Humma.ai
Summary
Humma.ai is an AI startup aiming to build a social media platform using Empathetic AI™ models. Part of their vision for their social media product is to discover ways to integrate AI to help its users feel more connected with its community through a new feature coined as Bumping, which helps users interact with each other based on recent activity on their community feeds.
Role
UI/UX Designer
UX Researcher
Moderator
Tools
Figma
FigJam
HubSpot
UserTesting
Zoom
Duration
6 months
Reimagining how people can connect online.
Humma.AI is an emerging social platform committed to creating safe, culturally competent online spaces, with a strong focus on privacy, community, and inclusivity.
Over the course of the quarter, the our team completed mixed-method research to understand online community connection and inform Humma.AI’s community experience, specifically for individuals who are underserved by mainstream social networks.
As one of the UX Designers and Researchers for this 6-month long project, I helped design and ideate a variety of new community-focused features - including the Community Hub, Activity Radar, and Bumping - to create space for authentic, low-pressure connection.
The reality: People crave community, not content.
"Local community participation may influence socioeconomic mobility of individuals across their lifespan and also reduce large-scale socioeconomic disparities."
US Department of Health and Human Services, 2023 (Source link)
In 2018, only 16% of Americans reported that they felt very attached to their local community.
We heard a consistent story across our initial interviews, where we sought after people who actively use social media, or recently stopped.
What we found out from a large portion of our interviewees was that they were tired of the isolating and performative nature platforms like Instagram, X/Twitter, and Facebook feels so steeped in.
What they desired - or even reminisced on - was a digital experience that mirrored real relationships. Not an experience that only continued push them away.
Between what was learned from users, our own personal experiences, and the values Humma.ai and the stakeholder have, we knew that whatever we intended to design must be rooted in trust, curiosity, and emotional safety.
Underrepresented groups feel the pain the most.
All of us on the team, as well as the stakeholders we collaborated with, come from a diverse background of genders, sexualities, nationalities, and more.
With the pilot of Humma.ai intended to be in Los Angeles, CA - one of the most diverse cities in the world - it was no surprise that everyone involved in this project had some sort of experience of exclusion or inaccessibility on mainstream social media.
That is why we ensured our design process was rooted in inclusivity - in the way we accommodated interviews, sourced a diverse pool of participants, and ensured the designs and content never excluded anyone.
To build inclusively, we had to start where people already belonged.
We grounded our work in research, combining data with lived experiences. Throughout the course of this project, we prioritized four different research and testing methods:
24+ sources
LITERATURE REVIEW
20+ apps
COMPETITIVE ANALYSIS
10 participants
1-HOUR MODERATED INTERVIEWS
12+ Los Angelinos
INVITED TO CO-DESIGN WORKSHOP
Key Insights
People crave authentic dialogue, not surface-level engagement. Platforms that are built on attention and algorithms don't necessarily foster that these days.
Safety comes from control. Given the ability, if users can choose how they can connect with others, they're more likely to both feel and be safe online.
Real-time interactions feel more alive and human. The asynchronous nature of many social media platforms and feeds prevents from being the case.
Designing for belonging, trust, and flexibility.
From these insights, we shaped three core principles that guided every feature onward:
The Hub: Building a digital town square.
Every community thrives when there exists a "hub" of sorts. Where interactions can be full or fleeting, where passerbys can become friendships, and where people convene to participate.
To achieve this, we first designed the Community Hub — an open, conversational space for groups within the app where users can join in small activities, view events, or jump into group discussions without pressure.
Additional Screens
Activity Radar: Sharing moments, in real time.
The Activity Radar helps people discover what’s happening in their community in real-time - from live events, active polls, trending discussions, and even active chatrooms. Think of it like stepping into a community space and seeing little pockets of people and activities that you can jump right into.
It’s a quiet nudge to re-engage, showing nearby or in-app activity while maintaining privacy and consent.
After testing, we focused on clarity and calm - balancing visibility with respect for users’ emotional space.
Bumping: Encouraging spontaneous connection via AI.
Bumping adds a human touch to the digital world. When two users interact with the same post or event around the same time, the AI Concierge™ prompts a small, friendly gesture: a “wave,” a short note, or a chat invite.
Every element is opt-in, guided by tone preferences, and safeguarded by AI moderation for early interactions.
Additional Screens
At first, Bumping did not resonate with users.
It is well worth noting that Bumping was initially met with a mixed reception that was both expressed and observed during user testing. During both moderated and unmoderated testing sessions of the initial prototype, we received feedback and suggestions that turned into design changes.
Users either ignored or dismissed Bumping altogether.
The original prototype used mini pop-ups on the bottom right to notify the user for Bumping.
When asked, many users mentioned that closing or ignoring "pop-up" notifications have become second nature to them, due to their "intrusive" and "interruptive" nature.
How did we fix this? We integrated these notifications within the feed itself, in order for it to feel more seamless in the scrolling experience.
Users were afraid of how it could be taken advantage of.
For users that did eventually understand its concept, some users had expressed initial discomfort at how this might prompt unwanted messages or interactions.
While we aim to provide a space where everyone feels welcome, we understand that is not the current society we live in.
Two solutions we recommended to reduce noise or prevent bad actors was to re-evaluate this as an opt-in only feature and utilize consensual AI monitoring.
Final Results
80% of users after testing said they would use these features in real life.
Across 33 moderated and unmoderated user testing sessions, participants repeatedly described their experience testing these features with generally positive reactions, both on a functional and emotional level.
Users told us they finally felt they had permission to reach out, to be curious, and to connect without feeling like they were performing. That sense of safety and approachability validated our goal: design features that bring warmth and humanity back to digital interaction.
Bumping, Activity Radar, and the Community Hub were recommended for initial launch.
Despite our limited time frame and scope during this project, our testing and concept validation allowed us to formally recommend Humma.ai to prioritize the Community Hub, Activity Radar, and Bumping for their product pilot.
Each of these features solved a distinct need:
The Hub provided structure for open, honest dialogue.
Activity Radar helped people notice and join shared experiences.
Bumping transformed those overlaps into spontaneous, authentic connection.
Our core pillars of Belonging, Trust, and Flexibility stayed true.
From the very beginning, these three pillars acted as our north star - and by the end, every design decision traced back to them.
Belonging: People felt invited into communities that reflected their identity and interests.
Trust: Transparent onboarding and consent-first flows helped users feel safe sharing and responding.
Flexibility: Adjustable settings for tone, timing, and visibility gave people confidence to participate at their own pace.
Seeing users articulate these values back to us during interviews was the clearest sign we had succeeded.
Personal Learnings
AI Transparency: Explicit transparency and consent of AI being involved in various features was always appreciated, and encouraged confidence in using them.
Interruptions: Just like in real life, most people don't appreciate the sense of being interrupted. This helped us pivot away from the "notification" approach of features such as within bumping.
Change Is Slow: Ultimately, this is a social media product at the end of the day. It is at the whims of the users that use it. But we hoped that our goal of an empathetic design approach in a familiar setting may be the catalyst for better things to come.
Opportunities
Additional Concept Validations: on additional descoped/less-prioritized features, such as Question of the Day, Quests, Feedback Loop, and more.
Expanded Participant Pool: We focused on working closely with Los Angeles-based participants for the product's intent on launching exclusively in the county when it starts.
Demographic Gaps: We hoped to gain insight and feedback from demographics that were less represented in participant sourcing, to continue to ensure that features, language, and tools available within Humma.ai is as-inclusive as possible.
Appendix
To view the final report on this project, including expanded information on all sections, all sources, and data gathered, please contact me.
Credits
Design Team: Justin Gamiao, Bailey Herbstreit, Anya Menon, Matea Montanaro, Kayla Nguyen
Sketches: Matea Montanaro
Special Thanks: Humma.ai Team, UC Irvine, West Los Angeles College