A University of Washington seminar, “Calling BS in the Age of Big Data,” promises to help students develop a BS detector — and it’s become a global phenomenon, with universities as far away as Australia planning to teach a version of it this fall.

Share story

Did you hear about the researchers in China who said they’d developed an algorithm that could predict whether somebody was a criminal by scanning a photo of their face?

The researchers used “fancy machine learning” to eliminate human biases and come up with a scientific way to determine criminality by examining facial features, University of Washington professor Carl Bergstrom told his class in Mary Gates Hall one day last month.

“What do you guys think?” Bergstrom asked.

Watch online

The UW is providing free online access to the BS lectures on the UW Information School’s YouTube channel. (Note: Videos feature frequent repetition of a minor profanity.) Here’s the link: http://callingbullshit.org/videos.html

In unison, more than 100 students responded out loud: “Bullshit!”

Most Read Stories

Unlimited Digital Access. $1 for 4 weeks.

Exactly.

Biology professor Bergstrom and Information School professor Jevin West are co-teaching a seminar this quarter that’s all about BS — or, precisely, “Calling BS in the Age of Big Data.”

The 10-week seminar class has achieved the academic version of a chart-topping pop single: At the UW, it reached its 160-student capacity shortly after registration opened this spring.

Its reach across the globe is even more impressive. Colleges in Canada, France, Portugal, England and Australia have contacted the professors about teaching a version of the course this fall. A partial list: Carnegie Mellon University, Hofstra College, Iowa State University, Stanford University, Columbia University, UCLA, Oxford University (England), Universite de Nantes (France), Murdoch University (Australia), University of Guelph (Canada) and Instituto Piaget (Portugal).

Currently, a Catholic high-school teacher in Minnesota and a tech college in Copenhagen are teaching the course, and the University of Alaska Anchorage will teach it this fall.

The course syllabus is freely available, as are YouTube videos of the lectures.

West and Bergstrom are especially enthusiastic about helping high-school teachers use the BS course materials to develop classes. “Ultimately, that’s very important to us, because this is material that shouldn’t wait until your senior year of college,” West said.

The professors attracted widespread attention by using a profanity in the title of the course, but the interest is real. So far, the first video of their class lecture has been viewed more than 12,000 times.

Calling BS is not a class about fake news, and it’s not a political- science course. Rather, the professors want students to develop a skeptical eye about big data and its limitations, particularly in the sciences. They’re teaching how to recognize results that are just too good to be true.

In their fifth lecture, Bergstrom and West explored why Google’s much-vaunted ability to predict a flu outbreak turned out to be, well, BS.

They picked apart the Chinese criminality project. They examined the mystery of why — with the exception of September 11 — the 11th day of the month is referenced less frequently in books and literature than almost any other day of the month, at least as measured in a Google book-scanning project. (Think you know the answer? Read to the end.)

One of their main themes: If the data collected for use in an analysis isn’t solid, the result won’t be, either. “Garbage in, garbage out,” West told the class.

In the case of the algorithm that claimed to predict criminality, the Chinese researchers used pictures of incarcerated and innocent people to create a mathematical formula to identify people in each category.

But, as Bergstrom noted, the non-criminals were wearing faint smiles; the criminals had subtle frowns etched into their faces. “I have this hypothesis that this is a smile detector,” he told the class.

As for Google Flu Trends — Google’s instantaneous map of influenza hot spots, created using the geographic locations of searches for flu-like symptoms — the company appeared to be predicting outbreaks as much as two weeks faster than the Centers for Disease Control, which relied on confirmed reports of influenza from doctors and medical centers.

“When I read this article, I remember taking it to my class and saying, ‘Big data really is amazing,’” West said. “Big data is changing the world.”

But Google Flu Trends soon became the poster child of big-data hubris and overreach, West said. As time ticked along, researchers realized that the tracker was overestimating flu cases most of the time, and in one case, overestimated it by more than 50 percent.

“In fact, what they found is that if you just took a simple model, like tracking the temperature, it would do a better job” of predicting flu outbreaks, West said.

The two professors bring a sense of humor to their lectures.

“We’re making it fun — making it entertaining, theatrical, to make it less intimidating,” West said. Because the class is made up of students from 40 different majors, the material is short on technical detail and long on developing a common-sense BS detector.

Next fall, Calling BS will be offered as a three-credit course at the UW, and the professors hope to include guest speakers — climatologists and doctors, for example — along with videos from experts across campus who can talk about what BS looks like in their fields.

“I want to tell them meaningful stories in the class, that they’ll tell people after dinner,” Bergstrom said. “These are the conversations we want to start.”

They’re also making available a sanitized version of their materials (which uses the word “bull” instead of BS) that could be used in middle or high school. And they’re hopeful that a new law Gov. Jay Inslee signed last month, which encourages schools to teach media literacy, will draw teachers to their site.

As for why the 11th day of the month is so infrequently referenced in scanned texts: Google’s algorithm confused the number 11 with other look-alike characters — capital I’s, lowercase l’s, even the letter n — and often failed to recognize the number in the books it scanned. Give yourself an A if you got that right.