Lauren Leek says predictive algorithms can make culture feel alike

In short: Data scientist Lauren Leek argues that systems built to predict what people will click or buy can quietly push everyday culture toward more familiar, easier-to-copy options.

What's going on

Lauren Leek makes this case in an essay called Temperature Zero for Culture. She says the issue is not that personalization always gives everyone the exact same thing. It is that prediction systems often prefer options that are easier to forecast, recommend, fund, and repeat at scale.

She points to London pubs and shopping streets. In one analysis of 640 London high streets and about 18,000 food businesses, she found that two streets shared around half of their “food profile” on average. Chains were a major reason streets looked similar, because chain menus and storefronts repeat across neighborhoods.

She also looks at pop music lyrics. Using more than 4,000 Billboard Hot 100 hits from 1950 to 2015, she reports that the richness of vocabulary inside songs fell by more than a quarter starting in the early 1960s. Repeated lines rose by nearly a third. She adds a key point, songs became more repetitive within themselves, but it was not simply that all songs became identical to each other.

Leek describes a feedback loop. If a system is tuned to predict clicks or sales, it will often promote familiar, “legible” choices (like a safe bet). Those choices then get more attention and investment. People end up choosing from a narrower menu, which gives the system even more evidence to keep recommending the same kinds of things.

What to watch

Leek’s warning is that what gets “flattened” depends on what a system is optimized for. As more businesses use prediction tools for financing, recommendations, and planning, it may get harder for unusual options to reach an audience in the first place.

Source: NYTimes