Mermind is a mental health platform for college students offering a space for judgement free venting, track their moods and seek support. Jump to results ↓

Role

Researcher,

Product Designer,

Visual Designer

Timeline

6 months

Collaborators

Satvik Shrivastava

Me!

The Problem


College represents a crucial developmental period, where students face a tangle of numerous challenges across academic, social and personal domains. In a significant proportion, these stressors interact with genetic and psychosocial adversities, increasing the likelihood of mental health issues.


The existing support infrastructures (e.g SWC in IIT-B) fail to provide and match the safe and necessary space for students. Existing mental health tools are also either too generic, or too exposed to feel safe. We noticed a lack of awareness, existing stigma and no effective space to address these issues.

Research Strategy


Our research was grounded in interviews and conversations with students and mental health professionals - personas representing reality; as well as literature on digital mental health, positive habit formation and the role of AI as it increasingly becomes ever-present. This surfaced key learnings and insights that shaped everything downstream:


Students found comfort in relatability and wanted to choose between that and solutions, not have one forced on them. They were afraid of being 'seen' but wanted community and a sense of belonging. However, misinformation and stigma were also rampant, especially with social media, which is why we considered anonymity to ease things.

Insights & Thinking


Research consolidated into seven opportunity areas, each framed as open design questions rather than a solution


Personalized Guidance & Support · Facilitating Awareness & Reducing Stigma · Emotional Tracking & Insights · Comfort vs. Solutions Pathways · Professional Support Network & Infrastructure · Venting & Community Support · Ethics & User Privacy

An Experiment to Validate our Findings


Anonymity as a spectrum, We designed an experiment to test the feasibility of an anonymous venting channel at the university level, inviting users to share their frustrations or rants in a completely anonymous setup. Rather than asking students if they'd use anonymous venting, we tested it in an unmoderated setting.

Leveraging AI


"Do we really need a machine to respond too when there are humans responding?"


Our early plan leaned towards a humanlike companion until a discussion with a mental health professional stumped is. We reprioritised the AI aspect to be transparently non-human, summarising and surfacing information rather than generating advice and never encouraging self-diagnosis.


We even tested this with a custom GPT instructed on our intended tone & tested with real user interactions, allowing us to assess its effectiveness of responses within our mental health framework.

57% · Genuine Vents


Validation: Students vented about their body image issues, missing their old friends, frustration about workplace, general mood.

22% · Random


Expecting the Unavoidable: People like posting random things on the internet and it if it helps them cope in someway, it’s mostly okay.

21% · Trolls


Need for Moderation: Anonymity is powerful and there always are trolls who’d try to bring other people down. Moderation is layering and has to be thoughout.

57% · Genuine


Validation: Students vented about their body image issues, missing their old friends, frustration about workplace, general mood.

21% · Trolls


Need for Moderation: Anonymity is powerful and there always are trolls who’d try to bring other people down. Moderation is layering and has to be thoughout.

22% · Random


Expecting the Unavoidable: People like posting random things on the internet and it if it helps them cope in someway, it’s mostly okay.

Anonymity earned honesty, but it also invited abuse, at a real rate, not a hypothetical one. This result validated the core premise and shaped the moderation model that followed.

Leveraging AI


"Do we really need a machine to respond too when there are humans responding?"


Our early plan leaned towards a humanlike companion until a discussion with a mental health professional stumped is. We reprioritised the AI aspect to be transparently non-human, summarising and surfacing information rather than generating advice and never encouraging self-diagnosis.


We even tested this with a custom GPT instructed on our intended tone & tested with real user interactions, allowing us to assess its effectiveness of responses within our mental health framework.

We scoped toward a reminder system, curated and moderated resources, campus-facing posters, professional escalation for urgent vents, and an anonymous venting core for this build. The idea was for users to be able to select their own path and pace, while keeping it open for everyone.

We scoped toward a reminder system, curated and moderated resources, campus-facing posters, professional escalation for urgent vents, and an anonymous venting core for this build. The idea was for users to be able to select their own path and pace, while keeping it open for everyone.

How the System holds together


We built a solid conceptual framework of the system - showing the relationship between different system areas, how they interact with the user, data and each other.

Click here to see the framework on user data.

Visual Design


We drew inspiration from Tamagotchi-style pets before landing on a fish-based metaphor, tying naturally into "school" and ocean imagery that matched the name MerMind. That metaphor carries through the whole design and UI: mood states expressed through the fishes, soft gradients, and a card-based layout that keeps a heavy subject feeling light to navigate.

The Final Product


The final system - comprising of the MVP application and supporting print & design media - was prototyped and tested with users. You can try out the prototype for yourself here ↗ or watch the videos below!

Leveraging AI


"Do we really need a machine to respond too when there are humans responding?"


Our early plan leaned towards a humanlike companion until a discussion with a mental health professional stumped is. We reprioritised the AI aspect to be transparently non-human, summarising and surfacing information rather than generating advice and never encouraging self-diagnosis.


We even tested this with a custom GPT instructed on our intended tone & tested with real user interactions, allowing us to assess its effectiveness of responses within our mental health framework.

Validation


We showed and tested the prototype with students across campus, as well as mental health professionals. A lot of praise was received for the initiative and solutions, as well as the structure and design of the apps. We received a lot of valuable feedback about onboarding and data collection, personalisation, safety features and data presentation.

What next?


This project was a great learning experience and something I would like to take forward in a meaningful way. However, somethings I would work on are auditing and simplifying the feature breadth, refining the mood tracking aspect, improve the data representation and design for empty and edge-case scenarios, that the timeline didn't leave room for.