Skip to main content

Daniel McFadden delivers Moses Lecture

Weaving Human Choice into Transportation

 In a field once dominated by Origin-Destination tables and cordon counts – simplistic data ancestors from another world compared to today’s available methods – the transportation industry now holds more data at its fingertips than ever before.

From highway sensors and automated toll systems to GPS tracking of individuals and vehicles, contemporary transportation researchers enjoy access to a deep array of information that can drive planning and management of travel by land, sea, or air. There is, however, one critical, complex variable that can derail even the most thoughtful, informed, and data-fueled plans: human choice.
 
“Transportation is a technological behemoth bedeviled by human behavior,” Nobel Memorial Prize laureate and professor Daniel L. McFadden told a crowd of some 100 Northwestern University students, faculty, and transportation industry personnel at the 2019 Leon N. Moses Distinguished Lecture in Transportation on November 12 at Northwestern’s Lutkin Memorial Hall.
 
Hosted once every two years by the Northwestern University Transportation Center (NUTC), one of the world’s leading interdisciplinary education and research institutions devoted to the better movement of materials, people, energy, and information, the Moses Lecture honors the legacy of the late Leon N. Moses, whose 46-year tenure as a Northwestern faculty member included a four-year stint as NUTC director.
 
McFadden’s appearance follows the likes of previous presenters such as Northwestern economist Robert Gordon and University of California, Irvine, economist Jan K. Brueckner in bringing compelling questions and insights to the Moses Lecture forum.
 
“In the areas of transportation and behavior, McFadden’s name stands out,” said NUTC director Hani Mahmassani, a former student of McFadden’s at the Massachusetts Institute of Technology.
 
Choice in Transportation
 
Titled “Attend! Consider! Decide! What Planners and Machines Must Learn to Predict Travel Behavior,” McFadden’s 90-minute program detailed notable early endeavors in transportation data designed to inform planning and management, including a 1970s era effort to develop systematical causal models of individual travel behavior.
 
“Transportation has been a source of problems, ideas, and solutions throughout my career,” said McFadden, who cited his work on 1973-1977 project to predict Bay Area Rapid Transit ridership as a particularly meaningful initiative.
While research has evolved and newfangled technological tools have propelled fresh thinking in subsequent decades, McFadden said human choice and behavior continue to test the transportation industry.
 
From the bicycle peloton to trading pencils, McFadden, the presidential professor of health economics at the University of Southern California, illustrated examples of choice and its influence on behavior.
 
“We are challenged by choice,” said McFadden, who was awarded the Nobel Prize in Economic Sciences in 2000, along with James Heckman, for the development of theories and methods for analyzing discrete choice. “People like to have choices, but don’t want to make choices.”
 
And those individual choices, whether it’s deciding to change lanes on the highway or to drive rather than bike to the office, complicate transportation design and management.
 
Beyond the Machines
 
Though planners, researchers, and policy makers alike might be tempted to think machine learning alone will solve longstanding transportation quandaries, McFadden said technology, while promising and helpful, isn’t such a clean problem solver. Like humans, he noted, machines tend to attribute causal structure to correlations and “overfit” their “model” of system outcomes.

“Enthusiasts of machine learning are naïve about what these machines can do,” he said. “One has to be skeptical [about machine learning], though there is great opportunity in it.”

To harness the potential of data resources and processing power to better predict behavioral response to interventions and inform transportation decision-making, McFadden said human planners and their machine counterparts must be prepared to dig into human behavior to “sharpen the focus of policy interventions to improve transportation efficiency and environmental performance.”

He suggested, for instance, training machines to recognize various structures of behaviors or feeding machines different patterns to spark better predictions about what will happen when a freeway lane closes or a transit authority considers altering bus or train service.
 
“Behavioral travel demand studies, updated to be truly behavioral, should … use simulated population behavior generated by behavioral studies as core elements in learning,” McFadden concluded.

Back to top