Minerva

Solving conversational anxiety through ethical AI.

PROBLEM
Over 74% of Gen Z and Millennials experience social anxiety regarding in-person interactions which contribute to the highest rate of depression and isolation of any generation.
TARGET
Design a “Conversational Guide” that will address the issue and target a 70% adoption rate within a user’s first 3 sessions.
Three smartphones displaying an app teaching communication skills with topics including open-ended questions and active listening.
Smartphone screen showing an app with a glowing wavy shape and text reading, 'Thanks for asking. My week has been going OK so far, how about yours?' and a button labeled 'This is an example of an Open-Ended Question.'
ROLE
Research
Product Design
TOOLS
Gemini
Adobe
Figma
PLATFORM
Native App
OVERVIEW

We have lost the art of conversation.

Gen Z and Millennials face many challenges in interpersonal communication that contribute to increased rates of anxiety and depression.  
Challenges

Digital Over-reliance

Gen Z has grown up with smartphones and texting as primary communication tools. This leads to a comfort with asynchronous, text-based interactions and a lack of practice with real-time, in-person conversations. 

Lack of Curiosity

Social media often encourages presenting a curated version of oneself, leading to a focus on personal narratives and broadcasting. This can translate into real-world conversations where individuals talk more about themselves and struggle to ask genuine questions about others.

Conversational Anxiety

Due to a lack of experience in face-to-face conversations, many Gen-Z individuals experience high anxiety around in-person interactions, favoring online or text-based conversations, further exacerbating the issue.

How can we use artificial intelligence to bring about real life connection?

Enter: Minerva
The goal was a "Conversational Guide" combining bite-sized modules with interactive practice. Users master core techniques like active listening and storytelling through simulated conversations. This provides real-time feedback, allowing users to build the confidence needed to translate digital fluency into impactful in-person communication.
RESEARCH

Everything starts with research

Extensive research yielded a 25 page document outlining the issues faced by many Gen-Z and Millennial individuals with regards to conversation.It uncovered a strong desire to connect but a lack of knowledge about where to start or what to do.

» 52 survey responses via Google Forms using Reddit and Slack channels

» 65 articles and research papers referenced

I distilled the research into 6 main issues

1
Over-reliance on Digital Communication & Lack of Face-to-Face Practice
Problem: Gen Z has grown up with smartphones and texting as primary communication tools. This leads to a comfort with asynchronous, text-based interactions and a lack of practice with real-time, in-person conversations. Many are more comfortable texting than talking on the phone or in person.
Impact: Decreased comfort and confidence in face-to-face settings, leading to awkwardness and difficulty in sponataneous interactions. They may struggle with thinking on their feet as they're used to having time to craft responses.
2
Difficulty Interpreting and Utilizing Non-Verbal Cues:
Problem: Online communication largely lacks non-verbal cues (facial expressions, tone of voice, body language, eye contact). This leads to a diminished ability to read and send these crucial signals in person.
Impact: Misunderstandings, difficulty gauging others' emotions, and a reduced capacity for empathy. "The Gen Z stare" is a recent TikTok trend that highlights this issue, where a lack of appropriate eye contact or expressions can be misinterpreted.
3
Shortened Attention Spans and Expectation of Immediate Gratification
Problem: The fast-paced nature of online interactions (quick replies, endless scrolling) has shortened attention spans. People expect immediate responses and can get bored during real-time conversations that require sustained focus.
Impact: Difficulty with active listening, easily distracted, and a tendency to give one-word or dismissive replies in person (e.g., "Noted," "Okay") that can come across as disengaged.
4
Self-Centeredness and Lack of Question-Asking:
Problem: Social media often encourages presenting a curated version of oneself, leading to a focus on personal narratives and broadcasting. This can translate into real-world conversations where individuals talk more about themselves and struggle to ask genuine questions about others.
Impact: Superficial conversations, difficulty building rapport and deeper connections, and perceived lack of interest in the other person.
5
Anxiety and Fear of Social Situations (especially for Dating):
Problem: For many, the controlled environment of online dating (where profiles are curated and messages can be edited) can make the transition to offline, spontaneous interactions daunting. There's anxiety around making a good "first impression" in person, managing awkward silences, or navigating potential rejection.
Impact: People might avoid in-person dates, or when they do go, they struggle to keep the conversation flowing naturally, leading to unfulfilling experiences. Ghosting, for example, can be a result of not knowing how to end a connection gracefully in person.
6
Lack of Formal Communication Training:
Problem: Many Gen Z individuals haven't had explicit training in formal communication skills, particularly in professional or more structured social settings. This can extend to basic etiquette and understanding how to adapt tone and content to different contexts.
Impact: Using overly casual language in inappropriate settings, struggling with small talk, and difficulty articulating thoughts clearly and comprehensively.

The issues were synthesized

The issues were condensed into a matrix of concerns raised by survey respondents and then processed using a MOSCOW method of “Must”, “Should”, “Could”, and “Won’t”.  It helped really distill what the core purpose of the app was and what it wasn’t.
Digital whiteboard with sticky notes discussing communication challenges, impacts, and AI-based conversational coaching solutions including clarifying non-verbal cues, ethical AI, and practice run strategies.Prioritization matrix for features of a conversational onboarding and coaching system divided into Must, Should, Could, and Won't categories with detailed notes on functionalities like real-time feedback, gamified progress, language selector, and AI-generated avatars.

Personas were created

To help shape the target audience, we created to transform abstract user data into relatable, human archetypes to help build empathy and keep the product from being designed for the “average” user.
Liam - Minerva persona 1
LIAM - 20y/o
The Socially Anxious Student
GEN-Z
A college student studying computer science. Spends a lot of online. Has a close-knit group of gaming friends but struggles with face-to-face interactions, especially in new social settings.
PAIN POINTS
  • Feels awkward and self-conscious in real-life conversations.
  • Worries about saying the "wrong thing" or running out of things to talk about.
  • Finds it hard to initiate conversations or join ongoing ones.
  • Avoids social events that require extensive small talk.
  • Feels a disconnect between his online confidence and his in-person shyness.
GOALS
  • Learn concrete techniques to start and maintain conversations.
  • Reduce social anxiety and feel more comfortable in social situations.
  • Practice conversational skills in a low-pressure, private environment.
  • Improve his networking abilities for future career opportunities.
MOTIVATION
Wants to expand his social circle, feel more confident, and participate more fully in college life and future professional settings. He especially wants to feel more comfortable going on first dates.
"I just freeze up. I know what I want to say, but it just doesn't come out right when I'm face-to-face."
Maya - Minerva persona 2
Maya- 32 y/o
The Career Climber
MILLENNIAL
A budding marketing professional working at a mid-sized company. Highly competent in her work but struggles with "water cooler" conversations, networking events, and building rapport with senior leadership. Relies heavily on email and Slack for communication.
PAIN POINTS
  • Feels her lack of strong conversational skills is hindering her professional advancement.
  • Struggles with small talk and making genuine connections outside of formal meetings.
  • Finds networking events daunting and unproductive.
  • Wants to build stronger professional relationships but doesn't know how to initiate or deepen them verbally.
  • Worries about coming across as disengaged or unapproachable.
GOALS
  • Master the art of professional small talk and active listening.
  • Learn how to build rapport and demonstrate empathy in professional settings.
  • Improve her ability to articulate ideas clearly and persuasively in spontaneous conversations.
  • Network more effectively and confidently.
MOTIVATION
Desires career progression, wants to be seen as a strong leader, and believes good communication is key to influence and success. Appreciates efficient, practical tips she can apply immediately.
"I'm great at my job, but I feel like I'm missing something when it comes to connecting with people one-on-one and it’s affecting my professional presence.”
GOALS & HYPOTHESES

Setting Goals & Hypotheses

Good design is not arbitrary – it is based on an end goal.  With that in mind I split my goals into two main categories:  Adoption & Retention.
Adoption
Achieve a 70% "Vocal Activation" rate within the first 2 minutes of the first session
HYPOTHESES
  • Using a haptic "pulse" synchronized with the AI's greeting creates a physical sense of presence that encourages a verbal response.
  • A "Listen-Only" mode intro that transitions into a question increases comfort levels compared to an immediate prompt to speak.
  • Displaying a "Privacy Shield" animation (confirming no recording is stored) during the first mic-request increases opt-in rates.
60% completion rate of the "First Conversation Milestone" (Tier 1 Engagement)
HYPOTHESES
  • Visualizing the "Tier Meter" early in the conversation gamifies the interaction enough to push users past the initial awkwardness.
  • Providing "Suggested Responses" (ghost text) in the first session reduces "freezing" when the user doesn't know how to reply.
  • Incorporating "Micro-Celebrations" (warm color pulses) within the first 30 seconds reinforces positive behavior and encourages session completion.
50% of users set a personal "Communication Objective" during onboarding
HYPOTHESES
  • Presenting clear use-cases (e.g., "Dating," "Job Interview," "Small Talk") during setup increases the perceived utility of the app.
  • Linking the AI Mentor's personality to the user's chosen goal makes the experience feel personalized and worth starting.
  • Minimalist onboarding (under 4 screens) prevents drop-off before the user reaches the core value proposition.
Retention
40% of users return for a 2nd session within 48 hours of the first
HYPOTHESES
  • Delivering a "Next Skill Unlocked" notification 24 hours after the first session creates a curiosity loop.
  • The "24-hour Auto-Delete" policy creates a "use it or lose it" psychological trigger for users to review their transcript before it's gone.
  • Suggesting a "Practice Scenario" based on the user's previous performance (e.g., "Ready to try an open-ended question today?") increases relevance.
Increase "Average Conversation Duration" by 25% over the first 5 sessions
HYPOTHESES
  • Gradually increasing the AI's use of "challenging" conversational cues (like silence) forces the user to practice longer-form speaking.
  • Introducing "Branching Narratives" in the Coach mode (secrets to discover) encourages users to stay engaged longer to "solve" the persona.
  • Dynamic background color shifts (Cool to Warm) act as real-time feedback that keeps users incentivized to keep the "warmth" going.
30% of users reach a 14-day streak of conversations
HYPOTHESES
  • Transitioning the mentor from "Teaching" to "Critiquing" at Day 7 provides a new value layer that keeps the app useful as skills improve.
  • Periodic "Skills Assessments" (quizzes or tests) provide a sense of tangible growth that justifies continued subscription/use.
  • Real-world "Challenges" (e.g., "Today, ask one person a 'Why' question") bridge the gap between the app and real life, making the app feel essential.

Exploring concepts & shifting priorities

While many respondents cited dating and conversation issues, I avoided focusing on the actual date to prevent privacy concerns and phone dependency. Instead, I shifted the app's core purpose toward private, one-on-one training. Initial sketching and wireframing helped narrow the user-requested features into a simplified, focused set.
Wireframe mockups of a mobile app interface showing screens for open-ended questions, drawers with suggestions, and contextual buttons including a heart icon and purge memory option.
DESIGN
Smartphone displaying a blue screen with an ear icon in the center and a button labeled 'SUGGESTIONS'.

Creating a behaviour-change engine

Learning by Doing
The app focuses on two major components;a contextual digital mentor, and a nuanced conversational coach with customizable help settings.

Contextual Digital Mentor

Minerva’s core is a contextual digital mentor using an entertaining "learn with the flow" method. Users explore subjects through intuitive drawers and help buttons. The app interacts with users to teach effective conversation tactics, leveraging "micro-learning moments" to provide education as they go.
It provides guidance on topics such as:
  • Active Listening
  • Open-Ended Questions
  • Finding Common Ground
  • Curiosity and Genuine Interest
  • Sharing Without Monopolizing
  • Transitioning topics or ending the conversation

Nuanced Digital Coach

Respondents requested a safe space to practice without judgment. Minerva addresses this by generating AI personas with complex backstories, secrets, and defining memories.

By weaving these facets into the AI, Minerva creates authentic characters that feel like real people, offering users a richer, more effective conversation experience.

Building ethical AI

It was crucial to architect an AI-driven experience that was ethical in its use of data as well as how the agent interacted with people.  High AI attachment is linked to loneliness, social anxiety, and a desire for predictable, affirmative feedback.  

Additionally, over-reliance can erode human judgment, increase isolation, and create "digital companionship" that can be deactivated by the provider at any time.

Furthermore, privacy and data control were paramount with regards to GDPR.
Smartphone screen showing a chat conversation with alternating blue and dark blue message bubbles and a button labeled 'PRESS & HOLD TO PURGE MEMORY' below.
Smartphone screen showing a paused conversation with a countdown timer set to 23:59:00 indicating when the conversation will end and be deleted.
Smartphone screen showing chat settings menu with options to pause conversation, toggle suggestions, show transcript, start over, and purge memory now.
Conversations are auto-deleted in 24 hours
Users have clear controls to manually purge memory
Personas are dynamic and regenerate with every conversation keeping people from developing attachments
The LLM is trained to not lead on users
RESULTS
Mobile app screen showing a glowing heart icon with a message quote: '... and honestly I was pretty hurt by that.' Below, advice on active listening encourages nodding and paying attention to words, tone, and body language.

Micro-Celebrations

In Mentor-Mode, Minerva would provide visual and haptic feedback when someone does something right by pulsing a warm color.  This provides moments of delight and stimulates the desire to watch for visual cues from your conversational partner.
Icon denoting an important metric
Because of this feature, people indicated they were 70% more likely to adopt and continue usage of the product.
Minerva screen showing Tier 0 engagement
Minerva screen showing Tier 1 engagement
Minerva screen showing Tier 2 engagement
Minerva screen showing Tier 3 engagement
Minerva screen showing Tier 4 engagement
Minerva screen showing Tier 5 engagement
Mobile image of 6 different colored screens showing Minerva's color changing background

Tier progression

As you converse with Minerva, her display changes to indicate a growing level of engagement.  The better you do, the higher tier you progress to and the interface warms up from blue to orange.   If you disengage in conversation or practice close-ended tactics then the interface cools down again and the tier meter drops.
Icon denoting an important metric
Respondents indicated Tier Progression resulted in a 3X likelihood of retention
Minerva Screen showing a badge collected during conversation
Minerva Screen showing badges and awards from participation

Gamification

As the user’s conversation progresses, the collect badges and a title adding to the excitement and a sense of tangible accomplishment throughout your journey.
Icon denoting an important metric
Respondents indicated that badges and personas were 20% more likely to keep them on the app longer than 14 days
Minerva logo showing an owl

Minerva is an ongoing project

Minerva is still growing and data is still coming in regarding how it performs in the wild.  We are currently looking for funding.
Next Up
FINTECH
Hand holding smartphone showing a banking app with a $3,240.88 checking account balance and recent transactions.
Redesigning the world’s second largest neobank to target US membership.
VIEW CASE study
E-COMMERCE
Laptop, tablet, and smartphone displaying a luxury hotel booking website with scenic resort images and search options.
Increasing Mobile Engagement by 83% through strategic UX.
VIEW CASE study