An overly long grace keeps the system live after the user has abandoned the action

What it is

The user has already dropped both hands, turned to pick up a cup, or walked away, and in their own judgment the action is over. If the re-entry grace is still running, the next motion that enters the frame is treated as continuation by the same person who never left: responsive after abandon. An overlong window mistakes “already finished” for “still engaged,” so idle motion, someone else’s reaching hand, or smoothing a sleeve lands on a command channel that is still alive.

Why it happens

The grace timer only sees how long tracking has been down, not whether the person still means to operate. Intent to stop is usually faster than the sensor’s re-lock: arms can drop and speech can start within 300 ms, while a depth camera may need close to a second to lock again after a torso yaw. If the window is set to two or three seconds for “worst occlusion,” most deliberate abandons fall inside it. Stale engagement also inherits focus, the selected object, and uncommitted values, so a false trigger does not start from a blank desktop—it continues the previous intermediate state. This is not Midas touch from an enter condition that is too loose. The person was legitimately engaged; the error is a delayed exit.

Studying it

After a one-shot action, let participants leave or drop their hands at their own pace. The experimenter marks “I consider myself stopped”; the log marks when the system clears engagement. The lag of the system behind subjective end is the cost of overlong grace. During that lag, insert unrelated motion (a goodbye wave, catching a passed cup) and count how many become commands. Sweep grace from 100 ms to several seconds and plot recovery-from-occlusion success against post-abandon false triggers; do not report a single “recommended duration.”

Where it stops holding

In browse tasks where people linger, look, and think, subjective “abandon” is itself fuzzy, so overlong grace hurts less and short grace hurts more. Safety-related continuous controls (a hoist, table height in surgery) cannot afford stale engagement even in a browse setting. In a seated cockpit or workstation people rarely actually leave; a few seconds of grace mainly false-fires on conversational gestures, not on walk-away. Public exhibits have fast turnover; overlong grace admits the next visitor into the previous visitor’s session.

Applying it

• Cap grace below the measured time from a deliberate arm-drop to the next ordinary action, rather than above the longest occlusion; recover from occlusion by other means.
• If during grace the body yaws away from the screen or both arms hang past a relaxed pose, allow the window to end early.
• Walk “finish and leave”: every second the system still highlights engagement after the user has turned away in the recording counts as a defect.

Related

• Same group: C4.19.1 A re-entry grace period can keep engagement after brief tracking loss · C4.19.3 An overly short grace treats brief occlusion as deliberate disengagement · C4.19.4 Whether re-entry requires repeating the enter action depends on whether the grace has expired
• Adjacent: C4.02 The Midas touch problem · C4.16 Disengagement and timeout
• Search: overlong grace · stale engagement · false continuation