Relational Risk in Humanoid Robotics | Sara Filipcic & Roshni Lulla

In this talk

1. 0:09

   Deployment depends on the people affected

   Roshni Lulla contrasts impressive robot demonstrations with the harder question of whether people will welcome robots in daily life. In a hospital, a buyer may approve a pilot, while staff, patients, and families experience the robot’s effects. She says those people may not be consulted even though they bear the consequences of failures.

2. 3:10

   Assessing robots before and after launch

   Lulla says the Institute helps robotics companies understand how people respond to their products and evaluate them before deployment. Sara Filipcic describes checks of robot conduct around people, information collection and access, simulated failures, and performance in real settings over time. The resulting records can support discussions with buyers and regulators.

3. 7:06

   Lessons from social technology and empathy research

   Filipcic points to harms associated with social media and emotionally engaging AI companions as examples of what can happen when consequences are overlooked. She also describes Moxie, a child-focused robot whose devices stopped working after its company failed, leaving families without a planned transition. Her research distinguishes recognizing another person’s feelings from sharing them, and she says language models can be tuned toward antisocial responses.

4. 13:48

   Acceptance as part of product development

   Lulla recommends addressing acceptance early, arguing that documented answers on safety and privacy can help buyers decide and reduce later pilot rework. Filipcic says adoption depends on trust among decision-makers, staff, and the people who encounter robots each day. The Institute is developing an IEEE standard focused on trust and safety in human-robot relationships, with particular attention to homes, families, education, and children.