WritingEssay
Whose words were those
Why the most important question in a Bangladeshi consultation is who is speaking, and what it costs a doctor not to know.
No. 26 · June 2026 · 4 min read
In a Bangladeshi consultation, the patient is often not the one talking. A son speaks for his elderly mother. A husband relays his wife's complaint. A relative translates a dialect the doctor does not share. By the time a symptom reaches the doctor, it may have passed through one or two other people, each of whom added, dropped, or reframed something without meaning to.
This is not noise to filter out. The source of a symptom is part of the symptom. My mother has chest pain is a different clinical fact from my mother clutched her chest and would not say why, and a doctor who cannot tell which one he is hearing is reasoning on a story he cannot fully trust. In the accompanied visit, who said what is clinical information, and losing it is losing data.
My mother has chest pain is a different clinical fact from my mother clutched her chest and would not say why.
So Glyph treats the speaker as a field, not an assumption. The intake resolves who is holding the device, and tags every claim with its origin: patient-reported, attendant-reported, attendant-translated, attendant-observed. When the patient's own account and the attendant's diverge, the divergence is kept and shown to the doctor, not smoothed into a single tidy narrative that hides the seam.
A clinical AI built for a Western solo encounter has no place to put this, because it assumes the voice it hears is the patient's. That assumption is wrong here often enough to be dangerous, and getting it right is not a translation feature. It is a different model of what a consultation is: not one person describing herself, but a small group assembling an account, in which the doctor's job includes knowing who contributed which part. Glyph is built for that consultation, because that is the one that actually happens.