Utilizing Collaborative Filtering to Recommend Opportunities for Positive Afect in daily life

Abstract

In this position paper, we discuss the potential use of Recommender Systems to support interventions used in mental health care. As an example, we highlight an ongoing project which aims to develop a mobile application to increase positive afect and psychological well-being based on the "Three-good-things" exercise. In our system, users are asked to take photos of situations or activities which they felt satisfied about or were grateful for during each day. Afterwards, they are asked to reflect upon the activities and write short messages to describe their experiences. Collaborative Filtering is used to enhance the existing exercise by providing recommendations of nearby activities and locations which have the potential to increase positive afect for users based on their past experiences. This helps further highlight opportunities for users to experience small moments of happiness and gratitude in daily life.

CCS CONCEPTS

• Human-centered computing → Collaborative filtering ; • Applied computing → Health care information systems;

INTRODUCTION

Recent years have seen increasing interest in the use of digital interventions in Mental Health care. Such technologies allow treatments to be delivered directly to users who can access them at a time of their convenience, ofering greater flexibility and control over the therapeutic process and helping reduce the cost of mental health services. This has led to a number of online, mobile and even gamified therapies to be developed and used to treat a variety of mental health conditions such as depression and anxiety. However, it is challenging to create digital interventions which are engaging enough for users in the long term and practical enough to be integrated efectively into clinical practice [ 1 ].

Our own prior experiences of developing Gamified Digital therapies have highlighted some of these challenges [ 7 ]. In particular, one of the main dificulties which we encountered was in providing content which matches the characteristics and interests of each individual user of our digital therapy. For example, when developing the Ready-Set-Goals, a mobile application designed to encourage goal setting for clients undergoing Cognitive Behavioural Therapy (CBT), users found the pre-set goals provided by the application to not be flexible enough to meet their interests and their specific situations. When allowed to set open goals based on diferent life areas (work, education etc.) however, users found it too vague and it was dificult for them to come up with meaningful and suitable goals for the therapy [ 7 ]. To address this issue, we would need to develop a system which is able to suggest appropriate goals for users based on their individual characteristics, real-life situation (time, location etc.) and therapeutic aims. Similar issues have also been encountered in our development of digital interventions for other CBTs. Recommender systems could play a vital role in addressing such issues and overall, we believe that there is much potential for such systems in the domain of Mental healthcare.

While several studies have examined the use of Recommender Systems in health care, such studies tend to focus on recommending items such as medicine, health information or treatments based on the health profile of the patient [ 6 ]. However, few studies have examined the use of Recommender systems within mental health treatments. Therefore, in our research, we are interested in investigating whether such systems could be used to enhance interventions and exercises aimed at improving psychological well-being. Overall, our project aims to: (1) develop novel recommendation algorithms that could be integrated into existing CBT exercises to provide therapeutic content that are relevant to the personal context of each user (2) contribute practical knowledge regarding the efective implementation of such systems in a therapeutic setting, in particular to determine how user acceptance would be efected when recommender systems are integrated with diferent aspects of the therapy and (3) provide empirical evidence regarding the effectiveness of such systems through user experiments that compare the system with traditional therapeutic approaches. 2

RECOMMENDING OPPORTUNITIES FOR

POSITIVE AFFECT IN DAILY LIFE

An example of one of the systems in our project is a mobile application that is being developed based on the "three-good-things" exercise originally proposed by Seligman [ 3 ]. The exercise involves participants keeping a record of three things that went well for them each day of which they were happy about. By reflecting on these activities, participants are able to experience more thankfulness and appreciation in daily life and this has been shown to help increase emotional well-being, happiness, and resilience [ 4 ]. To further enhance the original exercise, our system plans to use collaborative filtering techniques to highlight latent opportunities for users to experience positive afect in daily life (activities which could provide feelings of gratefulness, happiness and satisfaction etc.) based on their past records. The aim is to encourage users to become more aware of previously unrecognized opportunities which could provide them with experiences of positive afect. 3

DEVELOPMENT OF THE SYSTEM

been identified as having potential to provide positive experiences. Data about the nearby Points of Interests would be extracted from OpenStreetMap (for example, the tags which describes the characteristics of the locations such as parks, trees, shrine). A similar technique which we previously used in [ 5 ] would be used to match the concept tag words with diferent POI locations. Users would be alerted to the presence of such locations and the potential activities which could be carried out when they move near there. 4

CONCLUSION

In this paper, we discuss the potential use of Recommender Systems to support Mental Healthcare. As an example, we describe an ongoing project which utilizes collaborative filtering to recommend opportunities for users to experience positive afect in daily life. This system is used as part of a mobile application which we hope to develop and use to improve well-being and happiness for general users and those with low self-esteem.

ACKNOWLEDGMENTS

This work was partially supported by Grant of Innovation Platform for Society 5.0 and JSPS KAKENHI Grant Numbers 17H01822 and 19H04118.

References

[1] Adrian Aguilera . 2015 . Digital technology and mental health interventions: Opportunities and challenges . ( 2015 ). https://doi.org/10.3989/arbor. 2015 .771n1012

[2] Asai Akari et al. 2018 . HappyDB: A Corpus of 100,000 Crowdsourced Happy Moments . In Proceedings of LREC 2018 . European Language Resources Association (ELRA), Miyazaki , Japan.

[3] Martin EP Seligman , Tracy A Steen, Nansook Park, and Christopher Peterson . 2005 . Positive psychology progress: empirical validation of interventions . American psychologist 60 , 5 ( 2005 ), 410 . https://doi.org/10.1037/ 0003 - 066X . 60 .5. 410

[4] J Bryan Sexton and Kathryn C Adair . 2019 . Forty-five good things: a prospective pilot study of the Three Good Things well-being intervention in the USA for healthcare worker emotional exhaustion, depression, work-life balance and happiness . BMJ open 9 , 3 ( 2019 ), e022695 . https://doi.org/10.1136/bmjopen-2018 -022695

[5] Panote Siriaraya , Takumi Kiriu, Yukiko Kawai, and Shinsuke Nakajima . 2018 . Using Open Data to Create Smart Auditory Based Pervasive Game Environments (CHI PLAY '18 Extended Abstracts) . ACM , 7. https://doi.org/10.1145/3270316.3271524

[6] André et al. Valdez . 2016 . Recommender systems for health informatics: stateof-the-art and future perspectives . In Machine Learning for Health Informatics . Springer, 391 - 414 .

[7] Marierose MM van Dooren , Panote Siriaraya , Valentijn Visch, Renske Spijkerman, and Laura Bijkerk . 2019 . Reflections on the design, implementation, and adoption of a gamified eHealth application in youth mental healthcare . Entertainment Computing ( 2019 ), 100305 . https://doi.org/10.1016/j.entcom. 2019 .100305