Gemini Chatbot App
Team: Joseph Lico, Miles Macklin, Jenna Scheen
Challenge: Design a Conversational Agent (CA) for health which will help mitigate the impact of the COVID-19 pandemic on a specific population or community of your choice. 
Our Solution: Gemini is a dual-personality, mobile chatbot app which analyzes users' written, audio, and visual journaling input to foster anonymous connections between BIPOC young adults
We want to learn about the disproportionate impact of COVID-19 on BIPOC adolescents’ mental health through conducting interviews and collecting survey data so we may alleviate emotional strain by simulating companionship through Artificial Intelligence.
We conducted 6 interviews with Asian, African-American, Indian, White and Latino participants within the age range of 17-21, as well as a mental health advisor from the University of Michigan Counseling and Psychological Services (CAPS). To gauge our target population's satisfaction with current mental health aides and products marketed towards quarantine use, we also conducted a 12-question survey garnering 35 responses.
After reviewing key findings from our survey data and narrowing down the primary concerns of each interviewee, our team  used affinity mapping to organize the interview data. We then categorized the interviewees' painpoints into 9 target areas:
1. Stressors of generational divide
2. Routine Changes
3. Mindfulness towards mental health
4. Impact of transition to online school
5. Efforts to cope
6. Accessibility Issues
7. Effects of increased screen time
8. Social media/social justice stressors
9. Unique needs of BIPOC seeking help
Following the data analysis, our team used virtual post-its on Mural to generate chatbot design solutions. Our brainstorming brought us closer to key features highlighted in the final design such as the spontaneity of creating anonymous connections between BIPOC with similar experiences, and locating resources for BIPOC-related mental health issues.
Affinity Diagrams 
Potential Chatbot features 
Following our 2 Concept Validation sessions, our team conducted 2 Think-Aloud Testing sessions for feedback on rough wireframes and our preliminary chatbot persona design. While we noted down opportunities for improving the chatbot's direct messaging and journaling features, the chatbot's frequent usage of emojis was well-received by our testers.
After attempting to visualize the chatbot features through rough wireframes and sample chatbot-user conversation, one question still hung in the air. 
How can we design a chatbot personality humanlike and complex enough to retain the attention of our target demographic while appealing to different users' needs and personas?
CHatbot Personality #1: Castor: Light Mode
Named after the second-brightest object in the Gemini constellation, Castor is the extroverted and blunt chatbot personality whom the user turns to for comic relief. Castor sends memes and various media connected through third-party apps in alignment with the user's personality, while cracking self-aware jokes about being artificial intelligence (AI) and having no free will. 
Chatbot Personality #2: Pollux: Dark Mode
Named after the son of Zeus and his corresponding star in the constellation Gemini, Pollux is the more open, agreeable, and nurturing chatbot personality whom the user turns to for gentle reassurance, support, and assistance in connecting to other BIPOC users seeking relief. 
Back to Top