Thursday April 19, 2018

#HHNYC: #Algorithms, #Art and Authorship

Jer Thorp (photo: S. Mazzocchi)

By Sherry Mazzocchi

When designing the 9/11 Memorial, architect Michael Arad didn’t want the names of the dead arranged in obvious patterns. He wanted them placed randomly.

That didn’t go over well with the next of kin, or first responders. Fireman and policemen wanted to be with their own kind. Companies wanted their employees together, so did family members. Assembling the roughly 2,900 names with all of the different relationships was a huge task.

But artist and educator Jer Thorp saw it as an extension of a “Miss Manners” problem—arranging people at a big event when you know something about who can and cannot be seated next to one another. Thorp, an artist and educator at The Office for Creative Research in NY, solved the problem with algorithms.

Speaking at the Hack/Hackers NY Meetup in Soho Thursday, Thorp defined algorithms as a set of instructions that “we do until a condition is solved—‘a do/until.’” The combination is similar to wedding vows, he said, which are “a type of algorithm that, if you execute them successfully, you both die still married.”

Algorithms solve problems that can’t be cracked with sheer computational power. They build systems that find answers in interesting ways. With a background in cell biology, he’s especially fond of genetic algorithms that allow for mutations to find the best solution.

To resolve the name placement problem, the memorial team found out what names should be next to each other on the memorial—or “meaningful adjacencies.”

The simplest arrangemens were pairs—such as brothers or mothers and daughters. But the towers were places where people spent a lot of time, so there were several other built-in relationships.

Thorp created an algorithm that clustered people who had to be together.

Of the 2,900 names, there were about 1,400 adjacency requests, including a set of 72 names that needed to be together. There were also 600 Cantor Fitzgerald employees. In addition, one name from the south pool had to be connected to another on the north pool. These were two friends; one was on a plane and the other in a tower.

The designers needed to span the buildings’ footprint without breaking connections. Thorp built a data visualization tool. That allowed the two friends to be the shortest distance possible from one another.

Thorp also created a second, space-filling algorithm. He said, “It’s the extension of what’s called a ‘backpack problem’—how do you get irregularly shaped pieces into a space?” That algorithm iteratively filled the space with names. The two algorithms created the optimal layout where names fit and adjacencies were fulfilled.

Another team of mathematicians also tried to solve the adjacency problem, but they were only 93 percent successful.

Thorp’s solution worked because it wasn’t solely a math problem. “We also understood that this was a design problem,” he said.

The names weren’t blocks—they were type. Names ending with “T” and starting with “J” wouldn’t fit around the half-inch expansion joints that were added to the parapet to prevent the metal from buckling. Punctuation also had to be taken into account. So they wrote an algorithm adjusting for that.

What seemed like a restriction ended up adding more room. “By being able to stretch and move a little bit, we were able to get over the rigid problem. And that’s why the mathematicians only got 93 percent,” he said.

“At the end of the day,” he said, “this was a super-human problem.” But it revealed the use of algorithms as both a creative and a computational instrument.

Thorp said people often believe algorithms aren’t human artifacts–that they are owned and operated by computers. Instead of claiming authorship, people blame bad computational results on something other than themselves. “I think that’s a deep fallacy that we need to get around,” he said.