Autis(CASP)

Autis(CASP) is an AI-Powered Autism Screening Assistant. A user friendly chatbot that helps users self-screen for an estimated level of severity on the autism spectrum based on a user's symptoms.

Comment

Inspiration

We created Autis(CASP) to help more people understand and be able to estimate autism early on. We know it can be hard to get a professional autism screening, so we made a friendly, easy-to-use AI chatbot that gives people 24/7 access to a autism screening tool to best estimate their case of autism and where they lie on the spectrum based on 7 criteria as outlines by "The Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition". We also had access to an award winning research paper, co-authored by one of our own members, which assisted us in creating ideas and queries for the project. Paper

What it does

The chatbot prompts users for symptoms relating to Autism and helps them understand what symptoms they or their patient may be experiencing with helpful prompts. It chats with you in a kind, understanding way, asks about things like how you interact with others or handle daily routines, and gives you a summary of where the patient fall on the autism spectrum or what level they are. To be clear, Autis(CASP) is not a doctor and can't diagnose autism. It is a helpful first step in estimating the presence of autism based on user input/data.

How we built it

  • On the surface we used Next.js package to create a user friendly UI and a modern chatbot that can interact with users. It utilizes React.js components in JavaScript with embedded CSS for UI effects.
  • The chatbot is run by Google's Gemini-1.5-flash generative AI API.
  • A python script runs the s(CASP)/Prolog queries taking input formatted from the chatbot AI.
  • We used s(CASP)/Prolog for the logic engine of the project ensuring proper logic is followed to come to a sound conclusion, based on proper scientific research, to report back to the user (avoiding the usual hallucinations of modern AI).

Challenges we ran into

Finding a topic was a bit difficult at first since it proved quite time consuming but we were able to develop our idea based on available autism research and delegated tasks. We ran into problems finding a suitable AI to run the chatbot and format the user responses/input. Our first choice, Meta's Llama LLM AI, was unfortunately not free as we discovered so we had to make a time consuming switch to Gemini AI. Creating questions that were clear but also sensitive and making sure the AI understood user input was also challenging.

Accomplishments that we're proud of

We created a stellar modern UI chatbot and and a solid working s(CASP) logic code to back our project based on real scientific data we collected from the "The Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition", the first of its kind, We are also proud of the work of one of our source: an award winning research paper co-authored by one of our members.

What we learned

We learned how s(CASP) works and how to connect it to a python script and an AI powered chatbot. We learned just how complex Autism can be whilst creating the logic code for the project.

What's next for Autis(CASP)

We hope to expand this project to further include other disorders perhaps overhauling it into Ment(ASP) where it can handle queries for multiple mental disorders. Perhaps afterwards we can expand to physical disorders and other illnesses. This could transform into a doctor's assistant available to provide logic based real estimation tailored to each individual user/patient at a mass scale on demand 24/7. This has the potential to democratize access to mental and physical healthcare, while also being scalable for use in hospital systems around the globe--bolstering healthcare accessibility and affordability for all--and reducing significantly wait times for diagnosis and treatment. Of course, nothing replaces a human doctor, so their constant input would be highly beneficial to this project. Another aspect we seek to improve is utilizing a custom LLM instead of a GPT AI to further enhance the user experience and build medical knowledge overtime. That way, Autis(CASP) and it successor Ment(ASP) can be more reliable as it utilizes a proprietary LLM trained on and continually trains on medical data.

Presentation

Built With

Share this project:

Updates