Interaction capability spreads across many devices instead of one place

What it is

The immediate consequence of ubiquitous computing: controlling an environment no longer runs through one terminal but is spread across multiple entry points in the space — the wall switch, the desk speaker, the phone in the pocket, the sensor on the door each carrying a slice of interaction capability. No device is "the remote"; each is just one patch of the environment's interaction surface.

This inverts the one-person-one-computer structure: in the PC era interaction lived on the box on the desk, and users knew where to find it. In the distributed structure, "where do I go to do it" is itself the question — the same action (turn off the light) may exist at three entry points, each covering a different scope.

Why it happens

Distribution is not a design failure but the resultant of three forces:

• Physical optimality. Entry points closest to the point of use save the most effort — bedside controls the bedside lamp, the entryway controls the house. Proximity to movement paths is a spatial efficiency that necessarily multiplies entry points.
• Capability specialisation. Voice suits status queries, phones suit complex configuration, physical switches suit eyes-free operation; each occupies the scene where it is strongest. There is no universal entry point — only specialised ones, each holding a corner.
• Heterogeneous acquisition. Devices are bought year by year, each shipping its own entry point (its own button, its own app); nobody plans the whole. The entry-point layout is a fossil record of purchase history.

The structural result is a fragmented coverage matrix: entry points × functions form a sparse matrix where each point covers a patch, with overlaps and holes between patches — some functions reachable three ways, some unreachable, some only via detours (open the app, drill three levels). The user's mental map of "what is doable where" is maintained by trial and memory, and the matrix keeps deforming with every new purchase.

Studying it

• Entry-point inventories: enumerate every interaction entry point in a home and its reachable functions, build the entry-function matrix, and measure overlap and hole rates. Multi-device environment research — Dearman and Pierce's "It's on my other computer!" work is representative — robustly finds users frequently mistaken about which device holds which capability, a burden distribution sharply amplifies.
• Entry-point choice experiments: offer the same task through multiple entry points and observe selection strategy (nearest? habitual? most recently learned?), measuring choice time and error — the most direct cost of distribution is the time spent choosing where to act.
• Longitudinal tracking: how entry habits redistribute when a new device arrives, and whether old entry points fall into disuse (function intact, entrance abandoned).

One methodological caution: entry-point choice is steeped in household movement patterns and furniture layout; stripping the spatial variable for the lab strips out the core of the distributed structure — such studies belong in real spaces.

Where it stops holding

• The burden scales with overlap inconsistency, not entry-point count. Five points over five disjoint function sets is manageable; three points over overlapping functions that behave differently is immediately unmanageable — the primary variable is inconsistency, not number.
• Residents and visitors face different boundaries. Residents internalise the entry map over months; a visitor faces a completely unreadable environment — the same structure has sharply different usability for the two, and evaluation must separate them.
• Highly mobile tasks benefit from distribution. Actions needed everywhere (lights off, pause music) genuinely gain; low-frequency complex tasks (configuration, troubleshooting) carry pure burden. Don't average the two directions together.

Applying it

• Guarantee multi-point reach with identical behaviour for common actions: high-frequency actions (on/off, pause, volume) available at every natural point of use, with strictly identical behaviour across points — overlap is not the problem; inconsistent overlap is.
• Converge complex tasks to a single entry point: configuration, troubleshooting, and history keep exactly one full-function entry (usually the phone app); other entry points link to it as shortcuts, never maintaining parallel simplified copies.
• Plan entry points when a device arrives: state explicitly how its entry overlaps existing ones, and where they duplicate, name primary and shortcut — don't let purchase history decide the structure by default.
• How to check: have a first-time visitor complete three basic tasks (lights, temperature, music) and count the wrong entry points attempted. Wrong-entrance attempts are a direct readout of the structure's confusion.

Related

• Same group: Z1.07.2 Distributed interaction needs consistent entry points to stay manageable · Z1.07.3 Losing the aggregation point costs control of the whole system · Z1.07.4 Over-aggregation into a single entry point builds single-point-of-failure risk
• Nearby: Z4.10 Voice, app, and physical switches coexisting · Z1.06 Device ecosystems and coordination
• Search terms: distributed interaction · multi-device ecosystems · interaction surfaces · entry point selection