In the Age of A.I., What Is Taste? And Do We Still Have It?

AI taste algorithms replace human curators, New Yorker asks who decides style

Sarah Mills | GlobalBeat

The New Yorker published a 6,500-word essay examining whether artificial intelligence recommendation systems are eroding individual aesthetic judgment.

Writer Kyle Chayka argued that algorithmic feeds on Spotify, Netflix and TikTok condition users to accept pre-filtered choices rather than form personal taste.

The piece revived a decades-old debate about mass culture versus personal discernment, now focused on machine-learning models that predict preferences from behavioral data.

Spotify’s Discover Weekly reaches 120 million users, Chayka reported, citing company figures released in February. The playlist blends collaborative filtering with audio analysis to surface tracks similar to a listener’s history. He wrote that the feature drives 20% of total streams on the platform, a metric Spotify confirmed to investors last year.

Netflix deploys 1,300 “taste communities” that group viewers by habits rather than genre, the article noted. An internal slide deck leaked in 2023 showed the firm tests different thumbnail images for the same title up to 30 times to maximise click-through. Chayka quoted an unnamed engineer saying the goal is “eliminating the need to choose.”

TikTok’s For You Page refines recommendations every 3 minutes from dwell time, replays and swipes, according to company blog posts cited in the essay. Chayka observed that creators who reverse-engineer the signal often adopt identical visual tropes, flattening creative range. He cited data-tracking firm Sensor Tower showing average daily usage on the app reached 95 minutes in 2024.

The writer interviewed 23-year-old Atlanta resident Maya Patel, who said she deleted Spotify after noticing her playlists contained “the same breathy female vocal every Monday.” Patel told the magazine she had not independently sought new music in 2 years. Chayka used the anecdote to illustrate passive consumption replacing active exploration.

Art historian Elizabeth Prettejohn told the magazine that 19th-century department stores triggered comparable anxieties. Prettejohn said ready-made fashions once allowed middle-class shoppers to “purchase taste by the yard,” echoing current fears about algorithmic shortcuts. The comparison framed curation anxiety as recurring whenever technology widens access to culture.

Chayka quoted sociologist Pierre Bourdieu’s 1979 study “Distinction,” which argued that taste functions as social currency. Bourdieu wrote that cultural choices reinforce class boundaries, a dynamic Chayka said algorithms now automate by segmenting users into marketing clusters. He warned that machine sorting could entrench existing inequalities by limiting exposure to unfamiliar styles.

The article distinguished between “recommendation” and “curation.” Recommendation predicts what a user already likes, Chayka wrote, while curation introduces the unexpected. He cited record-store clerks, art dealers and DJs as human curators who risk personal credibility when suggesting the unfamiliar. Algorithms, by contrast, optimise for retention metrics, he noted.

Critics quoted in the essay accused streaming services of steering users toward cheaper licensed content. Music-industry analyst Tatiana Cirisano told Chayka that Spotify’s discovery features favour tracks for which the company pays lower royalty rates, citing a 2023 Midia Research report. The magazine said labels now pitch songs labeled “algorithm-friendly,” prioritising consistent tempo and length under 3 minutes.

Chayka acknowledged his own reliance on algorithmic feeds. He confessed that 40% of his weekly listening originates from Spotify, a percentage he logged manually over 1 month. The disclosure undercut nostalgic calls for pure human discernment, instead urging readers to interrogate how much autonomy remains within machine-mediated choice.

Technology ethnographer Chanda Prescod-Weinstein argued that algorithms deprive marginalized creators of discovery, the article noted. Prescod-Weinstein told Chayka that recommendation systems trained on historical preference data reproduce dominant cultural norms, making it harder for Black and queer artists to surface without already existing buzz. The scientist urged platforms to audit outputs for demographic skew.

The essay closed by suggesting small acts of resistance: disabling autoplay, seeking physical media, and manually constructing playlists for friends. Chayka wrote that genuine taste formation requires friction, inconvenience and occasional disappointment. He concluded that reclaiming agency starts with “choosing to choose,” even when an easier feed awaits one click away.

Background

Algorithmic culture can be traced to 2006 when Netflix launched a $1 million public contest to improve its recommendation engine, the article recalled. The competition popularised collaborative filtering, a technique that infers preferences by comparing users with similar histories. Spotify hired members of the runner-up team in 2013, accelerating industry-wide adoption of machine learning for content delivery.

Concerns about mechanised taste surface periodically with new distribution formats. The New Yorker noted that 1950s broadcast television triggered fears of passive couch-potato culture, while 1990s compact-disc “shuffle” buttons were criticised for fragmenting the album. Each cycle debates whether technology liberates or constricts individual judgment, a pattern Chayka argued repeats today with AI feeds.

What’s Next

European Union regulators plan to enforce audit rules for large recommendation systems starting in August, Chayka noted. Platforms operating in the EU must disclose risk assessments and allow external researchers access to key algorithms, provisions that could reveal how taste profiles are constructed. Compliance filings are due 6 months after the Digital Services Act takes full effect.