Overview of Implicit Interactions
In this episode of Data Skeptic: Human Computer Interaction, Kyle Polich speaks with Wendy Ju, professor at Cornell Tech, about designing systems that interact naturally with people—especially robots and autonomous vehicles. The conversation centers on implicit interactions: the nonverbal, often unnoticed ways humans communicate intent, expectation, and response in everyday life. Ju explains how her research uses prototypes, simulations, and “Wizard of Oz” methods to study future technologies before they are fully built, with the goal of making machines behave in ways people can understand and trust.
Key Themes and Takeaways
Designing for real human behavior
- Ju argues that many products fail because builders focus on making systems work technically before testing how they work socially.
- Her work emphasizes staging future scenarios and observing human reactions early, rather than waiting until the final product stage.
- She frames interaction design as deeply contextual: what works in a lab may fail in the real world.
Implicit interaction as a framework
- Implicit interactions are exchanges that happen without explicit commands or language.
- Examples include:
- moving toward an automatic door to signal it should open,
- sliding a cup forward to request more coffee,
- using subtle cues while driving to indicate intent.
- Ju maps these interactions along two axes:
- attention: background vs. foreground,
- initiative: reactive vs. proactive.
Machines need social signaling
- A major insight from Ju’s work is that machines often lack the subtle signals humans rely on to coordinate safely.
- Examples include:
- turn signals on cars,
- motion cues from a robot that indicate whether it will move,
- audio or visual signals that communicate intent before a system acts.
- Without these cues, users may feel confused, unsafe, or forced to guess.
Notable Research and Examples
The “Wizard of Oz” method
- Ju uses hidden human operators to simulate autonomy in early-stage testing.
- This lets researchers study interactions as if the system were autonomous, even when a person is secretly controlling it.
- The method helps reveal how people respond to robots or vehicles in realistic settings without requiring a fully finished system.
The ghost driver protocol
- One of Ju’s best-known examples is her work on autonomous vehicles.
- She and her team created a setup where a person dressed as a car seat appeared to be driving an autonomous car, while the car seemed empty from the outside.
- This allowed them to study how pedestrians and cyclists interact with driverless vehicles.
- Key finding: many people do not make eye contact with drivers as often as commonly assumed; instead, they often rely on cues like the car’s motion and deceleration.
“Normative behavior” vs. normal behavior
- Ju distinguishes between:
- normal behavior: statistical average behavior,
- normative behavior: the informal actions people take to enforce social norms.
- Examples:
- New Yorkers subtly correcting tourists in crowded pedestrian spaces,
- people standing on the right side of escalators,
- drivers learning local merging norms when entering a new city.
- These small corrections help newcomers adapt to shared expectations.
Autonomous Vehicles: Main Concerns
What current AVs still lack
- Ju says many autonomous systems still model people too crudely, often as blobs or simplified objects.
- They often fail to understand:
- orientation,
- intent,
- social context,
- timing of human movement.
Safety and public trust
- Autonomous cars are often described as safer by selectively citing “at-fault” accidents.
- Ju points out that they can still have high rates of rear-end collisions and other interaction failures.
- She criticizes the idea that “people will adapt” to machine behavior, arguing instead that machines should adapt to human norms.
Why signaling matters
- A vehicle that slows down or pauses can unintentionally signal to pedestrians that it has seen them.
- This can cause people to step into the road too early, leading to dangerous timing conflicts.
- Ju stresses that safe autonomy requires clear, early, and socially legible signaling.
Academia, Design, and the New Cornell Department
Why design needed its own home
- Ju says academia is often organized around traditional disciplines that don’t fit design research well.
- She sees the new Cornell design tech department as a place for:
- cross-disciplinary research,
- design centered on real-world problems,
- collaboration across engineering, materials, environments, and interaction design.
The kind of students it serves
- The department welcomes students from many backgrounds, including engineering, bioengineering, law, and design.
- Ju says the best candidates tend to have:
- a strong point of view,
- unusual insight,
- the ability to see problems others miss,
- and the ability to communicate those insights clearly.
Methods and Philosophy
Empirical, field-based design
- Ju describes herself as an empiricist.
- She believes real-world field testing is essential because interaction is highly contextual.
- Her research combines:
- prototype testing,
- simulations,
- observational studies,
- and theory-building from observed behavior.
Free cues we take for granted
- Humans rely on signals that may seem invisible until they disappear:
- engine revving,
- body posture,
- vehicle motion,
- subtle changes in timing.
- Electric and autonomous vehicles often remove these cues, which means designers need to intentionally replace them.
What’s Next for Wendy Ju
- Ju says her work on cars, robots, and human-behavior modeling is very much ongoing.
- She is helping build a broader research network focused on understanding people before designing computational systems for them.
- She also wants to bring more researchers into this space and expand the community working on implicit interaction and human-centered autonomy.
Where to Follow Her Work
- LinkedIn: her primary social platform
- BlueSky: occasional updates
- Vimeo / YouTube: videos of research projects and demonstrations
- She also publishes papers, often with accompanying visual materials to make the work more accessible
Bottom Line
This episode argues that the future of robotics and autonomous systems depends less on raw technical capability and more on understanding the unspoken rules of human interaction. Wendy Ju’s work shows that if machines are going to coexist with people in public spaces, they must learn to communicate as humans do: subtly, contextually, and with the right amount of initiative.