Designing an Artificially Intelligent Agent for Medicine On-Demand in Nigeria

My Bentley University course on "Designing Interfaces for Artificial Intelligence" worked on a Harvard School of Public Health (HSPH) project that aims to bring on-demand healthcare to Nigeria. The first half of the semester was spent determining how artificial intelligence could be used to assist this venture. After many discussions and revisions with the HSPH team of doctors, we determined that the most useful intelligent agent is one that filters requests in emergency or non-emergency categories, with a human triage nurse on hand to make judgement calls on unclear cases The system will route primary care doctors to non-emergency requests. The user workflow and complete system architecture are shown below.

The following wireframes are intentionally low-fidelity in order to show what is capable and which functions matter, but allows the HSPH team to continue working on the venture beyond the semester with a dedicated software development team.


The app opens on a screen that immediately offers the value proposition of sending a doctor to the user's location. Inspired by Uber, the home screen lets users choose their location and call their service provider with a click of one button. Separating from the Uber model, the interface asks users if they have any emergency symptoms, split into illness or emergency symptoms. Since the value proposition is to provide medical services on-demand, the app also lets users speak with a doctor by phone if that suits their needs.


If a user selects any of they symptoms that may be an emergency indicator, the service lets users call a triage nurse to determine if the symptom they selected is a true emergency indicator or if the patient can be seen by a primary care doctor. The screen also shows the user the location, address, and phone number of the nearest hospital, and provides a single click-to-call button. (Nigeria does not have a 911-type service that would call an ambulance to the user's location, so they must find their own way to the nearest hospital)


If a user selects no emergency indicators, then they get a confirmation screen which shows the name and photo of the primary care doctor. The interface provides options to cancel the visit and view the doctor's travel status. Many of the patients using this service are wealthy and live in secured neighborhoods that could require passing through security, so the interface provides a field to enter this necessary information. Tapping "being describing symptoms" takes users to the screens shown below. 


Nigerians are hesitant to provide any personal information to a software service, so the app only asks for the basics to call a doctor and increase the chances of a successful use case. After confirming the doctor, however, the app attempts to learn more about the patient. This information is not necessary since the doctor and take it during the visit, but if the app can get the information from the patient, then the doctor's visit can be more efficient, as well as providing a heads-up to the doctor. The app also begins creating a medical history on the patient, since it is highly unlikely the service will have one to begin with. Only after describing their medical history does the app say "submit" and stop asking for information. After this the app progresses to the screen that shows the doctor's driving status.


The full prototype includes several more screens. If you are interested in seeing those, I am happy to walk through these in person or over email.