Data Skeptic

Implicit Interactions

45 min

Wendy Ju, a professor at Cornell Tech, discusses her research on implicit interactions in human-robot interaction and autonomous vehicles. She emphasizes the importance of understanding unstated, nonverbal cues in design, and advocates for early-stage testing using methods like Wizard of Oz to improve how autonomous systems communicate and behave. Key insights include how people don't always make eye contact with drivers, challenging common assumptions, and the need for systems to signal their intent proactively.

Summarized by Podsumo

🎧 Listen 🎙️ Ask about this episode

✨ Key Takeaways

1

The 'ghost driver' method was invented to study how pedestrians interact with autonomous vehicles by having a person in a car seat costume drive, revealing that people often don't look at the driver and instead rely on subtle cues like deceleration.
2

Implicit interactions are defined by tensional demand (attention) and initiative (reactive vs. proactive), with successful systems moving between foreground and background to avoid overwhelming users.
3

Autonomous vehicles get rear-ended 10 times more often than human drivers due to poor perception and modeling of human intent, highlighting a need for better signaling and testing.
4

Testing early with prototypes, simulations, and deception (e.g., Wizard of Oz) is critical to uncover hidden needs, like a robot needing to lift before moving to signal its intent.
5

Wendy Ju advocates for a new design department at Cornell that bridges disciplines to address real-world problems, seeking students with strong vision and cross-disciplinary skills.

💬 Notable Quotes

"There is a way in which the way that things transpire is based on an understanding which is different than reality. So there's that level of what we could call deception."
"The thing that we all know and we all remember that we all do when we interact with drivers at the intersection turns out not to be the normal interaction. That is the error recovery."
"I don't think academia is actually well set up for design to make research contributions, which is what I've wanted for a really long time."