Abstract
In line with statistical predictions, the age-correlated condition of dementia may become a major societal challenge in the 21st century.
As technology-supported reminiscence therapy is a potentially effective way to maintain the well-being of people with dementia, we propose a reminiscence recommender system that aims to lower the caregiver burden and allow for the efficient conduction of individually tailored reminiscence sessions.
This paper describes the underlying technologies of the MemoRec system as well as the promising results of the preliminary study.
Access provided by CONRICYT-eBooks. Download conference paper PDF
Similar content being viewed by others
Keywords
1 Introduction
With the worlds’ population growing ever older, the age-correlated condition of dementia will presumably be one of the key health-related challenges of the 21st century. As statistical institutions predict a significant rise of the elderly population, many citizens are also likely to develop dementia symptoms [1].
In this light, interventions for persons with dementia (PwD) focusing on well-being – such as activating pleasant memories in so-called reminiscence therapy (RT) sessions (with analogue or digital content) – are regarded as efficient [2, 3] and good practice in line with human-centered care for PwD [4].
The MemoRec assistive reminiscence system aims to enhance – the highly sought-after [5] – technology-supported RT approach by providing caregivers with an automatic tool that helps them to conduct spontaneous RT sessions tailored to the personal life themes of the participating PwD. In essence, MemoRec uses a specific recommender algorithm to match the PwD’s individual biographies – which were digitized and transferred into a graph database in a previous step – with potentially relevant audio-visual content (audio, video, text) from a dedicated content pool. It then presents the ranked results to the caregiver in an easy to use front-end. Based on that, the caregiver should easily be able to trigger reminiscence by showing appropriate content to the PwD, regardless of knowing the personal biography of the PwD or not.
2 State of the Art
Originating back in the late 1950s, the theory of psychosocial development postulates a struggle between “ego integrity” versus “despair” in elderly people. In order to turn this “crisis” to a favorable outcome, a person has to feel a sense of accomplishment pertaining to his or her past life [6]. Evidently, affirmative “life-reviewing” memories are an integral part of this well-being-fostering process [7]. For PwD, the creation of new memories is impaired, but the recall of autobiographical memories – especially from a specific time frame – still works to a certain extent [8]. Besides the correct time period, RT contents have to adhere to the personal life themes of the PwD (e.g. profession, birth and living places, interests, hobbies, etc.) and their individual significance in order to trigger reminiscence [9].
On the technological side, recommender systems [10] are frequently used in search engines (e.g. Google), social IT (e.g. social networks, dating portals, movie databases), commercial IT (e.g. web shops) or content-providing systems (e.g. streaming portals), in which users seek a filtered, efficient and effective prediction regarding items in a vast information pool and for a certain context. Depending on the prediction type as well as other parameters (e.g. the dataset size), several techniques can be implemented independently or in combination to create an appropriate “user-item-recommendation” function. Commonly used state-of-the-art algorithms are the dynamic learning approaches of “collaborative” filtering and “content-based” filtering or the more static domain-specific “knowledge-based” filtering [11].
Current multimedia RT approaches, e.g. the CIRCA project [12], put a heavy focus on different devices and different forms of interaction, but show a lack of automatic – potentially workload-reducing – recommendation of the reminiscence content itself. Albeit digital RT content seems to be well-suited for recommender systems, caregivers have to choose or even compose the contents for the session manually in most cases. As proof of concept, one of the few recommender systems for RT content, “REMPAD”, uses content-based recommendation with overall positive results [13]. Nonetheless, it does not use different media formats – a gap that MemoRec aims to fill.
3 System Description
As part of the multimedia RT project InterMem [14, 15], the life theme based recommender system MemoRec was conceived as a component of an interactive RT system. Following the identification of a life theme ontology (e.g. “profession”, “home”, “language”, “interests”) extracted from 40+ PwD biographies, a first iteration of the test system was developed in a user-centered design approach.
At its core, MemoRec not only provides a recommendation back-end, but also a “control” front-end for the caregiver and a RT “presentation” front-end for the PwD. Both front-ends are implemented using the Ionic web-app framework. The back-end includes the Neo4j graph database – holding the PwD life themes as well as the content information –, a life theme parser, the recommender algorithms themselves and an event-coordination database. The front-end for the caregiver lets him or her choose the highest ranked content items from an ordered list (and also rate their effectiveness afterwards), whereas the RT front-end presents the chosen images to the PwD on a suitable display. The system’s modules are loosely coupled over standardized interfaces (e.g. RESTful, JSON, CSV interfaces) in order to allow for an easy integration of new modules (see Fig. 1).
The main algorithm is a modified personalized page rank (PPR) algorithm [16] working on the life theme holding user-item-graph database. The structure of the graph consists of three parts: user-nodes (PwD), item-nodes (content) and life-theme-nodes – the latter nodes being connected with both user-nodes as well as item-nodes. Life-theme-nodes thus represent the individually weighted interests ranging from “−1” (lack of interest) to “3” (vested interest) of the particular PwD. As the caregiver requests recommendations for a certain PwD, a chosen amount of “random walkers” begin to traverse the graph starting from the particular user-node. Depending on the PwD’s preferences, the walkers are more likely to visit higher-weighted life-theme-nodes and thus more frequently end up at more likely suited item-nodes. The PPR algorithm counts the frequency of the visits to the item-nodes and calculates a ranked list to be shown in the front-end. If positive feedback is given by the caregiver, the particular item-node is being directly connected to the user-node, further raising the probability of frequent visits to this important content-item.
4 Methods and Results of the Initial Feasibility Study
As large display surfaces seem to be effective for activating PwD [17], MemoRec was initially tested using a custom-built 4 × 55-in. “monitor wall” in one of the InterMem dementia care partner institutions. In addition, a Microsoft Surface Pro 12-in. tablet PC served as control front-end for the caregiver. The one-shot case study sample consisted of n = 4 of the institutions’ PwD. The system was “fed” the life themes of each test person and then used by the caregiver as an assistive reminiscence content tool in the ensuing RT sessions (up to 30 min for each PwD).
Every session was directly observed (from a vantage ground outside of the field of vision of the caregiver-PwD-tandem; see Fig. 2) while textual notes were taken for later qualitative data analyses. In addition, MemoRec itself logged the time, content item, rating, as well as the PwD ID. As a final step, interviews assessing the usability and the (subjective) quality of recommendations were conducted with the caregiver.
On one hand, the recommended audio-visual content items seemed to elicit positive feelings and trigger reminiscence in most cases. On the other hand, some recommendations were misleading, particularly regarding the life-theme “home”: although one particular PwD’s home was Bavaria, MemoRec gave a picture from the Black Forest a top three ranking, as the life-theme-node “home” didn’t hold the correct property. Ultimately, the interview yielded a predominantly positive perception from the point of view of the caregiver: MemoRec was described as intuitive as well as easy to use while having a high potential when improved properly.
5 Discussion
In summary, the preliminary study showed that it is feasible to use MemoRec as an assistive RT tool to automatically find well-suited reminiscence content for a given PwD, but for valid test results, further iterations, more objective measures (incl. a baseline condition) and a higher number of test persons are needed (especially regarding the “learning from feedback” feature of the recommender that cannot be tested otherwise).
Further complications (and with that potential points of improvement) are the recommender-specific “cold start problem”, as the number of items in the content pool were not sufficiently high (i < 100) and the “knowledge engineering problem”, as the life theme ontology did not seem to be optimal, as seen in the misleading “home”-related recommendation.
Apart from algorithm-related improvements, an automatic multimedia web content crawler to widen the content pool as well as an automatic tagging system are to be implemented as future modules. Such additions may help the caregivers to further concentrate on the important inter-personal aspects of RT and be less burdened by the time-consuming collection, maintenance and manual selection of reminiscence content.
References
Alzheimer’s Disease International: World Alzheimer Report (2016). https://www.alz.co.uk/research/WorldAlzheimerReport2016.pdf. Accessed 21 Mar 2018
Pinquart, M., Forstmeier, S.: Effects of reminiscence interventions on psychosocial outcomes: a meta-analysis. Aging Ment. Health 16, 541–558 (2012)
Lazar, A., Thompson, H., Demiris, G.: A systematic review of the use of technology for reminiscence therapy. Health Educ. Behav. 41, 51S–61S (2014)
Kitwood, T.M.: Dementia Reconsidered: The Person Comes First. Open University Press, Maidenhead (2012)
Sixsmith, A., Gibson, G., Orpwood, R., Torrington, J.: Developing a technology ‘wish-list’ to enhance the quality of life of people with dementia. Gerontechnology 6, 2–19 (2007)
Erikson, E.H.: Identity and the life cycle: selected papers. Psychol. Issues 1, 1–171 (1959)
Butler, R.N.: The life review: an interpretation of reminiscence in the aged. Psychiatry 26, 65–76 (1963)
Fromholt, P., Larsen, S.F.: Autobiographical memory and life-history narratives in aging and dementia (Alzheimer type). In: Conway, M.A., Rubin, D.C., Spinnler, H., Wagenaar, W.A. (eds.) Theoretical Perspectives on Autobiographical Memory. ASID, vol. 65, pp. 413–426. Springer, Dordrecht (1992). https://doi.org/10.1007/978-94-015-7967-4_24
Subramaniam, P., Woods, B.: The impact of individual reminiscence therapy for people with dementia: systematic review. Expert Rev. Neurother. 12, 545–555 (2012)
Aggarwal, C.C.: Recommender Systems: The Textbook. Springer, New York (2016). https://doi.org/10.1007/978-3-319-29659-3
Adomavicius, G., Tuzhilin, A.: Toward the next generation of recommender systems: a survey of the state-of-the-art and possible extensions. IEEE Trans. Knowl. Data Eng. 17, 734–749 (2005)
Astell, A.J., Alm, N., Gowans, G., Ellis, M.P., Dye, R., Campbell, J., Vaughan, P.: Working with people with dementia to develop technology: the CIRCA and Living in the Moment projects. PSIGE Newsl. 105, 64–69 (2009)
Yang, Y., Caprani, N., Bermingham, A., O’Rourke, J., Collins, R., Gurrin, C., Smeaton, A.F.: Design and field evaluation of REMPAD: a recommender system supporting group reminiscence therapy. In: O’Grady, M.J., Vahdat-Nejad, H., Wolf, K.-H., Dragone, M., Ye, J., Röcker, C., O’Hare, G. (eds.) AmI 2013. CCIS, vol. 413, pp. 13–22. Springer, Cham (2013). https://doi.org/10.1007/978-3-319-04406-4_3
Klein, P., Uhlig, M.: Interactive memories. In: Proceedings of the 9th ACM International Conference on PErvasive Technologies Related to Assistive Environments, PETRA 2016 (2016)
Interactive Memories. http://www.intermem.org. Accessed 21 Mar 2018
Lofgren, P.A.: Efficient algorithms for Personalized PageRank (2015)
Bejan, A., Gündogdu, R., Butz, K., Müller, N., Kunze, C., König, P.: Using multimedia information and communication technology (ICT) to provide added value to reminiscence therapy for people with dementia. Zeitschrift für Gerontologie und Geriatrie 51, 9–15 (2017)
Acknowledgments
We especially thank all of our test persons, colleagues and partners, as well as the Federal Ministry of Education and Research (project InterMem, 16SV7322).
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2018 Springer International Publishing AG, part of Springer Nature
About this paper
Cite this paper
Bejan, A., Plotzky, C., Kunze, C. (2018). MemoRec – Towards a Life-Theme-Based Reminiscence Content Recommendation System for People with Dementia. In: Miesenberger, K., Kouroupetroglou, G. (eds) Computers Helping People with Special Needs. ICCHP 2018. Lecture Notes in Computer Science(), vol 10896. Springer, Cham. https://doi.org/10.1007/978-3-319-94277-3_79
Download citation
DOI: https://doi.org/10.1007/978-3-319-94277-3_79
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-319-94276-6
Online ISBN: 978-3-319-94277-3
eBook Packages: Computer ScienceComputer Science (R0)