Related Experiment Video
Updated: Aug 21, 2026

Virtual Agent for Real-Time Motivational Interviewing by Integrating Adaptive Nonverbal Behavior and Language Models
Published on: December 23, 2025
Engagement, Satisfaction, and Therapeutic Alliance With an AI Conversational Agent for Parents: Quantitative
Karin Mostovoy1, Blanca S Pineda1,2, Arjun Bharat1
1Department of Psychology, Palo Alto University, 1791 Arastradero Rd, Palo Alto, CA, 94304, United States, 1 (800) 818-6136.
Background:
Parent management training (PMT) is an evidence-based intervention for addressing child behavioral difficulties; however, caregivers often need additional guidance when implementing skills in daily life. Pat is an AI conversational agent designed to augment a therapist-led PMT program by providing caregivers with real-time guidance, reinforcement, and answers to parenting questions between sessions.
Objective:
This study explored caregivers' engagement with Pat, satisfaction, and the therapeutic alliance formed between caregivers and Pat during online group PMT.
Methods:
Data from 88 caregivers of children aged 3 to 14 years participating in an online PMT program were analyzed at 3 time points (wk 4, 8, and 12). Engagement with Pat was measured through the number of messages exchanged with Pat, modules completed, in-the-moment support chats, and days the platform was used. Satisfaction with the intervention was measured using the Net Promoter Score and the Therapy Attitude Inventory. The therapeutic alliance between caregivers and Pat was assessed using the Working Alliance Inventory-Short Form Revised.
Results:
Caregivers exchanged a high number of messages with Pat (mean 383.09, SD 185.64 by wk 4 and mean 726.47, SD 405.41 by wk 8) and completed multiple required and optional modules (mean 5.41, SD 1.93 at wk 4; mean 10.49, SD 9.18 at wk 8). Net Promoter Scores remained high and stable across the study (wk 4=77.8, wk 8=76.7, and wk 12=75.0), with no significant change over time (F2,114=0.10; P=.91). Therapy attitudes increased significantly from week 4 (mean 4.22, SD 0.47) to week 8 (mean 4.49, SD 0.37) and remained stable through week 12 (mean 4.49, SD 0.48, F2, 114=5.69; P=.004). Therapeutic alliance ratings with Pat were consistently high across time points (mean 4.10, SD 0.63 at wk 4; mean 4.10, SD 0.55 at wk 8; and mean 4.04, SD 0.63 at wk 12), with no significant differences across weeks (F2, 114=0.182; P=.83).
Conclusions:
These findings show that caregivers meaningfully engaged with Pat, remained highly satisfied across time, established a strong therapeutic alliance, and held positive therapy attitudes. Overall, the hybrid group model integrating human therapists with Pat may represent a promising and efficient strategy for enhancing engagement and providing scalable and real-time support to parents.