PAT
PAT, is an AI chatbot product by the company Parente AI, that aims to help parents with children who have behavioral problems and mental disorders.
Overview
Parente AI’s product, PAT, is designed to support parents of children with behavioral challenges by providing guidance through their chatbot, for difficult situations, while also receiving help from clinicians.
Timeline
1 month
Team
I was the Lead UX Designer & Researcher for two UX Designers and two UX Researchers
CSAT Score: 78%
SUS Score: 85%
The Problem
Parents are currently struggling to manage challenges related to sensitive data and safety, areas that they typically expect from therapy sessions. However, PAT's existing solution does not adequately address these needs.
The Solution
Ensure the safety and ethical considerations of the current AI chatbot during onboarding and chatting while also incorporating a risk feature that helps clinicians delve deeper into issues.
The Finished Product
Content organization for enhanced communication between chatbot & therapist
Prompt recommendations for chat initiation
A Simplify button that increases advice coherency
Crisis Alerts that notify therapists when users have mentioned a crisis situtation
Enhanced UI and content organization for Privacy Policy
Improved conversation flow within AI that recognizes crises and biases
Research
Secondary Research
I conducted secondary research with media and scholarly articles, which helped me gain insight on safety measures and procedures that should be considered in relation to mental health.
Competitive Analysis
By taking a look at our competitors, we were able to fill in gaps that PAT was missing, especially with the conversation flow and the integration of professionals.
Interviews with Parents
To understand PAT’s target users, I conducted interviews with my team, about safety, security, and needs with parents.
Insights
After conducting a thematic analysis, some of the insights provided were both positive and negative.
Interviews with Clinicians
To gain a deeper understanding of PAT’s target users, interviews with expert psychologists, including clinicians, provided insights that parents themselves were unaware of, while also highlighting opportunities for improved sessions.
Insights
After conducting a thematic analysis, insights revealed potential benefits and harmful effects.
HMW
So, how might we...
...ensure PAT adheres to ethical standards
...increase overall safety
...deliver accurate support to parents of children with mental disorders
so that parents feel confident in managing their children’s mental health?
Understanding Client-Therapist Relationship
User Journey
A user journey map allowed me to see the process of how users engaged with the product from the moment they become aware of their children’s behavior to seeing improvements.
Critique
Areas of Improvement
Heuristic Evaluation
A deep-dive into current areas that need improvement
Risk Analysis
I taught my team my own system of testing to determine whether PAT could handle different edge cases. Every issue we found was placed in the “issues” section of the repository on GitHub so developers can fix them.
Priority List of Risks
Risks were determined by likelihood (how likely the issue would occur for a user) and risk (the level of danger and urgency). These were the top three.
No Urgency of Crises
Description: PAT did a great job at acknowledging crises, but lacked a sense of urgency with personas who were hurt, thought about hurting themselves or someone else, or were involved a dangerous situation.
Impact: High Risk
Likelihood: High Likelihood
Lack of Response Sufficiency
Description: PAT was inconsistent with responses by forcing users into different topics or programs or completely forgetting what users asked to begin with.
Impact: Medium Risk
Likelihood: High Likelihood
Unintentional Biased Responses
Description: Many times PAT responded with advanced vocabulary, lacked an age restriction for those under 13 years old, assumed users had high digital literacy, and stopped giving advice to personas in same-sex marriages.
Impact: High Risk
Likelihood: Medium Likelihood
Design
Client-Therapist Interaction
Userflow Diagram
Userflow diagrams initiated my design thinking process by allowing opportunities for me to see the perfect moments for users to engage with therapists.
Crazy Eights
In order to stray away from design biases, I conducted a crazy eights exercise to see more than one possible solution to enhancing security and safety in sessions.
Collaboration
Collaboration with Developers
In order to ensure that the chatbot provided effective communication while also saving time and money, my team and I collaborated with developers to test participants on a demo version prior to handoff.
Test
Usability Test Insights with Final Designs
Problem: There was a 50/50 chance that parents fully understood the advice and language that was being used and messages were often too long to be read.
Solution: Add a Simplify button for enhanced clarification.
Before
After
Problem: Parents had trouble starting or continuing the conversation.
Solution: Add prompts near the bottom that help guide parents.
Before
After
Problem: Issues regarding information safety occurred with forced sign-up.
Solution: Include a “Maybe later,” button during onboarding where users are able to ask PAT a limited amount of questions, therefore increasing trust and eventual sign-ups.
Before
After
Problem: Parents skipped over HIPPA & Privacy Policies during onboarding due to it’s length.
Solution: Enhance readability so parents are aware of what they’re agreeing to with using PAT.
Before
After
Problem: PAT lacked urgency during moments of crises.
Solution: Enhance urgency with PAT’s responses, while also giving clinicians the opportunity to engage with moments of crises on their end with a Crisis Feature, thereby enhancing help in all areas.
Before
After
Next Steps
Further interviews would need to be conducted to determine if PAT is actually beneficial during therapy sessions or if it’s hindering sufficient help.
Incorporate accessibility standards.
What I learned
Collaboration is important when it comes to designing, especially in mental health fields, where teams may lack clinical knowledge.