The division of labor must be predictable

What it is

People pick up a tablet expecting to retouch with a stylus, sit at a computer expecting to align a timeline, and raise a phone expecting it to be the viewfinder. Predictable division means each device’s role in the job can be guessed before touching it, and stays still across sessions. Continuity Camera makes the iPhone a camera for the Mac with a clear split: the computer wants a picture, the phone takes it. If the same product sometimes offers a full editor on the tablet and sometimes hides the timeline down to preview-only, nobody can decide which device to pick up. Predictability is about stable roles, not how the capability list is sliced, and not how a missing feature is explained.

Why it happens

Cross-device use runs on a role schema: the living-room screen is for watching, the pocket device is for control and capture, the desk is for finishing. The schema comes from physical cues (how large, whether there is a stylus, whether it sits on a tripod) and from the last few successes. Every break—tablet suddenly cannot draw, phone suddenly demands a full form, computer dumps the camera flow onto a screen nobody prepared—forces a replan. If roles drift with plan tiers, experiments, or a silent “this hardware is too slow,” people do not learn a division. They learn that they must try the device to find out what it can do today. Once an expectation exists, violating it costs more than never dividing: the step has already been prepaid to a device, and the walk is already spent when the role has changed.

Studying it

In ecology diaries, code “what I thought this device would do before picking it up” against “what it actually did.” Mismatch is a failed role expectation. In the lab, freeze the capability slice and vary only whether the role stays stable across sessions and whether it is labeled.

Independent variables: cross-session stability, idle-screen labeling (“this one shoots,” “this one edits”), alignment between physical form and role. Dependent variables: wrong-device picks, pick-up-and-put-down switches, whether the device named while planning the next step is the one that actually handles it.

“Which device do you usually use for which step” elicits the schema, not satisfaction. Do not train the role with a brand tutorial and then measure predictability—that measures training.

Where it stops holding

A new form factor entering the same job (glasses, an in-car screen) has no schema yet and needs an explicit role statement. Experts rewrite the split (editing the whole film on a phone); predictability must allow override rather than lock the official role. When a device is borrowed for something else (computer casting in a meeting, tablet handed to a child), the role pauses and the remaining device must still finish the task. In a household, one screen may be a parent’s cinema and a child’s console; read the role from the current user, not from the room.

Applying it

• Give each signed-in device a stable role in the product and state it in one sentence on that device’s idle or home surface, not only in settings.
• Treat a role change as an explicit event (“from now on this tablet only previews; edit on the computer”), not a silent slice change from a performance probe or remote config.
• Keep form and role pointed the same way: stylus tablets take stylus steps, camera phones take capture. Do not swap them so that picking the device produces a stall.
• Verify by asking a week-old user, with the app closed, which device they would pick for the next retouch pass and the next export. Every wrong name is a role that never became a schema. Then change one device’s slice with no notice and watch for an immediate “why not today” pause on pickup.

Related

• Within the group: K8.03.1 Allocate tasks by input capability and screen size · K8.03.3 Missing capabilities need explanation, not silent hiding
• Adjacent: K4.07 Division of labor and handoff with the phone · K6.11 Division of labor among cluster, center stack, and secondary displays
• Search terms: device role · multi-device ecology · role schema