EXAMPLE PAGE - SCHOOL BROCHURE - DEMOCRACY - Flipbook - Page 52
LIFE WILL BE BETTER
WITH FAIRER MACHINE
LEARNING AND AI
The problem: Humans may be capable of rational,
logical decision-making, but plenty of details can
derail the process, says Karthik Kannan, Thomas
Howatt Chaired Professor in Management at Purdue’s
Krannert School of Management. For example, prior
research has shown that judges mete out harsher punishments when they’re hungry or tired — seemingly
regardless of the merits of a given case.
The promise: Machine learning and AI offer a better
approach, at least in theory. Create the right algorithm,
and robots will churn out better answers. They’re endlessly productive, never get tired, and are logical by design.
Unintended consequences: Machines may have the
benefit of being machine-like, but biases baked into
the system can get compounded with machine learning. Karthik notes that Amazon’s machine-learning
recruiting engine, for example, systematically penalized women applying for roles, even when engineers
aimed to make the programs neutral to gender-related
terms. In addition, algorithms typically aim for identical outcomes among specific subgroups — even if one
subgroup has to work significantly harder to achieve
the same outcomes as another.
A new way forward: Kannan’s research focuses on
improving AI and machine learning not simply by
manipulating the criteria used to ensure fairness — like
“color blindness” — but to focus on the process, including ways that people might adapt their behavior as a
result. “AI-based tools won’t remove all problems,” he
says. “They’ll take away some, but they’ll introduce
different types of errors.
Our goal is to minimize the problems that
might occur because of
machine learning.”
50 PUR D U E A LUMNUS
LIFE WILL BE BETTER
WITH SAFE DRUGS THAT CAN
REPLACE OUR DIMINISHING POOL OF
EFFECTIVE ANTIBIOTICS
A range of conditions that were once treatable with a simple course of antibiotics — certain strains of staph, gonorrhea, and C. diff, for example — have evolved to
resist the antibiotics that once decimated them. It’s not just unfortunate. In thousands of cases, it’s proven to be deadly. This new reality puts doctors in a difficult bind.
But new research by professor of microbiology Mohamed Seleem is showing particular promise. Seleem and his team are screening more than 4,000 FDA-approved drugs to
see if they might effectively treat certain antibiotic-resistant infections.
The benefit of such an approach is twofold, says Seleem. First, antibiotics act as
something like a nuclear bomb to bacteria in the body, demolishing everything in their
path. Other treatments are more likely to act as surgical strikes, zeroing in only on the
harmful bacteria and leaving the rest unharmed.
Second, because Seleem is testing drugs that have already been approved by the FDA,
any promising findings will likely see a smoother path for eventual use. “This is a safe,
cost-effective way to repurpose a drug that has already been approved. It’s fast, efficient,
and it could save lives,” says Seleem.