When Data-Driven Decision-Making Leads to Bad Decision-Making and Worse Problem-Solving

Good data science begins with good questions.

Question:  Who owns the decision made by data-driven decision-making?  People at the beginning or the end of the algorithm-building process?  (Business Analysts, Application Architects, Developers)  People at the beginning or the end of the human implementation of the decision?

As a frequent flyer, I was both fascinated and horrified by what happened on United Airlines flight 3411 from Chicago to Louisville.   As reported, the flight had been fully boarded, when United Airlines decided that they needed to bump four boarded paying passengers in order to replace them with United employees who would be needed to crew a future flight in another location.  When the United gate agents couldn’t persuade passenger David Dao to “voluntarily” get off the plane, they summoned Chicago aviation police to put him into the hospital.  Dr. Dao’s fellow passengers recorded the event, and the videos “went viral.”  Matters were made worse when both United’s current CEO and its parent company’s former CEO made public statements that the public perceived to imply that a beating is what a passenger should expect if they push a contract dispute with the airline.  The consumer news cycle became dominated with analysis of the passenger videos. Meanwhile the business press criticized the CEO statements, the resulting PR crisis, and airline management’s clear problem with strategic vision.

The press has kept repeating a phrase that four passengers were “randomly selected” to be bumped off the flight.  The four passengers weren’t randomly selected, they were selected by an algorithm.  Airline experts have acknowledged that algorithms are used to select passengers primarily based upon frequent flyer status, ticket price, and passenger arrival time, as well as other factors such as handicap status.  In other words, passenger data is used to help drive a secret decision as to which passengers should be bumped.  In my opinion, the airline gate agents may have stuck to their perception that they were making a data-driven decision enforced by processes, when they should have been focused on solving a problem.  Effective problem-solving demands thinking and rewards creativity.  Effective problem-solving requires perspective.  The full problem may be more than just be the problem in front of you, but also the resulting issues in the future.  (In United’s case, a social media backlash in China over perceived abuse of an elder Asian-American had triggered Chinese government criticism in a place United perceives as important to strategic business growth.)

When building new data-driven decision-making processes, we must ensure that the humans making the actual decisions can see where the boundary is between decision-making and problem-solving.

Having spent years working with services contracts, I have gotten some chuckles reading consumer misconceptions voiced as social media comments.  Many commentators believe that an airline passenger is essentially buying or renting a part of an airplane.  In reality, passengers are entering into a complex contract for services, which the airlines call a “contract for carriage.”  Like most services contracts, these contracts for carriage include all kinds of terms covering both parties’ responsibilities.  While the mechanics of working with the airline include popular terms such as “reservation,” “flight,” and “seat,” the passenger is really contracting to be moved from Point A to Point B as the airline is best able.

(Image provided by geralt at Pixabay.)