Curo | Devpost

What Inspired Us

As four young men navigating the complexities of modern life, we are empathetic to stark realities of the mental health challenges facing society today. As four students passionate about the power of data and technology, we are also well aware of the potential of the Internet of Things to inform psychiatric intervention — a potential which has been validated by clinical literature [1]. On one hand, the inspiration for our project emerged from a deeply personal place—a realization that each of us, or someone we know, could silently be on the brink of despair. On the other hand, our idea emerged from the growing body of literature about the potential for digital phenotyping in psychology, whereby our digital data can inform diagnoses. We decided we wanted to tackle the issue of reactive psychological intervention, inspring us to develop a more proactive, data-based method to prevent the many horrible consequences of mental illness, specifically depression. Our project is more than an app or a service; it's a movement toward creating a safe, inclusive, and accurate mental-health tool that furthers trust, accessibility, and feasibility in .

What We Learned

One of the most enlightening aspects of our journey was the discovery and understanding of digital phenotypes and their potential in identifying mental health issues. This concept quickly became a cornerstone of our project. Digital phenotypes refer to the collection of digital data that relates to an individual's behavior and interactions with digital devices and platforms [1]. We learned that these data points, when analyzed thoughtfully, offer a wealth of quantitative information that can help in the early detection of mental health issues. Moreover research in this space is showing staggeringly promising results in how good digital phenotypes are as predictors of mental illness.

The Power of Quantitative Data

Our exploration into the realm of digital phenotypes revealed the vast potential of using quantitative data to gain insights into an individual's mental health. This data isn't just numbers and statistics; it's a reflection of behavior patterns, social interactions, and even changes in mood and mental state over time. For instance, variations in movement, sleep, and patterns of device usage can all serve as indicators of an individual's psychological well-being [1]. Numerous instances of research pointed to the fact we could find concrete, validated ways to utilize available quantitative data such as location to create metrics on key mental illness predictors like sociability. For instance, one group was able to employ machine learning on GPS- and phone usage derived user features to predict depression with 80% accuracy. This study reported population means for these features (such as location variance, location entropy, and mean screen time per day), which informed the statistical methods that we utilized in our app [2].

Gamifying Mental Health: A Creative Fusion

Harnessing the power of data science and innovative algorithms, we embarked on a journey to redefine mental health support by integrating the concept of gamification. Our aim was to transform the daunting task of managing mental health into an engaging and motivating experience. By gamifying mental health, we sought to break down the barriers of traditional mental health care and create a more accessible and appealing approach for users through a simple daily score that could be tracked. We want to create a positive reinforcement mechanism for behavior associated with good mental health and provide personalized motivational feedback as soon as any increase in negative behavior occurs

The Culminating Solution: CURA

We monitor several key digital data markers of our users such as location, screen time, sleep and more, all of which were chosen due to their strong peer-reviewed researched correlation with positive or negative increased risk of depression [2]. We used many data science methods such as advanced unsupervised learning algorithms to cluster user data such as location to get actionable insights such as the number of places the user has visited in the last week, which have been shown to correlate to mental state. We then present the user with an aggregated score on how good their habits have been in the past week based on 8 key metrics, as well as their habits generally compared to the population means of depressed and non-depressed individuals [2]. We present these two data points to the user in a very gamified manner to reward increases in positive habits. To also leverage the benefits of personalized based medical interventions [3], we have carefully prompted and created a chat to support the user and give personalized motivational and informative messages to the user. If the user exhibits abnormal negative habits and behaviors we alert them to resources where they can get help.

How We Built Our Project

Team Structure and Roles Data Science and Algorithms Expert: One of us took on the challenge of diving deep into data science and algorithms. This role was crucial for developing the scoring system that underpins our gamification approach. By analyzing user interactions and behaviors, we created a sophisticated model that personalizes the user experience, offering tailored challenges and feedback.

Application Development Trio:

API Integration Specialist: Focused on connecting our app with various external services and data sources. This role involved making API calls to fetch and send data specifically using Apple’s API calls for phone data and GPT API for our feedback. This ensured our app could interact seamlessly with third-party services and our backend infrastructure. Frontend Developer: Dedicated to crafting the user interface and experience. This involved using Swift to create intuitive and engaging layouts that would keep our users motivated and engaged in their mental health journey. Full Stack Architect: Served as the bridge between frontend and backend development. This role entailed overseeing the app's overall architecture, ensuring that both the client-facing and server-side components worked harmoniously together. Technical Architecture Our project's backbone was a carefully planned technical architecture that emphasized efficiency, scalability, and user experience. Swift for Frontend and Backend: We chose Swift as our primary development language due to its robustness, performance, and seamless integration with Apple's ecosystem. Swift allowed us to create a fluid and responsive interface while also handling backend logic with efficiency. Firebase Database: For our database needs, we utilized Firebase. Its real-time database capabilities and easy integration with Swift made it an ideal choice for storing and retrieving user data quickly and securely. Google Cloud Functions: To incorporate our data science models into the app, we leveraged Google Cloud Functions. This allowed us to execute our algorithm-driven scoring system in the cloud on a timer to receive scores daily while seamlessly updating the firebase data.

Development Process

Our development process was iterative and agile, focusing first on establishing a solid architecture that would support both the immediate and future needs of the project. Architecture Planning: We started by laying out the architecture, ensuring that our choice of technologies would allow us to build a scalable and maintainable app. Division of Labor: With our architecture in place, we divided the workload according to our individual strengths and areas of expertise. This division allowed us to work efficiently, with each team member focusing on their respective domain. Integration and Testing: As the app began to take shape, we prioritized integration and testing, ensuring that each component functioned as expected and that the user experience was smooth and engaging. Incorporating Data Science: The final step was to integrate the data science algorithms via Google Cloud Functions, allowing our app to offer a truly personalized and gamified mental health experience.

Challenges We Faced

Ideation: Finding the Sweet Spot The initial phase of ideation was perhaps the most daunting challenge we faced. Our goal was not only to innovate but also to ensure that our project would have a real impact on mental health. Striking the right balance between novelty, technical feasibility, and genuine usefulness took countless brainstorming sessions, research, and discussions. We aimed to create something that wasn't just another app but a revolutionary approach to mental health support. Navigating API and Privacy Constraints with Apple Integrating with third-party services through APIs while adhering to Apple's stringent privacy guidelines presented a significant hurdle. Apple's ecosystem prioritizes user privacy and security, which, while beneficial for users, imposed limitations on how we could collect and process data. We had to meticulously plan our API calls and data handling processes to ensure compliance with these guidelines without compromising the functionality and user experience of our app. One of the drawbacks of our app is that it requires user data that they might prefer to keep private, so we are sure to notify them what it is for and when we are using said data. Algorithm Creation and Data Integration Developing the algorithm that underpins our scoring and recommendation system was a complex task. Balancing the desire to provide valuable, tailored advice with the understanding that we are not the actual individuals experiencing these mental states required a thoughtful approach to algorithm design. We leveraged extensive studies and data science techniques to analyze patterns and behaviors while ensuring our recommendations remained sensitive to the nuances of individual experiences. Personalization vs. Generalization A critical challenge in our development process was finding the right balance between personalizing the user experience and acknowledging the vast differences in mental health conditions and responses to treatment. We recognized that while our algorithm could identify patterns and suggest actions, the subjective nature of mental health meant that one size does not fit all. This was perhaps the most difficult. The last thing we wanted to do was tell someone how they were feeling, or what mental state they were in. To mitigate this, we focused solely on quantitative data that we could present in a way to show them how they might be changing their behavior – and if it was for the worse by our measures – we notified them of this. Facing and overcoming these challenges was a pivotal part of our project's journey. Each obstacle provided us with valuable lessons on the importance of flexibility, user-centered design, and the ethical considerations of developing health-related technology. Further, the technical challenges forced us all to take an open to learning approach in order to adapt to every part of the project.

Built with What languages, frameworks, platforms, cloud services, databases, APIs, or other technologies did you use?

Citations: [1] Montag, C., Sindermann, C., & Baumeister, H. (2020). Digital phenotyping in psychological and medical sciences: A reflection about necessary prerequisites to reduce harm and increase benefits. Current Opinion in Psychology, 36, 19–24. https://doi.org/10.1016/j.copsyc.2020.03.013

[2] Opoku Asare, K., Moshe, I., Terhorst, Y., Vega, J., Hosio, S., Baumeister, H., Pulkki-Råback, L., & Ferreira, D. (2022). Mood ratings and digital biomarkers from smartphone and wearable data differentiates and predicts depression status: A longitudinal data analysis. Pervasive and Mobile Computing, 83, 101621. https://doi.org/10.1016/j.pmcj.2022.101621

[3] Teeny, J. D., Siev, J. J., Briñol, P., & Petty, R. E. (2021). A Review and Conceptual Framework for Understanding Personalized Matching Effects in Persuasion. Journal of Consumer Psychology, 31(2), 382–414. https://doi.org/10.1002/jcpy.1198