Leak from Otherworld

Augmented reality is often described as a technology that overlays digital information onto the physical world. This description captures the wearer's experience, but not the social situation around them. When a person follows invisible navigation cues through a crowd, records through smart glasses, reacts to a virtual object on a sidewalk, or leaves a persistent spatial annotation in a public place, the experience does not remain sealed inside the device. It changes posture, movement, attention, speech, audio, display behavior, spatial organization, and social interpretation. Augmented reality is never fully private. It leaks.

We use leaking to describe any mechanism by which AR content, state, sensing, presence, or spatial infrastructure becomes perceptible, inferable, or consequential beyond the primary user's intended experience. Some leaks are intentional and designed: projection-based AR, companion spectator views, QR markers, public displays, face displays, and shared spatial audio can make parts of an experience available to others. Other leaks are incidental: open-ear audio bleed, distinctive gestures, visible headset form factors, bystander anxiety around cameras, or crowds gathering around location-based game events. A third class is infrastructural and often invisible: shared world models, persistent spatial anchors, proximity discovery, AR cloud registries, and cross-platform spatial content discovery can make one person's augmented layer available to another person's device even when unaided bystanders see nothing.

This leakage matters because AR is moving from isolated demonstrations and single-user headsets toward public, persistent, and multi-wearer environments. Earlier bystander problems often assumed a simple asymmetry: one headset wearer inhabits an augmented scene while nearby non-users observe fragments through the physical world. Emerging systems complicate this assumption. If bystanders also wear AR devices, the relevant boundary is no longer only between the augmented wearer and the unaided observer. It is also between multiple co-located augmented realities: different apps, platforms, identities, permission models, spatial maps, and persistent content layers occupying the same place. A wearer of App A may become a bystander to App B. A person may be represented in a spatial model they cannot inspect. A public square may contain invisible anchors, contested overlays, or device-to-device cues that only some participants can perceive.

HCI has strong conceptual resources for parts of this problem: spectator experience, seamful design, social translucence, contextual integrity, embodied interaction, bystander privacy, and collaborative mixed reality. Yet these literatures remain fragmented. Work on spectators often focuses on performance and public displays. Work on privacy often focuses on capture and consent. Work on cross-reality systems often focuses on collaboration between intended participants. Work on standards and AR cloud infrastructure focuses on interoperability. What is missing is an integrated account of how augmented realities cross boundaries between private perception, public space, social interpretation, and device-mediated worlds, and how those crossings might be governed.

This paper addresses that gap through two linked studies. Study 1: Mapping systematically maps the design space of AR leaking through a scoping review of 201 works from 1992-2026. The mapping identifies 17 leaking channels across four macro-categories: environmental externalization, device-mediated signaling, social-behavioral transmission, and device-to-device leaking. We analyze these channels through permission, intentionality, modality, persistence, fidelity, containability, addressability, directionality, and governance. The map shows that leaking is not a single issue of visibility. It is a multi-dimensional design space in which different forms of perceptual, social, and infrastructural crossing create different risks and opportunities.

Study 2: Futuring uses the mapped gaps as material for protocol design. Through protocol-futuring workshops with stakeholders from AR/XR research, WebXR and spatial-web standards, AR security, location-based game communities, public AR art, accessibility, and public-space governance, we examine how future AR systems might disclose, negotiate, constrain, and contest leaking. Participants use scenarios, protocol cards, and red-team prompts to imagine future mechanisms such as reality notices, bystander preference tokens, spatial anchor leases, join/observe/mute negotiation, externalization budgets, and redress receipts. This second study turns the map into a design research agenda for crossing-reality protocols.

Together, the studies ask: when two or more realities share a sidewalk, classroom, gallery, transit station, or game event, what should each person, device, app, venue, and community be allowed to reveal, hide, contest, and remember? The paper makes four contributions to CHI:

1. It introduces leaking as a first-class HCI concept for AR: the crossing of content, state, sensing, presence, or spatial infrastructure beyond a primary user's private perceptual frame.
2. It provides a systematic mapping study of AR leaking mechanisms, identifying 17 channels across environmental, device-mediated, social-behavioral, and device-to-device regimes.
3. It reports a protocol-futuring study that develops and stress-tests candidate governance primitives for future multi-wearer and public AR systems.
4. It contributes design implications for building AR systems that coexist in shared social space: systems that make perceptual boundaries legible, support quiet refusal, govern persistent anchors, protect bystanders without devices, and provide practical redress.