What privacy-preserving features does PRIVATAR offer for self-representation and classroom presence?

Avatar-based self-representation lets the child choose how they appear in the classroom—such as a face overlay with emoji, an animated 2D avatar, or a motion-driven 3D avatar—with all image processing performed locally and a live preview shown before transmission. Gesture recognition lets classmates hide their own faces by giving the robot a thumbs-down, triggering a privacy overlay. Spatial no-go zones prevent movement into sensitive areas, and there are visual and audio cues to indicate when additional observers are in frame.

What are the five core design requirements derived from the participatory design workshops for PRIVATAR?

Transparency and access control; Privacy-aware mobility; Privacy-preserving self-representation; User-controlled privacy management; Social and contextual signaling.

Where and when was PRIVATAR demonstrated and published?

The PRIVATAR demonstrator was presented at MuC 2026 (the German-language human-computer interaction conference) and published as a demo paper at Mensch und Computer 2026 in Duisburg (DOI: 10.18420/muc2026-mci-demo-277).

Paper

PRIVATAR: Privacy-friendly Mobile Avatars for Sick School Children

Reading time: approx. 2′ 42″

Design Ethics

Human Centered Design

Universal Design & Inclusion

The Challenge

Sick children attending school via telepresence robots are more than a technical problem. The robots capture classrooms, teachers, and classmates through their sensors — while the ill child simultaneously exposes their own background: their home, or their hospital room. Privacy here affects everyone at once, making it an inherently social challenge.

Existing systems address this dilemma with static, paper-based consent forms. That's not enough. Privacy is context-dependent and situational — it shifts multiple times over the course of a single lesson. What's missing are dynamic, intuitive mechanisms that give all stakeholders active control — without extra effort, without additional devices, without barriers.

Our Approach

At MuC 2026 — the leading German-language conference on human-computer interaction — we presented our PRIVATAR demonstrator live. It is the result of three participatory design workshops with teachers, school administrators, media educators, parents, and children. From these workshops, we derived five core design requirements:

Transparency and access control — Who is part of a session? Who can listen in?
Privacy-aware mobility — Free movement through the classroom, with clear spatial boundaries around sensitive areas.
Privacy-preserving self-representation — Expressing social presence without having to reveal your face or your surroundings.
User-controlled privacy management — Active control for everyone, without accounts or additional hardware.
Social and contextual signaling — Simple ways to signal attention, availability, and readiness to interact.

In collaboration with the University of Göttingen (Computer Security & Privacy), the University of Duisburg-Essen (Interactive Systems), and the University of Bonn (Humanoid Robots Lab), we translated these requirements into a working system.

Our Solution

The PRIVATAR demonstrator connects three components: the OrionStar Minibot as a physical avatar in the classroom, a browser-based web app for the child at home, and an Android application running on the robot itself.

What sets it apart:

Avatar-based self-representation. The child chooses how they appear in the classroom — as a face overlay with emoji, as an animated 2D avatar, or as a motion-driven 3D avatar. All image processing happens locally on the child's device. A live preview shows exactly what is being transmitted — before it's transmitted.

Gesture recognition for classmates. Anyone who doesn't want to appear in the child's video simply shows the robot a thumbs-down. The gesture activates a privacy overlay on their face — no account, no extra device, no detour.

Spatial no-go zones. The robot continuously monitors its position in the room. Sensitive areas — such as the teacher's desk or personal workspaces — are defined as protected zones. As the robot approaches, movement is automatically restricted.

Transparency about additional observers. If another person enters the child's camera frame, the child receives a visual notification. Audio level indicators show whether the current conversation is happening in a 1-on-1, group, or class-wide context.

The publication — PRIVATAR: Privacy-friendly Mobile Avatars for Sick School Children — appeared as a demo paper at Mensch und Computer 2026 (30 August – 2 September, Duisburg). DOI: 10.18420/muc2026-mci-demo-277