Iñigo Urteaga got his first glimpse of the power of statistical modeling at New York’s Stony Brook University while feeding algorithms corrupt datasets and exploring what elements of ground truth could be recovered. “You can play God and see what the truth is,” he said.
Now an associate research scientist in the applied physics and mathematics department at Columbia University, Urteaga is working with a team using similar methods on a project with ovulation-tracking company Clue to explore how self-reported data might be used to better predict the timing of menstrual cycles.
“If we can’t explain our work to them, it’s hard for them to accept or trust it.”
“We’re only seeing the so-called corrupt versions of the data — because users aren’t always tracking consistently — and we’re using that to try and make predictions,” he said.
To him, the biggest challenge is ensuring the results of any model he works on can be clearly explained to the end user, especially when it can impact health.
“If we can’t explain our work to them, it’s hard for them to accept or trust it,” Urteaga said.
Part of the problem is that the algorithms these data inform are never 100% correct. And while some models hide that uncertainty, others use thoughtful language to explain the findings and, most importantly, outline their limitations, he said.
The most common misunderstanding he encounters about his research is that algorithms are perfect, and as a result, produce perfect solutions to problems. By including those limitations, he said, researchers can better ensure that the end users of their models fully understand their meaning and won’t be surprised if and when the algorithm misses the mark or fails to provide the answer they were looking for.
Urteaga, originally from Spain, can be found riding his road bike when he isn’t working or studying. If he had his pick, he’d rather be at a live rock concert, but with many of those cancelled due to Covid-19, activities like barbecues and outdoor picnics have helped him stay centered. “Those have kept my sanity,” he said.
— Erin Brodwin