Abstract

Digital persuasive systems that rely on static, non-personalized reinforcement (reminders, rule-based tracking) frequently fail to sustain engagement once user motivation declines, particularly under acute stress. This paper presents the object-oriented architecture of a Nostalgic Persuasive System, an adaptive engine that detects a user's emotional and stress state from reflective journaling text and responds with a personalized, generationally-targeted nostalgic media recommendation. The architecture integrates six functional layers: data acquisition, model training/preprocessing, text-based emotion classification (a fine-tuned DistilRoBERTa model), text-based stress detection (a fine-tuned RoBERTa model), a dual content-recommendation layer (LightFM for movies, content-based vector similarity for songs), and a contextual-bandit adaptive-learning layer (LinUCB) that refines recommendations from user feedback. The system's functional/non-functional requirements, class structure, and behavioral design are detailed following Design Science Research guidelines. The architecture was implemented and deployed as a live web application, generating 268 real interaction logs over a 7-day field evaluation with 49 active participants, showing substantially higher habit adherence in the personalized condition (90.1%) than in a non-personalized control (41.3%). This paper contributes a reusable architectural blueprint for emotion-aware, adaptive persuasive systems.

References

• H. Oinas-Kukkonen and M. Harjumaa, “Persuasive systems design: Key issues, process model, and system features,” Communications of the Association for Information Systems, vol. 24, no. 1, Article 28, 2009.
• T. Alahäivälä, H. Oinas-Kukkonen, and T. Jokelainen, “Software architecture design for health BCSS: Case Onnikka,” in Persuasive Technology, LNCS vol. 7822, Springer, pp. 3–14, 2013.
• A. McGowan, S. Sittig, D. Bourrie, R. Benton, and S. Iyengar, “The intersection of persuasive system design and personalization in mobile health: Statistical evaluation,” JMIR mHealth and uHealth, vol. 10, no. 9, e40576, 2022.
• K. Zhang, Y. Jiang, K. Xi, and N. Sun, “Application of persuasive system design in mobile health interventions for chronic disease management: A mini review,” Frontiers in Public Health, vol. 13, 1718371, 2025.
• P.G. Kidman, R.G. Curtis, A. Watson, and C.A. Maher, “When and why adults abandon lifestyle behavior and mental health mobile apps: Scoping review,” Journal of Medical Internet Research, vol. 26, e56897, 2024.
• A. Dennis, B.H. Wixom, and D. Tegarden, Systems Analysis and Design: An Object-Oriented Approach with UML, 6th ed., Wiley, 2020.
• M. Kula, “Metadata embeddings for user and item cold-start recommendations,” arXiv:1507.08439, 2015.
• A.I. Schein, A. Popescul, L.H. Ungar, and D.M. Pennock, “Methods and metrics for cold-start recommendations,” in Proc. 25th ACM SIGIR Conf., pp. 253–260, 2002.
• P. Lops, M. de Gemmis, and G. Semeraro, “Content-based recommender systems: State of the art and trends,” in Recommender Systems Handbook, Springer, pp. 73–105, 2011.
• P. Auer, N. Cesa-Bianchi, and P. Fischer, “Finite-time analysis of the multiarmed bandit problem,” Machine Learning, vol. 47, no. 2–3, pp. 235–256, 2002.
• L. Li, W. Chu, J. Langford, and R.E. Schapire, “A contextual-bandit approach to personalized news article recommendation,” in Proc. 19th Intl. Conf. World Wide Web (WWW '10), ACM, pp. 661–670, 2010.
• Y. Liu, M. Ott, N. Goyal, J. Du, M. Joshi, D. Chen, O. Levy, M. Lewis, L. Zettlemoyer, and V. Stoyanov, “RoBERTa: A robustly optimized BERT pretraining approach,” arXiv:1907.11692, 2019.
• V. Sanh, L. Debut, J. Chaumond, and T. Wolf, “DistilBERT, a distilled version of BERT: Smaller, faster, cheaper and lighter,” arXiv:1910.01108, 2019.
• R.W. Picard, Affective Computing, MIT Press, 1997.
• C. Sedikides and T. Wildschut, “The sociality of personal and collective nostalgia,” European Review of Social Psychology, vol. 30, no. 1, pp. 123–173, 2019.