Recognition gaps become de facto exclusion

What it is

A hotel room folds lights, HVAC, and curtains into a voice-primary path. The desk sells “just speak.” A guest in their seventies says “make it cooler” three times; the system hears something else or nothing, and they hunt for a switch the room no longer has. The interface is specified as open to everyone. Completion is open to the group the recognizer prefers. That is de facto exclusion by recognition gap: the feature was not switched off; the task cannot be finished; it was not given. Children in a classroom that only accepts spoken roll-call meet the same door.

Why it happens

When voice is the only path, or the advertised default, word error rate stops being a decimal on a quality sheet and multiplies into “can this attempt finish.” An 8% sentence-error rate for younger adults can still be repaired in a turn or two. The same skill at 30–40% for older adults or children burns the dialogue in timeouts and recovery, and people leave. With no second channel, leaving is exclusion. Ads and unboxing film an adult getting it right once; exclusion is invisible in the story. Excluded people never appear in success logs, so growth still looks healthy. The gap moves from a measurement issue to a who-counts-as-a-user issue.

Studying it

Change the dependent measure to task completion by cohort, not WER. Same goal (drop the room two degrees, say “here” at roll-call), child / older / younger-adult groups on the voice-primary path. Log finish, abandon, and whether a hand went looking for a non-voice control. Independent variable: whether a peer touch or panel exists. Hotel rooms and classrooms are the right fields: the former often hide the panel, the latter often make speaking the only legitimate act. Ask “do you think this feature exists” — people who cannot finish will say it is broken or not for them. Both answers are exclusion, not low satisfaction.

Where it stops holding

A screen or a wall control that stays reachable makes voice failure merely slow; exclusion does not hold. Skills whose tail completion has been aligned with younger adults have turned de facto exclusion off; stratified WER may still exist. Treating “older people dislike new tech” as the cause of exclusion recodes recognition-driven exits as attitude. Expert users (clinicians on a long-used voice charting system) are not in this consumer room/classroom story. Exclusion is about the main path, not a demo in which an adult speaks the child’s line.

Applying it

• In scenes where a child or an older adult is the owner, voice must not be the only path. Hotel rooms keep reachable light and HVAC controls; roll-call keeps a gesture, a roster, or touch.
• Before copy says “just speak,” look at tail completion, not the adult demo film.
• When recovery is exhausted, land on a non-voice exit. Do not spend a third turn on “please try again.”
• How to check: in a real room or classroom, the target age band finishes the main task alone; observers do not speak for them. If they cannot finish and cannot find another control, treat it as exclusion. Add a channel; do not add a slower prompt.

Related

• Same group: M4.04.1 Word error rate is substantially worse for children and older adults · M4.04.3 Evaluate by cohort, not by the overall number
• Nearby: M4.09 Recognition differences for children and older adults · C7.06 Hands-free and eyes-free use · M1.06 Repair strategies
• Search terms: de facto exclusion by recognition gap · cohort completion · age-stratified word error rate