Can a Recommender System Support Treatment Personalisation in Digital Mental Health Therapy? A Quantitative Feasibility Assessment Using Data from a Behavioural Activation Therapy Game
Robert Lewis, Craig Ferguson, Chelsey Wilks, Noah Jones, Rosalind Picard
CHI'22: ACM Conference on Human Factors in Computing Systems
Session: Late Breaking Work (LBW)
Abstract
Digital mental health interventions often have low user uptake and engagement. Recommender systems are a potential way to improve the digital therapy experience as personalised recommendations may help users to identify therapy tasks they find enjoyable and helpful. Using a dataset containing 23,476 ratings collected from 973 players of a mental health therapy game, this work demonstrates how recommender system algorithms can predict a user's therapy task ratings more accurately than simpler models that predict the average rating for a task, adjusted for the biases of the specific user and specific task. Collaborative filtering algorithms (matrix factorisation and k-nearest neighbour) outperform this baseline with a 12.6-12.9% improvement in mean absolute error (MAE) and context-aware collaborative filtering algorithms (factorisation machines) outperform with a 13.5-13.9% improvement in MAE. These results suggest that recommender systems could be a useful tool for tailoring recommended therapy items to a user based on a combination of their past preferences, current context, and the ratings of similar users. This scalable approach to personalisation -- which does not require a human therapist to always be in-the-loop -- could play an important role in improving engagement and outcomes in digital mental health therapies.
WEB:: [ Ссылка ]
Pre-recorded presentations of CHI 2022
