Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Adaptive LLM Agents: Toward Personalized Empathetic Care

Priyanka Singh, Sebastian Von Mammen

TL;DR

The paper addresses how to personalize mental-health support using AIS-stratified LLM agents that operate in three profiles (L, M, H) and adapt over time via longitudinal AIS signals. It combines clinically grounded frameworks with a design-fiction approach, treating the AIS-based architecture as a diegetic near-future prototype embedded in home, work, and daily-life contexts. A four-phase methodology (assessment, multi-profile deployment, adaptive feedback, design-fiction exploration) underpins initialization, adaptation, and safety rails, detailing how agent stance evolves and how safety is maintained. The work highlights governance, autonomy, and societal implications of widespread AIS-guided companionship, arguing for responsible design and empirical validation to complement speculative deployment analyses.

Abstract

Current mental-health conversational systems are usually based on fixed, generic dialogue patterns. This paper proposes an adaptive framework based on large language models that aims to personalize therapeutic interaction according to a user's psychological state, quantified with the Acceptance of Illness Scale (AIS). The framework defines three specialized agents, L, M, and H, each linked to a different level of illness acceptance, and adjusts conversational behavior over time using continuous feedback signals. The AIS-stratified architecture is treated as a diegetic prototype placed in a plausible near-future setting and examined through the method of design fiction. By embedding the architecture in narrative scenarios, the study explores how such agents might influence access to care and therapeutic relationship. The goal is to show how clinically informed personalization, technical feasibility, and speculative scenario analysis can together inform the responsible design of LLM-based companions for mental-health support.

Adaptive LLM Agents: Toward Personalized Empathetic Care

TL;DR

The paper addresses how to personalize mental-health support using AIS-stratified LLM agents that operate in three profiles (L, M, H) and adapt over time via longitudinal AIS signals. It combines clinically grounded frameworks with a design-fiction approach, treating the AIS-based architecture as a diegetic near-future prototype embedded in home, work, and daily-life contexts. A four-phase methodology (assessment, multi-profile deployment, adaptive feedback, design-fiction exploration) underpins initialization, adaptation, and safety rails, detailing how agent stance evolves and how safety is maintained. The work highlights governance, autonomy, and societal implications of widespread AIS-guided companionship, arguing for responsible design and empirical validation to complement speculative deployment analyses.

Abstract

Current mental-health conversational systems are usually based on fixed, generic dialogue patterns. This paper proposes an adaptive framework based on large language models that aims to personalize therapeutic interaction according to a user's psychological state, quantified with the Acceptance of Illness Scale (AIS). The framework defines three specialized agents, L, M, and H, each linked to a different level of illness acceptance, and adjusts conversational behavior over time using continuous feedback signals. The AIS-stratified architecture is treated as a diegetic prototype placed in a plausible near-future setting and examined through the method of design fiction. By embedding the architecture in narrative scenarios, the study explores how such agents might influence access to care and therapeutic relationship. The goal is to show how clinically informed personalization, technical feasibility, and speculative scenario analysis can together inform the responsible design of LLM-based companions for mental-health support.

Paper Structure

This paper contains 16 sections, 1 figure, 1 table.

Table of Contents

  1. Introduction
  2. Background
  3. Clinical grounding of AI-supported therapy
  4. Digital interventions and the treatment gap
  5. From generic chatbots to structured conversational agents
  6. Acceptance of illness as a personalization axis
  7. Design fiction as a lens on future deployment
  8. Methodology
  9. Agent design, goals, and adaptation
  10. Implementation as diegetic prototype
  11. Scenario construction and design-fiction lens
  12. Scenarios of AIS-Stratified Care Futures
  13. Hypothesis 1: Low acceptance at home
  14. Hypothesis 2: Moderate acceptance at work
  15. Hypothesis 3: High acceptance in daily life
  16. ...and 1 more sections

Figures (1)

  • Figure 1: Therapeutic interaction model with adaptive agent acceptance. An initial AIS assessment routes the user to L, M, or H stances. Longitudinal signals feed an adaptive loop that blends agent behaviour while enforcing safety rails and supporting oversight.