Algorithms Make Surprisingly Good Creative Directors

This programmer didn’t know what to do for her thesis–so she designed a computer program that would tell her.

Everyone’s been there–struggling to think of an idea. Whether that’s for a new design, a creative project, a piece of writing, or even what to do this weekend, being blocked is a quintessentially human experience. So when it came time for Nicole He, a student at NYU’s Interactive Telecommunications Program, to decide what to do for her thesis project, she thought: Why can’t a computer do this for me?


It started as a joke. But writing a computer program that would generate ideas for creative projects sounded like a worthwhile challenge to pursue. And so for her thesis, He created an algorithm that would identify the best art project to do at a particular moment in time, which would then print out a receipt with a prompt for a project–then He would create an art piece that fulfilled the computer’s requirements. She documented the project, called The Best Art, via Twitter and Tumblr, completing a total of 27 projects over several weeks.

[Photo: courtesy Nicole He]
“Make fake flowers feel loved,” the algorithm ordered on one day. He bought a bouquet of real flowers and presented them to a bouquet of fake ones. “With a pair of earrings, design something that wastes,” it decreed on another. She created earrings made of $20 bills. “Using Trump’s foreign policy, create something essential,” it demanded. She created a list of resources to help Syrian refugees. “Build art out of a photo album and a thinly sliced raw beef,” it commanded. He filled an old school photo album with slices of raw beef.

To build the algorithm, He first thought about what conditions might influence a human trying to come up with an idea–how they’re feeling, what the weather’s like, and perhaps what’s happening in politics. Then, she imagined what factors might influence a computer coming up with art projects using similar proxies–the current weather in Brooklyn, her computer’s horoscope (she says it’s a Gemini), the number of Trump tweets that day, and the distance between the Earth and the International Space Station. Using these parameters, the algorithm comes up with a number to describe that day, against which it measures 1,000 different project ideas.

To create the potential project ideas, He scraped words from Kickstarter, U.S. patents, and old ITP projects, then let the algorithm plug those words into sentence structures she’d written. Then, the algorithm would perform a sentiment analysis on each prompt to determine its level of positivity or negativity and assign it a number based on this sentiment. Of the 1,000 project ideas the algorithm generated each time the program was run, whichever number was closest to the number describing the “best” art for that day would be spit out via a receipt.

“I spend all my time telling my computer what to do,” He says. “One thing that was sort of fun for me to explore originally was the idea of reversing the role of a human and computer in the creative process.”


[Photo: courtesy Nicole He]
For He, her somewhat arbitrary methodology and the decisions she made are meant to point out just how much algorithms are dictated by their designers. “The whole thing is sort of absurd, and it’s also a very much a reflection of my personality,” she says. “My hope is that people can think about how other algorithms, more important, more powerful algorithms, reflect the people who built them. They don’t come out of nowhere, they’re made by people.”

The Best Art algorithm also inverts our expectations of computers’ roles in our lives. He says that over the course of the project, she found that its bizarre creative briefs stimulated her creativity. “It’s much easier to be creative when you have constraints. If someone tells you, make something creative, that’s way harder than if someone says, do something creative with thinly sliced beef,” He says. “I found the algorithm to be really helpful for the creative process.”

That might cause her to come back to the algorithm in the future. “It could be useful for a creative exercise if you want to make something but you don’t want to come up with an idea,” she says. And what creative person hasn’t had that experience?

He isn’t the only one to invert the relationship between artist and machine (or user and machine). The ongoing Twitter bot Art Assignment Bot tweets out art ideas every hour–and some people have been inspired enough to follow them. The Swedish artist Jonas Lund let an algorithm dictate what he should make in the 2013 work The Fear of Missing Out, but his algorithm analyzed successful contemporary artworks and gave him a series of instructions on how to make similarly “successful” art.

Meanwhile, He says that the reaction to the project online has been largely positive, and some people have commented on how they would have interpreted a prompt differently. She believes her project and the others like it help people think about their computers in a different light.


“People are actually interested in the idea of a computer as a character, as a personality,” she says. “We all sort of have these personal and intimate relationships with our computers.”

About the author

Katharine Schwab is an associate editor at Co.Design based in New York who covers technology, design, and culture.