The Translation App and the Village Clinic
A rural health clinic in Sri Lanka has begun using a free translation app to communicate with patients who speak a regional dialect. The app was trained mostly on urban Sinhala and English text. Staff report three problems. First, older patients' descriptions of pain are often translated as vague phrases such as 'feeling unwell'. Second, the app renders some respectful address forms as blunt imperatives, which patients find rude. Third, when the connection drops, staff fall back on a single bilingual aide who is present only three days a week. The clinic sees about 60 patients a day. The district health office wants to expand the app to two more clinics next month. A nurse argues the app saves time; a senior doctor argues it is unsafe.
1. Analyse the risks the app creates, distinguishing language problems from technology problems.
Read the worked answer
The risks fall into two categories. Language risks: the app's training data is skewed toward urban Sinhala, so it may not recognise regional vocabulary or older speakers' ways of describing symptoms, producing vague translations that could mislead diagnosis. It also flattens register, turning respectful forms into blunt commands, which damages trust and may cause patients to withhold information. Technology risks: dependence on connectivity means the tool fails unpredictably, and the fallback is a single aide available only three days a week, so safety varies by day. These are distinct: better training data would fix the first, while offline capability and staffing would fix the second. A strong answer notes that the two interact: when the connection drops, the aide is overloaded, and when the app mistranslates, staff may not notice because they cannot check the dialect. The answer should also weigh the nurse's point fairly: time saved is real, but speed is not the same as accuracy. Expanding to two more clinics before fixing either problem multiplies the risk.
2. Recommend conditions the district health office should set before expansion, and justify each.
Read the worked answer
Three conditions are reasonable. First, require offline operation or a printed phrasebook for common symptom descriptions before expansion, because connectivity failures are predictable and the aide cannot cover every day. Second, run a validation period in which bilingual staff compare app translations with human interpretation for a sample of consultations, and pause expansion if serious errors appear; this tests accuracy against the actual patient population rather than the app's marketing claims. Third, protect the interpreter role: the aide's hours should not be cut on the assumption that the app replaces her, and respectful address forms should be reviewed by native speakers of the regional dialect. Each condition targets a specific risk identified earlier: offline capability addresses connectivity, validation addresses mistranslation, and staffing plus register review addresses trust and safety. The answer should acknowledge cost and staffing limits and explain why expansion should be staged rather than refused outright.