A symptom can be easiest to describe in Twi, especially when English makes it sound flatter or less exact. A private conversational AI should let you use the words that come naturally, then clearly ask what may reach the cloud before you share something personal.
At 7:18 on a rainy morning in Kumasi, Abena stood in her kitchen holding a mug she had stopped drinking from. Her daughter’s school bag was open on the chair, one sock missing. Abena had felt a twisting pain low in her stomach since the night before, but “stomach pain” did not feel like the right description.
She typed, “Me yam mu retwa me, and it comes and goes.”
Then she paused.
The question was personal. If the pain became worse, she might need to speak to a clinician quickly. If it passed, she still did not want a health question drifting into a service without knowing where it went or how long it stayed there. The bad outcome sat in front of her phone: either she delayed seeking proper care because she could not explain herself clearly, or she shared more sensitive information than she intended.
The clearest words often arrive first
Code-switching is practical. A person may begin in English, reach for Twi when a feeling becomes hard to pin down, then return to English when describing timing, medication, or a plan for the day.
That is how people talk with family, friends, and trusted professionals. Requiring one language per turn creates an extra task at the exact moment someone is trying to find the right words.
A phrase such as “me yam mu retwa me” carries a shape and urgency that a speaker may struggle to recreate in formal English. The point is not to turn everyday conversation into a medical diagnosis. It is to give the person room to describe what they feel without translating themselves into vagueness first.
For health concerns, an AI conversation should never replace professional medical care. Severe, sudden, or worsening symptoms deserve urgent attention from an appropriate clinician or emergency service. But clear language can help someone organise what they are experiencing, prepare questions, or explain the issue to a family member before they make that call.
A private question deserves a visible choice
Abena did not need a long privacy policy while her daughter looked for the missing sock. She needed a plain choice before she continued: never send this to the cloud, ask me each time, or allow it for this session.
That moment matters because consent is part of the conversation. It should not be buried behind an account setting someone set months ago and forgot about.
Nkomo makes that choice explicit. You can talk or type in natural Twi and Ghanaian English, while deciding whether a specific conversation may reach the cloud. Its history stays on your device under your control. Turn history off, and it is purged immediately. You can also export your data or delete your account in one tap.
Ghana’s recent rejection of a proposed US health agreement over inadequate safeguards for sensitive health data is a reminder of the wider issue. Health information carries consequences beyond the moment it is typed. Asking where it goes, who can access it, and how long it remains available is ordinary caution.
For a fuller explanation of these choices, read a plain-English guide to cloud consent, on-device history and deletion in conversational AI.
A pause can be part of care
Abena chose to be asked before anything went to the cloud. She used her own words, added when the pain had started, and drafted a short explanation for a clinician. Then she put the phone down, found the sock under the table, and called for medical advice when the pain did not settle.
The useful part was small: she did not have to choose between speaking naturally and knowing what happened to her question.
That is a better standard for conversational technology, especially around personal subjects. The language should fit the person speaking. The privacy choice should be clear before the details leave their hands. And if something goes wrong, the app should show the error rather than quietly pretending the message arrived.
Keep the words, keep the choice
Personal questions are rarely tidy. They can begin with a symptom, move into a family concern, and end with a request for help finding the right next step. A person should be able to say “me yam mu retwa me” without first converting their experience into language that feels less true.
They should also be able to stop, look at the consent choice, and decide for themselves.
Comments
No comments yet.