Integrating Conversational Pathways with a Chatbot Builder Platform

Abstract

MIRA, a Mental Health Virtual Assistant, was developed during the COVID-19 pandemic to address rising mental health concerns and the demand for support and services. As a task-oriented chatbot, MIRA connected healthcare workers seeking mental health support with trusted services and programs. It achieved this by integrating Natural Language Processing (NLP) within a structured conversation flow, designed by experts in our multidisciplinary team to cater to healthcare workers. This work extends MIRA across diverse populations to improve accessibility to mental health services. The pilot version of MIRA was implemented in Python using an open-source chatbot development framework. Its conversational logic was hard-coded, and minor changes to the conversation flow necessitated extensive changes to the code base. Such a rigid system made it challenging to extend the functionalities of MIRA to cater to other languages or new populations, such as first responders, Veterans, Indigenous communities, or youth. In addition, building relevant and effective solutions that cater to community-specific needs requires insights from subject matter experts and close collaboration with a multidisciplinary team, including psychiatrists, domain experts and people with lived experience. Scaling MIRA’s functionalities to serve diverse populations requires being able to easily integrate additional conversational pathways without changes at the code level. Hence, we introduce a chatbot builder platform and a conversation flow visualizer to simplify the creation and modification of conversational flows by non-technical experts. Our tools support multidisciplinary collaboration on MIRA’s development process by enabling iterative improvements and rapid modifications to the conversation flow. Our work has been successfully deployed and is available for public use.