Autonomous Transit: Weighing Benefits and Considering Costs

By Matt Ferguson

On hearing the phrase “artificial intelligence and transportation,” the first thing that likely comes to mind is self-driving cars.  They sound a bit like science fiction come to life, but are within reach enough to seriously think about the potential positive impacts on people’s lives –  the convenience of driving without actually having to drive! – and the ethical questions this new technology raises.

As I’ve thought about transportation and artificial intelligence, however, I’ve found myself wondering less about how AI should work (such as how a self-driving car should react in a situation) and more about if certain tasks should be managed by AI at all.  I’m certainly not opposed to expanding the role of AI in transportation – I think AI has already made useful changes to how we travel and has the potential to make many more beneficial changes. Rather, I wonder about the consequences of removing people from processes. 

There are some immediately obvious benefits to handing off tasks to AI. An algorithm cannot get tired or bored or distracted, it will never be in a hurry to finish so it can beat rush hour traffic home; handing over tasks to AI removes the possibility of human error.  In the aviation industry, much of the flying time is already handled by autopilot, with human pilots taking the controls only for takeoff and landing, and there is potential for AI to take over tasks such as safety checks and fuelling, among others. Increased efficiency would likely benefit airlines’ profit margins as well as passengers’ travel experience. 

Rail systems, both over and underground, are also already being changed by AI. Driverless or partially automated underground or light rail trains operate in cities across the country with more expansions on the horizon, and in October 2021, Germany debuted the first fully driverless overground train. Much like AI in the aviation industry, the roles of AI in trains center on improving efficiency, potentially increasing the frequency and quantity of trains, and removing human error.  Using AI to improve public transportation could also help to significantly decrease carbon emissions by enticing more people to travel by bus or train rather than drive.

Removing human error and inefficiency, though, comes at a cost.  The tasks that AI might take over are largely jobs that are currently held by people; automating those tasks could have a significant impact on available employment, which then has further consequences for people’s abilities to support themselves and negatively impact quality of life.  Is the disruption to and potential loss of large sectors of employment worth the benefits that AI would bring? Who stands to benefit the most from AI and who would bear the brunt of the side effects?

Particularly in the case of public transit, I wonder about the interpersonal impact of having autonomous trains or buses.  Many times I’ve been on a city bus and overheard the bus driver give directions to a disoriented tourist, seen them wait a moment longer at a stop for someone running to catch the bus, or, on occasion, de-escalated a tense situation with a passenger.  These are largely small things, but especially after two years of life in the virtual world, I am more aware of the importance of those small moments of human interaction and kindness.

Finally, there is the question of what happens when something goes wrong – the algorithm miscalculates and doesn’t put enough fuel in the airplane, the electricity goes out and there’s no manual override to open the train doors, the AI system is hacked and a city’s entire public transportation system is held hostage.  In the case of malfunction or a system error, does the fact that it was an unsupervised machine that erred make it less bad than if a human made that same mistake?  Or, approaching the issue from a different perspective, why should fear of machine error prevent us from implementing AI in areas where it would be beneficial? At what point do the benefits outweigh the disadvantages?

No matter how much AI does or doesn’t come to play a role in transportation, there will likely always be some degree of risk, some margin of error that even the combination of human oversight and creativity and machine precision cannot escape. Similarly, there will be benefits and drawbacks that we collectively must weigh as we determine how and how much transportation becomes run by AI.  The important thing, I think, is that we start asking ourselves these questions now so we can shape the development and expansion of AI in an ethical way.