Clarity -- Journaling AI Application

What it does

Clarity is an advanced Journaling AI Application designed for enhanced mental well-being. It features a chatbot system for regular journaling, diverse prompts for self-reflection, and AI-driven personalised responses. There is an ability to make the entry on your own or aided by the chatbot, as well as a calendar to reflect on past entries.

How we built it

We build a prototype utilising Figma, and ideally would create a backend and frontend in the future using a combination of frontend development with HTML, CSS, and JavaScript (React).

With regards to the backend, we identified five different tasks engineered to enhance the user’s experience while making an entry to his or her journal. Firstly, we understand that one of the primary reasons for people to write personal diaries or journals is for self-introspection, reflection, and to vocalise one’s thoughts to a non-living entity to remain personal and private. This can be especially hard for teenagers and the modern youth, and a simple act of assurance and kindness to elevate their mood can go a long way!

Hence, we introduce Clarity- a state-of-the-art journaling interface that not only allows users to share their deep and personal thoughts, but also understands their position of vulnerability to uplift their moods with encouragement, positive reinforcements and friendly anecdotes.

If the user wishes, he or she can also initiate a conversation with Clarity, which will remain friendly and encouraging to ensure their experience of personal reflection and introspection is fully realised. At the end of the journal entry, the user will be awarded a badge which is generated using stable diffusion with the v1-4 model. The contents of the journal entry was first summarised using Zephyr 7B Beta and passed into the stable diffusion to reduce the context size for the diffusion model. We also prompt-tuned Zephyr to provide a highly descriptive and imagery encrusted summary to ensure the badge comes out to be the best representation of the user’s journal entry.

Fourthly, we also added a functionality to allow users to relive their best days through storytelling. This module looks into the entries of the past to collect associations of all happy memories to create a cohesive story for the user to read through. This provides a way for the users to relive their past experiences through the eyes of another. Additionally, the user can also request for other associations, i.e, creative memories, eventful memories, playful memories, leisurely memories, etc.

Finally, for our back-end database management system to not only store but also sort the data in a meaningful way, we included a mood predictor model: RoBERTa fine-tuned on mood classification. This will enable us to cluster similar moods together for quick processing to provide feedback to the user about his or her moods, behavioural patterns and trends. This will be especially useful in the future iterations of our application where the users can not only get weekly reports on their activities and moods but also for our application, Clarity, to become what it was meant to be - a companion.

And finally, as a trial module, we are also deploying a medical inference model that can provide advice on managing mental health, suggesting specific diets and nutrition, recommending actions to take personal care, etc. While this certainly includes the risk of false positives and negatives where our model may be taken advantage of, we felt that this could be a good innovation to explore within the context of the hackathon but when it comes to production and real-world implementation, we must be considerate of all possibilities to ensure the utmost safety of our users.

Challenges we ran into

One of the prominent challenges encountered during the development of our project was the absence of a dedicated frontend engineer. We had to rely on simulating the final product using Figma. Navigating this hurdle required us to leverage our collective skills and knowledge, especially considering that most team members were relatively new to the Figma platform.

Accomplishments that we're proud of

The achievement lies not only in the technical aspects but also in our commitment to promoting mental well-being. We are proud of taking on Figma, and also coming up with an idea that we all collectively agreed on without any issues.

What we learned

We learned a lot from our backend engineer, who explained how the product would ideally be hosted on a server. We entered with many incredible ideas, and realised just how difficult the implementations actually are. The entire hackathon has been an incredible learning process from start to finish, as two of us are undergrad CS majors who have very little experience with tackling app development. Working with a multidisciplinary team has been an incredible experience for all of us.

What's next for Clarity -- Journaling AI Application

The app would ideally include cognitive exercises, recommending activities like puzzles for improved cognitive functions, and offers mindfulness and meditation suggestions based on the user's emotional state. In the future, we would hope for its adaptive content delivery to refine prompts and activities based on user feedback, guaranteeing the AI adapts to evolving user preferences over time.