What are we building?
Mental Health Conversational AI Architecture for Billions, Starting with English & Hindi
The conversational Architecture of AI mental health model that we are working to build will be as follows:
Active Listening AI Model
The model will enable AI to really listen to what the user is saying. This means, the ability to understand the context, the concern, the intensity, and the sentiments that the user is expressing and the ability to respond, connect, empathize and engage accordingly with the user.
CBT Trained Model
Based on the type of cognitive distortion and the issues, the model should be able to identify, suggest and enable user to complete the right CBT exercises over a span of 6-8 weeks to enable recovery, functional improvements and real changes in their life.
Story & Context Specific Narrative Generator
Stories, visualization, and narratives that motivate inspire, help reflect, and enable a feeling of being understood, are crucial for a person dealing with mental health. Identifying what would inspire a person and when, and the ability to produce that in real-time, can really prove beneficial for people dealing with mental health issues; especially as a part of a wider program
Tracking & Behavior Management Model
The model should be able to track emotions, use this information to recommend action, generate conversations that help, and monitor activity levels.
A Skeleton-Rule-based Model
The key to ensuring that the AI works for people with mental health concerns & does this consistently at scale, is its ability to follow a mental health conversation approach. Opening conversations at critical points as well as the ability to follow a guided path at others, as well as the ability to avoid saying or suggesting things that can be or are harmful is the key. Hence looking at this, a structured amalgamation of sub-conversational agents in a specific flow will enhance the efficacy of the whole model as well as sub-parts. Also this would be key initially in ensuring that the overall AI program is always in alignment with the goal of effective, safe, and supportive mental health care recovery program.
Where We Are Currently & Next Steps?
IWIll has already developed and launched a limited-release AI based CBT model for depression and anxiety. The CBT model & exercises with 5% open conversations and empathic responses model are deployed and currently being improved upon in English.
Hindi CBT AI is now being developed with a grant awarded by Microsoft AI4A to IWill.
The program is ready for its next phase of development to advance models, listed above, build on existing available models where possible, and customize them to enhance the efficacy.
Our Market Access
We as one of the firsts mental health startups from India have access to several state governments in India with individual states having a population of 20 million+ population on average, several large healthcare listed companies, NGOs, corporations, and direct consumers. This market access gives us the ability to build and improve the model as a part of the communities, for who the product and solution are aimed.
Access to anonymized data, clinicians and SMEs, and internal resources is another aspect that helps us in envisioining and building one of the most advanced AI programs for mental health.
Our approach to Efficacy
We are using global standardized self-reporting scales of improvement for mental health for measuring and improving efficacy of our mental health program.
Once developed, IWIll will be undertaking a randomized controlled trial to establish the efficacy of its mental health AI model (IWill-M) for depression and anxiety and show improvements and recovery for users.
We have academic partners but are constantly looking for, and are open to collaborating with partners, teams, clinicians, and academic and research institutes to ensure the safe development, use, and growth of our AI model for mental health.
Current Partners & Call for Collaboration
In our journey of using technology, to build AI for mental health and create access, we have been supported by Microsoft, ACTGrants, CommonWealth Centre for Digital Health, The George Institute for Public Health & UNICEF.
We are especially looking to partner with research-driven large-scale AI partners, academia, and mental health communities that enable us to reach our goal and be able to build, scale and advance access to quality mental health, a stream where in India alone, there is 83% treatment gap and clinician availability is less than 0.001. More than 1 billion in the world need help today and that number is 300 million in India.
Vision is to share learnings, models and advancing AGI
What we build for mental health, especially empathic conversations and active listening will have applicability in other key fields where AI can benefit humans exponentially. We are more than willing to advance this work together and help build better models of the future that benefit humanity. We will do this by ensuring we are ethical and respect people and their privacy, digital space at all times.
Why are we building this architecture?
Key elements of an effective mental health program of care and therapist skills
A good mental healthcare program needs many interventions, activities, support, and working in parallel and in sync to help a person dealing with a crisis or a long-standing situation get reasonably better and clinically and functionally live to his/her best potential.
For the purpose of building a conversational program that is able to reliably do its job at par with humans and near therapists, these key skills are required.
A program like this would have a good empathic clinician helping the person with active listening. Active listening is important because it means not just listening to someone but the ability to connect, reciprocate and extend the conversation in a desirable direction to enable dialogue and healing.
A good clinician will use Cognitive behavioral therapy to help the person work on their painful thoughts and patterns in a systematic and sustained manner. This skill requires an understanding of the type of cognitive errors a person is dealing with, the triggers and contexts in which these arise, and the ability to recommend and help follow with accurate CBT exercises and work.
Inspiring tales & conversations
Connecting the dots and sharing stories, sharing inspiring anecdotes, and information is known to be helpful for someone dealing with mental health concerns especially when the person is feeling low, is unable to feel motivated and the odds feel like stacking against the person.
Coordination of Care & Right Behaviors
Social integration enhancement, reduction in behaviors that are unhealthy, and adoption of new behaviors and patterns to sustain and support recovery, are also coordinated by a good mental healthcare professional and/or a program.
Identifying red flags in time and suggesting the right step
Mental health issues despite the best care may sometimes escalate and any system’s ability to be able to identify these signs well in time and guide next step behavior, knowing when to fail or moving the person to higher order care is extremely crucial and albeit of utmost importance.
Measuring impact, clinical evidence, and future course
Outcomes are the most important and Crucial Key Result that a program for mental health harnessing the power of AI should deliver without fail. All the above skills that a mental health program consists enable this to happen, but measuring this progress continuously, taking steps basis the measure, and delivering outcomes as well as laying a plan for further support is what is key for a program to achieve the intended outcomes.