Related Experiment Video
Updated: Jun 19, 2026
08:22
Remotely Supervised Transcranial Direct Current Stimulation: An Update on Safety and Tolerability
Published on: October 7, 2017
8.2K
Safety and User Experience of a Generative Artificial Intelligence Digital Mental Health Intervention: Exploratory
Timothy R Campellone1, Megan Flom1, Robert M Montgomery1
1Woebot Health, San Francisco, CA, United States.
Journal of Medical Internet Research
|May 23, 2025
Summary
Generative AI in digital mental health interventions (DMHIs) shows promise for user engagement and empathy. With proper safety guardrails, generative AI can be safely implemented in DMHIs without compromising user experience.
Area of Science:
- Digital mental health interventions
- Artificial intelligence in healthcare
- Human-computer interaction
Background:
- Generative artificial intelligence (AI) offers potential for enhanced user experiences in digital mental health interventions (DMHIs).
- Uncontrolled AI implementation poses risks of user harm and trust erosion.
- Limited rigorous evaluations existed for safely integrating generative AI into DMHIs prior to this study.
Purpose of the Study:
- To explore and compare user relationship, experience, safety, and technical guardrails of a generative AI-powered DMHI versus a rules-based DMHI.
- To assess the efficacy of technical guardrails in managing user inputs and AI outputs within a DMHI context.
- To evaluate the safety and user experience of generative AI in a controlled digital mental health setting.
Main Methods:
- A 2-week randomized controlled trial (RCT) involving 160 adult participants.
- Participants were assigned to either a generative AI (n=81) or rules-based (n=79) conversational DMHI.
- Measures included user relationship (satisfaction, alliance, empathy accuracy), user experience (engagement, adverse events), and technical guardrail success.
Main Results:
- User relationship and experience metrics were largely similar between generative AI and rules-based DMHIs.
- The generative AI arm demonstrated superior accuracy in detecting and responding with empathy (98% vs. 69%).
- No serious adverse events occurred; technical guardrails were 100% successful, and participant sentiment towards AI improved.
Conclusions:
- Generative AI can be safely and effectively implemented in digital mental health interventions (DMHIs) when appropriate guardrails are in place.
- This study provides a foundational blueprint for developing robust technical and conversational guardrails for safe generative AI deployment in DMHIs.
- The findings suggest generative AI can enhance DMHIs while maintaining a positive user relationship and experience.
Keywords:
AIDMHIRCTartificial intelligencechatbotscontrolled trialsdigital healthdigital mental health interventionexploratorygenerativegenerative AImental healthrandomizedrandomized controlled trialrelationshipsafetysatisfactionuseruser experienceuser relationshipuser safetyuser satisfactionMore Related Videos
Related Concept Videos
01:13Randomized Experiments
The randomization process involves assigning study participants randomly to experimental or control groups based on their probability of being equally assigned. Randomization is meant to eliminate selection bias and balance known and unknown confounding factors so that the control group is similar to the treatment group as much as possible. A computer program and a random number generator can be used to assign participants to groups in a way that minimizes bias.
Simple randomization
Simple...
Simple randomization
Simple...
01:11Blinding
Blinding is a commonly used method of not telling participants which treatment a subject is receiving. Blinding is a critical part of a randomized control trial or RCT. It reduces the bias that affects the results. In an RCT, blinding is used in the form of a placebo. A placebo effect occurs when untreated subjects falsely believe they have received the treatment and report improved symptoms. A placebo or a dummy treatment is administered to subjects to negate the bias caused by such an effect.