Monash University Confidential Companion
Researchers from Monash University wanted to create a bot that could act as a confidential conversational partner for at-risk young people. Goals included providing strategies to deal with stress, anxiety, depression, along with resources for young people dealing with other serious issues. Privacy and security was paramount, so Signal was chosen for its end to end encryption. The client also required a feature to alert an emergency contact in the event that the user expresses a desire to do self-harm.
Research common social and mental health issues experienced by Australian young people.
Integrate both user research and academic research into a cohesive content strategy to appeal to the target demographic.
Develop a conversation designed to engage users, provide therapeutic content, and direct users to external resources, depending on the nature and priority of the content.
Develop content to integrate proven support and therapeutic strategies including cognitive behavioral therapy, mindfulness, and meditation exercises.
Develop personality and character sketches for conversational interface.
Write localized, relatable content in character and on-brand across 40+ topics of conversation.
This general conversational interface was a challenge to create, as it needed to listen to users and provide useful responses to a wide range of inputs. This project was initially designed as a proof of concept, so I focused on developing the core, high-priority conversation flows and content required by the client, and left the possibility of further development open. I designed a conversational structure that would accommodate a broad range of user inputs at any time, while providing in-depth conversation in high-priority areas.
After content design, I developed a content strategy based on user research. I focused on the areas of content that would be most valuable to the user based on the user research and most valuable to the user from the therapeutic point of view based on input from our subject matter experts.
To determine what would be most valuable, our user research team conducted primary user research with the target demographic using both surveys and interviews. We also conducted secondary research into the likely users, focusing on subject matter that would be relevant. In addition, we drew on the expertise of our academic partners at Monash University, who provided insights into the needs and issues experienced by the potential users, in order to determine the highest priority topics to cover.
After the research phase of the project, I developed two separate personalities, sample content, and style guides, to for presentation to the client. Working with psychologists, we arrived at a guiding, empathetic personality, and an interaction style that would focus on listening rather than providing advice.
After iterating on the character, style guide, and sample content, to avoid any blind spots and nail down the tone and voice, the content development began. As a content heavy project, this bot required hundreds of lines of content. I directed the content team and edited the content, as we went through several rounds of review by an academic panel of subject matter experts to guarantee that the tone would provide the best experience and support for users, and any unexpected triggers could be avoided.
After content approval, the conversation and content was implemented using the proprietary Botanic bot platform and author's tool, and the bot was deployed on Signal.
This proof of concept entered the testing phase. Monash University researchers will put monitor users’ interactions with this bot in controlled environments before further iterating on the concept.