# US 2026/0311844 A1 · Method and system for mental health support

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2026/0311844 A1
- Document ID: US20260311844A1
- Status: Published application
- Kind: A1
- Application number: 19/078,342
- Filed: 2025-03-13
- Published: 2026-10-08
- Earliest claimed: 2025-03-13
- Assignees: Innsightful, Inc.
- Inventors: Farzad Ehsani; Demitrios Master; Noel Torres
- Web page: https://patentgrep.com/patent/US20260311844A1

## CPC

- G16H 20/70: relating to mental therapies, e.g. psychological therapy or autogenous training
- G16H: HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA

## Abstract

A system, method, and computer program product for automated mental health coaching utilizing a multi-agent architecture powered by large language models. The system employs a hierarchical framework of specialized agents that collaboratively deliver personalized, safe, and effective coaching interactions. A master agent manages overall dialogue flow and maintains conversation history while coordinating with a classifier that selects coaching templates from a template database based on real-time conversational context, ensuring structured yet flexible therapeutic guidance. An interaction agent generates contextually appropriate responses informed by the active template and dialogue history. A verification agent evaluates all generated responses for relevance, methodological consistency, safety, and appropriateness, and a suggestion agent produces alternative responses when verification criteria are not met. A trigger agent continuously monitors dialogue for indicators of self-harm or crisis situations, enabling rapid detection and escalation. Together, these components provide a robust, adaptive platform for automated mental health support.

## Claims

### Claim 1

A system for providing automated mental health coaching, comprising:
a master agent configured to manage dialogue flow and maintain dialogue history;
a template database storing coaching interaction templates;
a classifier configured to select active templates from the template database based on dialogue context;
an interaction agent configured to generate responses based on active templates and dialogue history;
a verification agent configured to evaluate generated responses across multiple criteria including relevance, linguistic coherence, emotional appropriateness, and context alignment;
a suggestion agent configured to refine responses through adaptive learning techniques when responses fail verification;
a trigger agent configured to monitor dialogue for crisis indicators;
an automated refinement mechanism configured to:
evaluate past responses in real-time;
optimize prompt engineering through adaptive learning;
implement probabilistic response scoring; and
evolve templates based on interaction feedback; and
feedback loops configured to enable continuous system adaptation and improvement based on interaction patterns and outcomes.

### Claim 2 (depends on 1)

The system of claim 1, wherein the master agent comprises a dialogue history storage module, a template tracking module, and a feedback processing module that enables dynamic template updates based on conversation flow.

### Claim 3 (depends on 1)

The system of claim 1, wherein each coaching interaction template in the template database comprises structured interaction steps, predefined response patterns, safety guidelines, and success metrics for verifying response quality and effectiveness.

### Claim 4 (depends on 1)

The system of claim 1, wherein the verification agent is configured to assess responses based on predefined criteria stored within each template to ensure responses align with therapeutic objectives while maintaining consistent quality standards.

### Claim 5 (depends on 1)

The system of claim 1, wherein the trigger agent maintains a database of crisis-indicating keywords and performs real-time analysis of user inputs to detect potential self-harm or crisis situations.

### Claim 6 (depends on 1)

The system of claim 1, wherein the system implements progress tracking capabilities that generate periodic assessments, monitor user engagement metrics, and maintain records of successful intervention patterns.

### Claim 7

A method for providing automated mental health coaching, comprising:
receiving user input through a text-based interface;
maintaining dialogue history of interactions;
selecting an appropriate coaching template from a template database based on dialogue context ;
generating responses based on the selected template and dialogue history;
verifying generated responses across multiple criteria including relevance, linguistic coherence, emotional appropriateness, and context alignment ;
refining responses through adaptive learning techniques when responses fail verification;
monitoring dialogue for crisis indicators;
implementing an automated refinement mechanism that evaluates past responses, optimizes prompt engineering, and evolves templates based on interaction feedback; and
enabling continuous system adaptation and improvement through feedback loops based on interaction patterns and outcomes.

### Claim 8 (depends on 7)

The method of claim 7, wherein selecting an appropriate coaching template comprises analyzing recent dialogue history, matching conversation patterns against template parameters, calculating matching scores for potential templates, and selecting the most appropriate template based on matching scores.

### Claim 9 (depends on 7)

The method of claim 7, wherein generating responses comprises considering the active template, complete dialogue history, user input patterns, and coaching objectives defined within the template.

### Claim 10 (depends on 7)

The method of claim 7, wherein verifying generated responses comprises assessing relevance to the active template, adherence to coaching methodology steps, safety considerations, and quality metrics including coherence and appropriateness.

### Claim 11 (depends on 7)

The method of claim 7, wherein monitoring dialogue for crisis indicators comprises analyzing user inputs for predefined keywords and patterns indicating potential self-harm or crisis situations.

### Claim 12 (depends on 7)

The method of claim 7, further comprising tracking progress by generating periodic assessments, monitoring user engagement metrics, and maintaining records of successful intervention patterns.

### Claim 13

A non-transitory computer-readable storage medium containing executable code for providing automated mental health coaching, the executable code when executed by a processor causes the processor to:
receive user input through a text-based interface;
maintain dialogue history of interactions;
select an appropriate coaching template from a template database based on dialogue context;
generate responses based on the selected template and dialogue history;
verify generated responses across multiple criteria including relevance, linguistic coherence, emotional appropriateness, and context alignment;
refine responses through adaptive learning techniques when responses fail verification;
monitor dialogue for crisis indicators;
implement an automated refinement mechanism that evaluates past responses, optimizes prompt engineering, and evolves templates based on interaction feedback; and
enable continuous system adaptation and improvement through feedback loops based on interaction patterns and outcomes.

### Claim 14 (depends on 13)

The non-transitory computer-readable storage medium of claim 13, wherein the executable code for selecting an appropriate coaching template comprises code for analyzing recent dialogue history, matching conversation patterns against template parameters, calculating matching scores for potential templates, and selecting the most appropriate template based on matching scores.

### Claim 15 (depends on 13)

The non-transitory computer-readable storage medium of claim 13, wherein the executable code for generating responses comprises code for considering the active template, complete dialogue history, user input patterns, and coaching objectives defined within the template.

### Claim 16 (depends on 13)

The non-transitory computer-readable storage medium of claim 13, wherein the executable code for verifying generated responses comprises code for assessing relevance to the active template, adherence to coaching methodology steps, safety considerations, and quality metrics including coherence and appropriateness.

### Claim 17 (depends on 13)

The non-transitory computer-readable storage medium of claim 13, wherein the executable code for monitoring dialogue for crisis indicators comprises code for analyzing user inputs for predefined keywords and patterns indicating potential self-harm or crisis situations.

### Claim 18 (depends on 13)

The non-transitory computer-readable storage medium of claim 13, wherein the executable code further comprises code for tracking progress by generating periodic assessments, monitoring user engagement metrics, and maintaining records of successful intervention patterns.

### Claim 19 (depends on 13)

The non-transitory computer-readable storage medium of claim 13, wherein the executable code for implementing an automated refinement mechanism comprises code for dynamically adjusting response generation parameters based on historical effectiveness metrics.
