Human-Machine Interaction (HMI) in Social & Public Environments

UX Prototyping & Student Mentoring | MIT Media Lab (2016-Present)

BACKGROUND  |  Increasingly, cities around the world are seeking ways to reduce their dependency on cars in an effort to curb congestions and pollutions.  Among the many available strategies that facilitate modal shift is the increasing pedestrianization and transformation of city roads to support active mobility (e.g. biking and walking). This shift, however, also introduces an emerging constraints to future autonomous vehicles: these new robotic systems will need to not just reduce in size to become more human-scale, but also succeed in operating in human-dominated environments by adopting to their corresponding behavioral, social, and cultural norms.  In this research, in contrast to the more traditional approach of training AI systems that use limited--or even biased--data sets, we take what Iyad Rahwan calls the “society-in-the-loop” approach[1] to creating AI systems.  In that, we look to the real world for inspiration to help us develop ways for a micro autonomous vehicle—using the prototype of the Persuasive Electric Vehicle (PEV)—to achieve three basic objectives: 1) convey its awareness of the physical and social environment; 2) convey its intention as a vehicle, and; 3) develop expressions that facilitate the building of trust and peace of mind with the people around it.  These three capabilities, we believe, serve a critical foundation to enabling a sustainable coexistence and collaborative relationship between humans and machine. 

PROCESS  | As early as early 2016, we deployed the PEV prototype in uncontrolled bike lanes and sidewalks around MIT as a test for initial reception from the student community.  By August 2017, the deployment shifted to bike lanes and sidewalks that were actively occupied by cyclists and pedestrians to allow potential scenarios in which uninformed persons could encounter the PEV in motion (Figure 1).  This provided us the opportunity to have conversations with people in-situ about the technology and about their expectations, fears and hopes.  To ensure that communities in other parts of the world--in other cultural and urban contexts outside of the MIT ecosystem in Cambridge, Massachusetts--can develop trust in the autonomous vehicle, we began to conduct pilots internationally through our network of collaborating cities and states, which consist of Andorra, Hamburg, Helsinki, Shanghai, and Taipei.  The most recent field study took place in March 2018, at Taipei’s Daan Forest Park, where we tested last-mile passenger pick-up scenarios on bike lanes during commuting hours, under broad daylight and at night. There, we observed the reactions and actions of a large sample of cyclists, pedestrians and joggers as they come in close contact with the vehicle.

RESULTS | The deployments in real-world bike lanes highlighted several critical scenarios in which some form of socially acceptable communication method would greatly facilitate the exchange between the PEV and people: when the PEV and a pedestrian or cyclist are on a path to potentially collide with one another, how might the PEV signal its awareness and intent to the approaching party?  If there are multiple parties approaching, how might the PEV behave differently than how it would to a single party? In addition to traffic yielding, we also explored interaction features combining Computer Vision, lighting, and audio for potential ways of encountering between the PEV and people in general scenarios (Video 1-2). An iteration of the PEV’s eye-contact and hand-gesture recognition feature became a live interactive exhibit at the Cooper Hewitt Smithsonian Design Museum in 2018 as a part of its future mobility exhibition (Video 3).