It's 9:40 at night and you're typing into a translation app: Please make sure the guest bath is really guest-ready this time, and don't forget the thing with the shower. You hit translate, paste it into a text, and send it to the woman who has cleaned your place for eight months. She replies with a thumbs-up. Two days later the shower door still has the same faint haze along the bottom track, and you feel something sour that's hard to say out loud. Is she careless? Is she ignoring you? Or does she think you're too picky to bother with?
Usually it's none of those. Your message failed before it was translated. The language barrier didn't create the problem. It just exposed one that was already in the sentence.
The tapping experiment that explains your shower door
In 1990, a Stanford graduate student named Elizabeth Newton ran a simple study. One person, the "tapper," picked a well-known song like "Happy Birthday" and tapped its rhythm on a table. A second person, the "listener," tried to name the song.
Before the listeners guessed, tappers predicted they'd get it right about half the time. Across 120 songs, listeners named three.
The gap comes from the tapper's head. While they tap, they hear the melody playing in their mind. The rhythm sounds obvious to them because the song fills in everything the taps leave out. The listener gets knocking on a table and nothing else.
Economists Colin Camerer, George Loewenstein, and Martin Weber had already named this pattern in 1989: the curse of knowledge. Once you know something, it's very hard to imagine not knowing it. You can't un-hear the song.
"The thing with the shower" is a tap on a table. In your head it plays the whole melody: the guest review from March that mentioned soap scum, the particular corner of the track, the white-vinegar spray under the sink, how it looks when it's right. Your cleaner receives five words. A translation app can turn those five words into fluent Spanish or Portuguese or Tagalog, and they're still just taps.
Why a second language makes it worse, and also makes it easier to see
Two people who share a first language can cover the curse of knowledge with a pile of shared context. Idioms, tone, and quick back-and-forth questions do a lot of quiet repair work. When a native English-speaking cleaner gets "the thing with the shower," they may text back the track or the glass? and the gap closes.
Across a language barrier, that repair channel narrows. Asking a clarifying question in a language you're less confident in costs something socially. Many people in service work, especially when the client holds the power, will choose a polite thumbs-up over a question that might make them look incompetent. Psychologists call a related bias the illusion of transparency: Thomas Gilovich, Kenneth Savitsky, and Victoria Husted Medvec showed in 1998 that people overestimate how much of their inner state others can read. You feel your frustration so clearly that you assume it comes through in the text. It doesn't.
So you end up with a host who thinks they were clear, a cleaner who thinks they complied, and a shower door neither of them is looking at the same way.
The upside is that a language barrier strips away the false comfort. With a fluent cleaner you could go years without noticing your instructions are vague. With a cleaner who reads your texts through a translation app, the vagueness shows up quickly. Fix it for them and you've fixed it for everyone.
The fix: stop describing and start showing
The answer to the curse of knowledge isn't more words. You can't hand over the song you're hearing, but you can hand over what produces it.
In cleaning, what produces it is a picture of the finished result. Cognitive psychologist Allan Paivio's dual coding theory proposed that information stored both as an image and as words is easier to remember than information stored only as words. The picture superiority effect, a well-replicated finding that images are recalled better than their verbal labels, points the same way. For your purposes the practical point is simple: a photo of a clean shower track doesn't need translating. It's the same in every language.
Think about the difference:
- Description: "Make sure the bathroom is spotless." (Spotless by whose standard? Which spots?)
- Specification: "Clean the bottom track of the shower door." (Better. But how clean is clean?)
- Demonstration: A photo of the track, close up, after you've cleaned it yourself, labeled with one word: así ("like this").
The third one gets past the curse of knowledge because it doesn't depend on her rebuilding the image in your head. You've put the image in her hand.
Write for a translation app, not a person
You'll still need words, so write them the way translation software handles best. That means short sentences, one action per sentence, concrete nouns, no idioms, and no sarcasm or softening that depends on tone.
"Could you maybe give the oven a little extra love when you get a chance?" can come out of a translation app sounding confusing, oddly intimate, or like it's optional. "Clean inside the oven. Every turnover." translates cleanly and leaves no doubt that it matters.
This can feel rude to a host who was raised to wrap every request in cushioning. It isn't. Plain language that respects someone's time is a courtesy. And the warmth you'd normally put into phrasing can go elsewhere: a thank-you in her language, a photo of a five-star review, paying on time without being asked.
Close the loop with pictures going both ways
The curse of knowledge works in both directions. Your cleaner also knows things you can't see. She knows the vacuum is losing suction, the guests moved the dresser, and the hallway bulb is flickering. If reporting those things means writing paragraphs in her second language, mostly she won't.
When photos go both ways, you get a shared vocabulary. You send the reference photo of how the bed should look. She sends a photo of how it looks today. If they match, you're done and no one had to write a sentence. If they don't, you can point at the exact difference, this pillow, this corner, without either of you hunting for words.
That's also a relief on a human level. Most of the sourness hosts feel toward a cleaner they "can't get through to" isn't about cleaning. It's the loneliness of not being understood, and her matching worry of never being sure she got it right. A shared photo standard takes much of that guessing out of the relationship.
Your next moves
- Run the tap test on your current instructions tonight. Read each line of your cleaning notes and ask: if I'd never seen this property, would I know exactly what to do and what done looks like? Circle every line that relies on something only you know, like "the usual," "the thing," "guest-ready," or "spotless."
- Photograph your five problem spots after you clean them yourself. Go to the property, clean the places that keep getting missed (shower track, under the bed, oven, the top of the fridge, the inside of the microwave door), and take a close-up of each at its best. These become your reference images.
- Rewrite one instruction set in translation-ready English. One action per line, concrete nouns, no idioms. Run it through a translation app, then translate it back to English. If the round trip comes back garbled, simplify again.
- Send the reference photos with one word each, not a paragraph. "Like this," in her language, and nothing more. Then ask her to send a matching photo of the same spot at the end of her next turnover.
- Learn three phrases in her language this week: thank you, it looks perfect, and any problems? The third one matters most. It makes reporting issues an expected part of the routine instead of something she has to bring up herself.
Where Stayput fits
This is the loop Stayput was built to make routine. Each property gets its own cleaner texts by SMS, so there's no app for your cleaner to learn, and every turnover ends with photo confirmation of the spots you care about, checked against what you expect. Restock alerts come back the same way, so she can tell you you're low on dish soap with a tap instead of a paragraph. It won't learn the language for you, but it means you and your cleaner are looking at the same picture every time. If that sounds like the missing piece, you can see how it works at stayput.lumenlabs.works.