Designing a Chatbot

Project: Project for course Human Factors in AI (Coursera)

Role: Product and UX/UI Designer

Overview: Design a prototype utilizing machine learning and artificial intelligence.

Tools: Figma

Deliverable: Presentation that outlines the ethical and privacy implications of machine learning based product. Include task analysis, overview of user experience design, privacy and ethical considerations. Also include prototype with screenshots of model that utilizes machine learning and artificial intelligence.

Medical Diagnosis Chatbot

For this project I decided to design a medical diagnosis chatbot.

Problem

Help patients identify their symptoms to determine if necessary to visit doctor

Overview of chatbot

Uses machine learning algorithm with natural language processing. Once user signs in to app with name and medical insurance if available, user can interact with chatbot.

Key task

Provide preliminary medical diagnosis based on input of symptoms and provide option to schedule a doctor’s appointment

Task Flow

Ask patient to select consultation or appointment?

a. If select consultation, ask patient to describe symptoms.

b. Provide possible preliminary diagnosis.

c. Ask how long symptoms present?

d. If symptoms present more than a week, recommend to schedule doctor visit.

e. Ask patient if would like to schedule doctor visit?

f. If yes, provide available appointment times for that

week.

2. If select appointment, provide available days and times that week. If none available,

offer following week availability.

User Experience Design

Key considerations of AI product user experience design:

User inputs (and any feedback loops)

Select consultation or appointment
Patient’s symptoms
How long symptoms have been present
Select appointment time

2. Transparency on the model

Uses Machine Learning algorithm with Natural Language Processing
Accesses database of symptoms to determine preliminary diagnosis

3. Communicating uncertainty

Provides only preliminary diagnosis with recommendation to visit a doctor.

Privacy Considerations

Protected health information
- only shared with system to determine preliminary diagnosis
- only shared with your healthcare provider when making an appointment
HIPAA-compliant

Ethical Considerations

Fairness
- Database of symptoms and diagnoses should be taken from wide range of ages, race, gender, and locations
- System should allow ability to accept wide range of insurance, including Medicaid or still provide guidance for those without medical insurance
- Chatbot is accessible tool that can be accessed on the web, which can be used easily by those with limited mobility or elderly

Accountability
- System will be routinely reviewed by doctors who treat a variety of patients, including children and adults, from a variety of locations to ensure accuracy of diagnoses

Transparency
- Database built from Mayo clinic, found to be a trustworthy and available source of the information
- Algorithm used is machine learning with natural language processing