FYI Book Club: ‘Weapons of Math Destruction’ gives readers cause for concern

Every number tells a story.

The most recent gathering of the FYI Book Club had readers discussing Cathy O’Neil’s “Weapons of Math Destruction,” a lament on the overuse and misuse of complex, often flawed algorithms and mathematical models — and the sometimes scary impact of numbers on our own lives.

“I had no idea so many algorithms were used,” said Jerry Bakker of Olathe. “People don’t understand the math models. They think that since it is math and the answer comes out of a computer, it must be true. But not necessarily.”

Added Pat Schneider of Kansas City: “The pervasiveness of gathering data and using it wherever a business can is a little frightening and annoying. We all complain about the Facebook algorithms and the ads on Google. These data miners are gathering information from so many places. It makes me want to go into Google and search all sorts of interesting things just to mess with the research.”

The room exploded in laughter, and attendees started listing outlandish search terms they would plug into Google.

“This mathematical tool of modern technology is increasing the disparity of income,” said Clif Hostetler of Kansas City, Kan. “It’s enraging how these algorithms cause businesses to target insecure and economically disadvantaged people.”

Janelle Sjue of Kansas City wasn’t surprised that lower-income populations are the target. “The algorithms look at correlation and not causation,” she said. “We have to separate the person from race.”

On the bright side: “GPS works well for people who need it. Spam filters clear out bad emails. And some courts are using new algorithms to reduce recidivism. But the answers are not always reliable, and people don’t understand the limitations of the mathematical models.”

Susan Tuncten of Kansas City also noted the helpful side of the financial algorithms. “On the positive side is the FICO (credit rating) score. It’s transparent. You know what to do to increase it or improve it,” she said. “That’s somewhat fair.”

Hostetler asked, however, “Was it more fair than the small-town banker who knew everybody? Maybe FICO is fairer than the good-old-boy network.”

For Peggy Brockschmidt of Kansas City, “The scariest chapter was the one on for-profit universities. The most vulnerable people are the ones who need the loans, and they get damaged.”

Said Tuncten, “There are so many variables among the students, it’s impossible to measure or compare accurately.”

Her daughter, Sara, chimed in: “These algorithms are created to make sense of data, but instead of improving our understanding, it’s as if people are changing their own behavior to make the algorithm a fact and not a guideline. College administrators are changing how they conduct academic business or oversee the educational aspects and institution’s culture.”

Attendees gathered at the Mid-America Regional Council (MARC) offices, where senior researcher Jeff Pinkerton talked about the use of data to help city agencies and other entities anticipate fluctuations in population, employment and transit, among other issues.

“Our computer model said only 7.3 of you would be here tonight. But I see 14, so we’ll have to go back and calibrate the formula,” Pinkerton joked.

He said MARC doesn’t use mathematical models as they are depicted in “Weapons of Math Destruction.” “We don’t really use such algorithms, but we do review the way we use our data,” he said. “We’re trying to take the real world and make numbers out of it but be cautious about how we present data and make the right decisions.”

Overall, readers said, they found O’Neil’s book accessible, engaging and informative.

“I loved the way she wrote about math,” Tuncten said. “And I wasn’t surprised at the way it impacts our daily lives. Data tends to make my eyes roll, but it was fun the way she wrote about data, and it was understandable.”

“O’Neil gave good guidelines for what constitutes a ‘weapon of math destruction,’ ” Schneider said. “The feedback loop is not appropriately designed or is completely missing. A good algorithm that’s good is one that could be changed. A bad one won’t be changed.”