As opposed to those used in the reminiscence therapies, this method does not rely on self-report and tracks objective behaviors, such as speech in everyday life, with no elicitation of reminiscence events. In 2017, Demiray et al were the first to examine reminiscence using a naturalistic observation method to enable investigating reminiscence in the real world. Studies with a focus on automated reminiscence therapy, in contrast, typically aim at eliciting reminiscence from users (eg, with the remote assistance of a therapist), rather than collecting spontaneous reminiscence events during everyday life settings. Moreover, the self-report method provides researchers only with the average frequency of an activity over a certain period of time. The use of self-reporting has potential limitations, such as recall biases, response styles, demand characteristics, social desirability, and limitations to introspection. The study of reminiscence in older adults has traditionally focused on (1) reflective self-reporting and life reviews and (2) automated reminiscence therapy, that is, “a nonpharmacological intervention involving the prompting of past memories, for therapeutic benefits, such as the facilitation of social interactions or the increase of self-esteem”, especially for dementia patients. Naturalistic Observation as a New Approach to the Study of Reminiscence Researchers who study aging emphasize a cognitive activity in old age such as reminiscence to be an essential part of healthy aging in fact, the use of memory interventions and reminiscence in therapies for older adults is common, emphasizing the relation between self-positive functions of reminiscence and well-being, according to Webster and Cappeliez’s tripartite model of reminiscence.
Many disciplines are interested in the study of reminiscence, such as nursing, social work, education, theology, psychology, and gerontology, with a strong focus on reminiscence in the context of aging. Reminiscing can be a volitional or nonvolitional process recollecting memories, an activity that may be private or involve others in the latter case, we refer to social reminiscence. Recalling or sharing valuable life experiences with third parties can support decision-making, bonding with others, and self-understanding. These experiences may refer to specific events (eg, the first kiss), repeated ones (eg, going to the gym every Friday), extended ones (eg, a Christmas trip), or even long periods of life (eg, living in a foreign country for some years). Reminiscence is the “naturally occurring act of thinking about or telling others about personally meaningful past experiences”. In this study, we embrace the healthy aging model by examining one such real-life activity: reminiscence. The novel WHO model encourages aging researchers to step outside the lab and into the real world to examine activities in everyday life, aiming to empower individuals to observe, measure, and take earlier action for their own health. Activities represent the interaction between person characteristics and environments they are understudied in traditional aging research. This model emphasizes the interplay of personal characteristics (eg, abilities), environments, and their interactions in producing functioning. With its first world report on aging and health, the World Health Organization (WHO) promoted a global paradigm shift in aging research by moving from a disease-focused model to a dynamic, contextualized, person-focused model of “healthy aging”. We computed the area under the curve (AUC), the average precision (AP), precision, recall, as well as F1 score and specificity measures on the test data, for all combinations of NLP features, algorithms, and learning strategies.
For each learning strategy, we performed cross-validation on a random sample of the training data set of transcripts.
Due to class imbalance in the data, we introduced three learning strategies: (1) class-weighted learning, (2) a meta-classifier consisting of a voting ensemble, and (3) data augmentation with the Synthetic Minority Oversampling Technique (SMOTE) algorithm. The data set comprises 2214 transcripts, including 109 transcripts with reminiscence.
The methods in this study comprise (1) collecting and coding of transcripts of older adults’ conversations in German, (2) preprocessing transcripts to generate NLP features (bag-of-words models, part-of-speech tags, pretrained German word embeddings), and (3) training machine learning models to detect reminiscence using random forests, support vector machines, and adaptive and extreme gradient boosting algorithms.