# Arachnophobia Exposure Therapy using Experience-driven Procedural Content Generation via Reinforcement Learning (EDPCGRL)

**Authors:** Athar Mahmoudi-Nejad, Matthew Guzdial, Pierre Boulanger  
**arXiv:** [2110.04146](https://arxiv.org/abs/2110.04146)  
**Date:** 2021-10-07  
**Summary page:** https://www.summarizepaper.com/en/paper/2110.04146v1/

## Key points

- Personalized therapy leads to better health outcomes
- Existing approaches for adapting therapeutic content often rely on hand-authored and pre-defined rules
- A new research paper proposes an innovative approach to automatically adapt therapeutic content based on physiological measures
- The study focuses on arachnophobia exposure therapy and introduces a method called experience-driven procedural content generation via reinforcement learning (EDPCGRL)
- Virtual spiders generated through EDPCGRL are used to match the specific needs of each patient
- The initial implementation of the study utilizes virtual or artificial humans due to limited access to real-life scenarios during the pandemic
- The EDPCGRL method quickly adapts to virtual humans with high accuracy compared to existing search-based approaches
- The authors of the paper are Athar Mahmoudi-Nejad, Matthew Guzdial, and Pierre Boulanger
- The research was presented at the 17th AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment 2021 (AIIDE-21)
- The paper provides further details about methodology and findings in eight pages accompanied by three figures
- This research contributes valuable insights into personalized therapy by leveraging physiological measures for automatic adaptation of therapeutic content

## Summary

In the field of personalized therapy, where therapeutic practices are tailored to individual patients, it has been observed that this approach leads to better health outcomes. Traditionally, therapists rely on their training, intuition, and patient feedback to adapt the therapy to each individual. However, existing approaches for automatically adapting therapeutic content to patients often rely on hand-authored and pre-defined rules which may not be applicable or effective for all individuals. To address this limitation, a new research paper proposes an innovative approach to automatically adapt therapeutic content based on physiological measures. The study focuses specifically on arachnophobia exposure therapy and introduces a method called experience-driven procedural content generation via reinforcement learning (EDPCGRL). This method utilizes virtual spiders generated through EDPCGRL to match the specific needs of each patient. Due to the ongoing pandemic and limited access to real-life scenarios, the initial implementation of the study utilizes virtual or artificial humans based on prior research in arachnophobia psychology. The results show that the EDPCGRL method is able to quickly adapt to these virtual humans with high accuracy compared to existing search-based EDPCG approaches. The authors of this paper are Athar Mahmoudi-Nejad, Matthew Guzdial and Pierre Boulanger. Their work was presented at the 17th AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment 2021 (AIIDE-21). The paper provides further details about their methodology and findings in eight pages accompanied by three figures. Overall, this research contributes valuable insights into personalized therapy by proposing a novel approach that leverages physiological measures for automatic adaptation of therapeutic content. By utilizing EDPCGRL and virtual spiders in arachnophobia exposure therapy, this study demonstrates promising results in terms of accuracy and efficiency compared to existing methods.