Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Script-Based Dialog Policy Planning for LLM-Powered Conversational Agents: A Basic Architecture for an "AI Therapist"

Robert Wasenmüller, Kevin Hilbert, Christoph Benzmüller

TL;DR

This work introduces Script-Based Dialog Policy Planning to enable LLM-powered agents to function as a safe, inspectable AI Therapist by coupling proactive conversational capabilities with a deterministic, expert-defined script. It defines a finite set of states and section-level instructions, enabling explicit transitions and traceability while allowing natural dialogue. Two implementation variants are evaluated using synthetic LLM-simulated patients, revealing that a single-actor variant is more efficient, whereas a multi-actor variant adheres more closely to the script, at the cost of conversational coherence. The results establish feasibility and provide a baseline for safety, evaluation, and future development of AI-driven behavioral health agents, with roadmap for richer scripts and human-in-the-loop validation.

Abstract

Large Language Model (LLM)-Powered Conversational Agents have the potential to provide users with scaled behavioral healthcare support, and potentially even deliver full-scale "AI therapy'" in the future. While such agents can already conduct fluent and proactive emotional support conversations, they inherently lack the ability to (a) consistently and reliably act by predefined rules to align their conversation with an overarching therapeutic concept and (b) make their decision paths inspectable for risk management and clinical evaluation -- both essential requirements for an "AI Therapist". In this work, we introduce a novel paradigm for dialog policy planning in conversational agents enabling them to (a) act according to an expert-written "script" that outlines the therapeutic approach and (b) explicitly transition through a finite set of states over the course of the conversation. The script acts as a deterministic component, constraining the LLM's behavior in desirable ways and establishing a basic architecture for an AI Therapist. We implement two variants of Script-Based Dialog Policy Planning using different prompting techniques and synthesize a total of 100 conversations with LLM-simulated patients. The results demonstrate the feasibility of this new technology and provide insights into the efficiency and effectiveness of different implementation variants.

Script-Based Dialog Policy Planning for LLM-Powered Conversational Agents: A Basic Architecture for an "AI Therapist"

TL;DR

This work introduces Script-Based Dialog Policy Planning to enable LLM-powered agents to function as a safe, inspectable AI Therapist by coupling proactive conversational capabilities with a deterministic, expert-defined script. It defines a finite set of states and section-level instructions, enabling explicit transitions and traceability while allowing natural dialogue. Two implementation variants are evaluated using synthetic LLM-simulated patients, revealing that a single-actor variant is more efficient, whereas a multi-actor variant adheres more closely to the script, at the cost of conversational coherence. The results establish feasibility and provide a baseline for safety, evaluation, and future development of AI-driven behavioral health agents, with roadmap for richer scripts and human-in-the-loop validation.

Abstract

Large Language Model (LLM)-Powered Conversational Agents have the potential to provide users with scaled behavioral healthcare support, and potentially even deliver full-scale "AI therapy'" in the future. While such agents can already conduct fluent and proactive emotional support conversations, they inherently lack the ability to (a) consistently and reliably act by predefined rules to align their conversation with an overarching therapeutic concept and (b) make their decision paths inspectable for risk management and clinical evaluation -- both essential requirements for an "AI Therapist". In this work, we introduce a novel paradigm for dialog policy planning in conversational agents enabling them to (a) act according to an expert-written "script" that outlines the therapeutic approach and (b) explicitly transition through a finite set of states over the course of the conversation. The script acts as a deterministic component, constraining the LLM's behavior in desirable ways and establishing a basic architecture for an AI Therapist. We implement two variants of Script-Based Dialog Policy Planning using different prompting techniques and synthesize a total of 100 conversations with LLM-simulated patients. The results demonstrate the feasibility of this new technology and provide insights into the efficiency and effectiveness of different implementation variants.

Paper Structure

This paper contains 26 sections, 5 figures, 1 table.

Table of Contents

  1. Introduction
  2. Method
  3. Requirements for an AI Therapist
  4. LLM-Based Proactive Conversational Agents
  5. Implementation Approaches
  6. Proactive Dialog Policy Planning
  7. Proactive Prompting Techniques
  8. Script-Based Dialog Policy Planning
  9. Script for Policy Planning
  10. Injecting Instructions via Prompts
  11. Section-Level Instructions
  12. Completion Assessment Step
  13. Experimental Setup
  14. Exemplary Script
  15. Implementation Variants
  16. ...and 11 more sections

Figures (5)

  • Figure 1: Proactive prompt-based dialog policy planning, version with an external Actor LLM: Steps on each conversation turn.
  • Figure 2: Structure of the exemplary script with multi-task sections incl. state transitioning rules ( "proceed with ..."). The complete script is provided in the technical appendix.
  • Figure 3: Script-Based Dialog Policy Planning, version with multiple LLM actors: Steps on each conversation turn.
  • Figure 4: Variant A: Exemplary dialog.
  • Figure 5: Variant B: Exemplary dialog.