DOORS AI Training with EasterSeals, Houston 

On March 25, our team members Anh Nguyen and Mason Granof traveled to Houston, Texas to deliver training on the relationship between mental health and AI with EasterSeals. Across two sessions, groups of veterans, healthcare providers, and peer-support professionals explored how Artificial Intelligence is reshaping the mental health landscape, and what both the risks and benefits look like in practice. 

This training is part of our Digital Outreach for Obtaining Resources and Skills (DOORS) program, which has been providing digital literacy training to both patients and providers since 2018. As AI tools become more embedded in daily life, DOORS has expanded to address the new questions providers are asking about technology in the therapeutic relationship. Our provider curriculum is designed to equip professionals with the foundational AI literacy they need to think critically about AI's role in their patients' mental health care.

Participants started by building a shared vocabulary around how generative AI actually works, including probabilistic text generation, sycophancy, hallucinations, safety guardrails, and benchmarking. With that foundation in place, the conversation turned to what providers need to watch for in clinical settings. We walked through risk mechanisms like emotional dependency on AI chatbots, patients receiving inaccurate health information, and the reinforcement of harmful thought patterns. From there, we discussed how to recognize when a patient's AI use becomes clinically relevant, how to document it in a structured and trackable way, and how to respond with a harm-reduction approach that meets patients where they are rather than simply telling them to stop.

Our discussion throughout the training reflected the ongoing debates within the field, where the rapid proliferation of this technology has presented novel opportunities and risks. One theme that came up repeatedly was debates about whether this technology could fill in service gaps in existing mental healthcare infrastructure. Some participants insisted that this technology would be extremely useful, particularly among more isolated rural communities, while others expressed concerns about feasibility and safety of applying this technology to mental health contexts. Other participants shared instances where their patients displayed increased cognitive rigidity following extended chatbot use, and expressed concern about potential long-term effects of the technology.