personalization — the idea that algorithms give each user more precisely what they want, liberating them from the generic mass culture that human editorial gatekeepers imposed on diverse audiences.
The research consistently complicates this narrative, revealing a paradox at the heart of algorithmic personalization: users receive more content aligned with their demonstrated preferences while encountering less content that would expand, challenge, or redirect those preferences — a combination that feels like more choice while producing narrower actual cultural experience.
Unlimited catalogs create decision paralysis, and algorithms solve this by reducing effective choice — presenting manageable options from impossible abundance while making users unaware of what they never see, which is to say, making them unaware of the choices they are not being offered.
The “accidental discovery” that characterized pre-algorithmic cultural consumption — browsing a record store, stumbling across an unfamiliar author in a library, watching a film recommended by a friend with different taste — produced exposure to cultural works that recommendation systems optimizing for preference similarity would never suggest, because those systems are designed to minimize the risk of disengagement rather than to maximize the range of experiences that exposure to diverse culture provides.
Research on Netflix’s algorithmic logic noted that it reduces opportunities for what scholars call “epistemic friction” — the productive difficulty that arises from engaging with diverse perspectives and content that resists immediate comprehension — precisely the quality that cultural education has traditionally identified as essential to cultural development rather than mere cultural consumption.
Toward Algorithmic Cultural Accountability
The cultural consequences of recommendation system dominance have begun generating regulatory and institutional responses that reflect a growing recognition that the commercial optimization logic of private platforms cannot be the sole determinant of cultural distribution at social scale.
The UK government’s inquiry into streaming services found that algorithms “could reflect biases that may subsequently reduce new music discovery, homogenise taste and disempower self-releasing artists” and called for greater oversight of the opacity of algorithmic curation — a regulatory recognition that the cultural effects of recommendation systems are a matter of public interest, not merely a commercial design decision.
The concept of “cultural citizenship” — the right of people to access diverse cultural expression as a dimension of civic life rather than merely as consumers of algorithmically curated entertainment — has gained traction in policy discussions, with some researchers arguing that recommendation systems should be required to incorporate diversity metrics alongside engagement metrics in their optimization objectives.
The European Union has included cultural diversity considerations in its Digital Markets Act and related regulatory frameworks, reflecting an institutional judgment that the cultural power of major platforms cannot be treated as purely commercial rather than as a matter with implications for democratic culture and social cohesion.
++ L'etica della progettazione dell'economia dell'attenzione
The technical solutions being proposed — diversity-aware recommendation algorithms, transparency requirements about recommendation logic, user controls over the degree of personalization — are promising but face the structural challenge that they require platforms to accept reduced engagement optimization in exchange for cultural goods that their business models have no direct incentive to value.
Conclusione
Recommendation systems influence culture not as neutral infrastructure serving pre-existing preferences but as active forces shaping what gets made, what gets distributed, what gets discovered, and what kinds of cultural ambition are economically viable in a world where algorithmic promotion determines commercial success.
The homogenization documented across music, film, news, and visual art reflects the predictable consequence of optimizing cultural distribution for engagement metrics rather than for the values — diversity, challenge, local specificity, artistic ambition — that human cultural curation has historically attempted to serve alongside commercial viability.
The creator adaptation to algorithmic norms — the shortened intros, the formulaic structures, the algorithm-friendly aesthetic choices — represents a cultural cost that is diffuse and difficult to quantify but reflects a real narrowing of the conditions under which certain forms of cultural ambition become economically possible.
What the evidence ultimately suggests is that leaving the cultural distribution of billions of people to systems optimized exclusively for commercial engagement, without transparency, accountability, or diversity requirements, produces cultural consequences that no democratic society would endorse if they were made explicit rather than embedded invisibly in infrastructure that most people experience only as personalized convenience.
Domande frequenti
1. What are recommendation systems and how do they influence culture? Recommendation systems are algorithms that select which content users see on streaming and social platforms. They influence culture by determining which creative works reach audiences, what artistic formats are rewarded commercially, and which cultural forms receive the algorithmic promotion that determines economic viability in the digital distribution era.
2. Do recommendation algorithms homogenize cultural taste? Yes, research consistently shows they do. Studies including work by Spotify’s own Research Scientists confirmed that algorithmically generated recommendations significantly reduce consumption diversity, with genre diversity data from Spotify’s Viral 50 Global chart showing a progressive narrowing between 2018 and 2024.
3. How do recommendation systems affect what artists create? Creators increasingly make artistic decisions based on algorithmic compatibility — shorter intros, more repetitive structures, mainstream genre positioning — to maximize the algorithmic promotion that determines commercial reach. This produces what researchers call “algorithmic precariousness” and a narrowing of the artistic forms that are economically viable.
4. Is personalization actually giving users more or less cultural diversity? Less, despite appearing to offer more. Recommendation systems reduce effective choice by optimizing for preference similarity rather than discovery, making users unaware of what they never see and eliminating the accidental exposure to unfamiliar culture that pre-algorithmic browsing produced.
5. What regulatory responses are emerging to address recommendation systems’ cultural effects? The UK government has called for greater oversight of algorithmic curation’s opacity. The EU has incorporated cultural diversity considerations into its Digital Markets Act. Researchers have proposed diversity-aware algorithms, transparency requirements, and user controls as technical responses to the commercial optimization logic that currently governs recommendation system design.
