A student seated and holding up her laptop

ASU students help build AI chatbot for crisis support center

If you’re facing a mental health crisis, domestic violence threat or eviction from your home, you need help fast. Crises happen 24/7, outside standard business hours and with no time to waste searching for services on the internet. Arizona State University’s Artificial Intelligence Cloud Innovation Center (AI-CIC) powered by Amazon Web Services built a production-ready solution awaiting rollout: the HOPE chatbot, an artificial intelligence solution for crisis support.

ASU student interns led by Arun Arunachalam, AWS senior solutions architect, developed HOPE (short for Helpful Online Personal Engagement) for WestCare. The large nonprofit behavioral health and human services organization provides programs for people dealing with substance use disorders, mental illness, homelessness, domestic violence and trauma.

“In a remarkably short time, the students helped us bring HOPE to life in a way that truly reflects our commitment to compassionate, accessible care,” said Diane Ludens, WestCare vice president of software development. “Their professionalism, speed and deep understanding of our work made this collaboration seamless and rewarding.”

Making HOPE accessible

HOPE is designed to offer convenient, round-the-clock help by answering common questions in both English and Spanish and guiding users toward the right WestCare services.

A digital front door to support, HOPE will complement human care without replacing it. The AI-powered chatbot is designed to help users explore programs such as outpatient substance use treatment, sober living, inpatient treatment support, domestic violence shelter resources and other critical support. It can also identify risk language and escalate users to crisis lines and other appropriate safety resources when a situation requires immediate intervention.

How it works

Arunachalam said the formula for the collaboration’s success was twofold: WestCare brought deep operational knowledge of the communities it serves, the realities of trauma-informed care and a clear understanding of what people need when they are reaching out in vulnerable moments. AICIC brought design thinking, rapid prototyping and technical expertise to the table that helped translate those needs into a working solution.

In one month, ASU student interns built the prototype for HOPE, which uses services such as Amazon Bedrock for conversational responses, AWS Lambda for backend logic and escalation triggers, DynamoDB for structured intake and session context, Amazon S3 for curated knowledge content, and API Gateway for secure connections between the frontend and backend services. The prototype centers on three core flows: client chat, safety escalation and routing, and knowledge and content management.

HOPE that’s safe, smart and scalable

HOPE distinguishes itself through several innovative features.

It operates exclusively on WestCare’s proprietary data rather than internet sources, greatly reducing the risk of hallucinations and giving users greater control over information accuracy.

Lahari Shakthi Arun, an undergraduate student in graphic information technology in the Ira A. Fulton Schools of Engineering, developed the chatbot’s personalized questionnaire interface that guides users to relevant services based on their specific needs. “The chatbot remembers how users answered the questionnaire and is very efficient,” she said.

Sahajpreet Singh Khasria, an undergraduate student in computer science in the Ira A. Fulton Schools of Engineering, developed built-in response protocols. “We had to modify its nature to be as helpful as possible to users,” he said. “If it detects any violent or harmful behavior, its responses are designed to calm the person down and direct them to the person they need to talk to for help.”

The open-source technology features a modular architecture that supports multilingual expansion, location-aware service recommendations and future website integration. It incorporates robust data privacy protection, content guardrails to prevent off-topic responses and location-based search functionality to identify nearby resources.

“We built it with reusability and scalability in mind,” Arunachalam said.

The HOPE prototype will be available on the upcoming WestCare Arizona website. Afterward will be rounds of measurement, iteration and refinement, including broader multilingual support, deeper integration with intake processes and expansion to additional WestCare programs and web properties. Based in Las Vegas, Nevada, WestCare serves communities across 17 states and four U.S. territories.

“I think it’s great we’re seeing nonprofits leverage this type of technology,” said Colleen Schwab, who leads the AICIC and works with education institutions, government and nonprofits to solve organizational or mission-related challenges with technology. “Younger generations and people who utilize AI in new and creative ways expect to have the same level of service or conversational engagement with chatbots.

“HOPE makes WestCare more approachable and easier to access, and people will get help much faster,” Schwab added.