Page 31 - spring18
P. 31

f it continues in its lane, it will plow into   It’s more prudent, then, to ask probing
           Iseveral pedestrians. If it swerves into the   questions such as: Is an algorithm biased? Is it
           adjacent lane, it will ram into a concrete barrier.   manipulative? Are there hidden moral or political
           In either case, injuries and even loss of life are   issues?
           likely.                                     An example of bias, Bambauer said, is seen in
              Although this scenario currently lives only   how scores from the computer program COMPAS
           on Moral Machine, a platform designed by the   (Correctional Offender Management Profiling
           Massachusetts Institute of Technology to “gather   for Alternative Sanctions) are used to create a
           a human perspective on moral decisions made by   risk scale for criminal recidivism. Although
           machine intelligence,” it might be coming soon   the program’s 137 variables do not include race
           to a street near you. And that has Jane Bambauer   or ZIP code, the scale still has bias built into
           both fascinated and concerned by what she   its computations, as Bambauer demonstrated
           describes as the current “Wild West” of artificial   through a series of bar graphs.
           intelligence.                               Even the perception of bias can turn people
              “The coordination of a world with both   away from trust in our institutions, she said,
           driven cars and driverless cars will be incredibly   showing how basic Google searches can be
           complicated,” Bambauer, a University of Arizona   interpreted by some as an indication of the tech
           law professor, told a group of science teachers and   giant’s political leanings.
           graduate students after her lecture in February   Manipulation can surface in something as
           on “Machine Influencers and Decision Makers”   innocuous as food reviews or as insidious as
           at Centennial Hall. The lecture was the fifth   “fake news.” Facebook’s news-feed algorithm
           in the College of Science series on “Humans,   “has incredible power,” Bambauer said, adding
           Data and Machines,” which has focused on the   that “the filter bubble is limiting you” in terms of
           convergence of the digital, physical and biological   big-picture perspective.
           worlds.                                     In any event, algorithms are complicated to
              Bambauer said the transportation industry   untangle.
           will be turned upside down in the same way   “All of the problems (with algorithms) are
           that it was when automobiles disrupted a world   interconnected,” she said. “Accuracy might
           of horse-drawn carriages, forcing both modes   increase bias. Data gathering might affect privacy.
           to share the road. She said it will be tempting   The problems are with setting priorities among
           to rush in and regulate — but much better to go   competing goals.”
           slowly.                                     Echoing what previous speakers in the
              “My default position with new technology is   series had said, she noted, “It’s easy to blame the
           not to do much heavy-handed regulation,” she   algorithms, but algorithms do what we ask.”
           told the teachers and students.
              “I’m a bit of a contrarian in my field. My
           impulse is to let companies figure out what’s
           working and what isn’t, before we regulate. There   Bambauer said the transportation industry
           are instances where, when trying to regulate in
           advance, you end up missing on innovations (that   will be turned upside down in the same way
           follow),” she said, citing the early World Wide
           Web as an example.
              Bambauer told her audience that it’s useless   that it was when automobiles disrupted a
           trying to fight the onslaught of algorithms.
           They’re pervasive, they’re not going away,   world of horse-drawn carriages, forcing both
           and they’re assessing our credit scores, career
           interests, health care and more.             modes to share the road.
              “We interact with machine-learning
           algorithms almost any time we do anything on
           the internet,” she said.




                                                                                                              SPRING 2018 29
   26   27   28   29   30   31   32   33   34   35   36