The Robotic Perspective in Human-Robot-Interaction

Human-robot interaction is typically designed from the user’s perspective – shaping a robot’s behavior, „personality,“ and relational dynamics to align with human expectations. As robots evolve from tools to companions [1,2,3,4], new opportunities—and challenges—emerge in designing meaningful interactions.

In my master’s thesis, I explored how the robot’s perspective might inform interaction design. Rather than adapting robots solely to human needs, I examined how technical characteristics of robotic systems—such as technical constraints and behavioral tendencies—can shape interaction.

I examined three types of learning-based robotic systems—user-feedback-driven reinforcement learning [5], inverse reinforcement learning [6], and instructive systems [7]—and interpreted them as distinct robotic characters:

The Wild One represents a reinforcement learning system that learns through trial and error, guided by user feedback. Curious and autonomous, this character exhibits unpredictable behavior, especially early on. Its potential for deep personalization comes with the demand for high user involvement and patience during its “training” phase.

The Attentive embodies an inverse reinforcement learning system that imitates user demonstrations and gradually refines its actions. It is fast, competent, and responsive to patterns—but limited in creativity. Its reliability stems from strict adherence to its teacher’s example, which constrains its behavioral range.

The Assistive reflects an instructive system capable of interpreting commands and executing complex tasks independently. It is efficient, dependable, and adaptive—yet socially distant, only engaging when explicitly asked. This character prioritizes function over relationship, offering minimal interaction unless required.

In parallel, I interviewed users of robotic vacuum cleaners to develop personas that reflect different human attitudes and expectations: the efficiency-driven, the entertainment-seeking and the multi-tasking.

By pairing each persona with a robotic character, I designed three speculative interaction scenarios. These were prototyped using hardware and software mockups—including a hacked Roomba—and presented as fictional advertising videos.

The project highlights how considering both user needs and robotic traits can inspire richer, more diverse interactions—where unpredictability and imperfection are not limitations, but opportunities for meaningful engagement.