Instruction prompt tuning (IPT) uses crafted prompts and in-context demonstrations (ICD) to guide large language models (LLMs) in performing previously unseen tasks, transferring knowledge while keeping the LLM mostly frozen. Social media multi-party conversation (MPC) analysis, on the other hand, has achieved remarkable performance on in-domain tasks including exact speaker identification. However, due to several challenges such as the lack of contextual generalization, absence of adequate ICD, and biases in LLMs, significant work remains to be done for quantifying mental health on social media MPC data using out-of-domain (OOD) knowledge transfer. In this article we propose DoRA , a novel dual-encoder demonstration retriever architecture designed to transfer an LLM’s knowledge about MPC modeling for previously unseen depression screening tasks. Our method pairs soft embeddings of MPC prompts with top-ranked ICD claims for depression screening, leveraging IPT-based OOD cross-task transfer using DoRA for the first time. Experiments conducted in zero-shot and few-shot settings across benchmark datasets using multiple LLMs demonstrate significant downstream performance with DoRA . Specifically, there is a 21.54% increase in recall for depressed utterance classification and a 21.11% increase in F 1 score for depressed speaker identification. These results highlight DoRA ’s potential as a screening tool for digital mental health applications.

Effective blood glucose regulation is essential for both diabetic and non-diabetic individuals. Finger-prick methods are inconvenient and painful, while continuous glucose monitoring (CGM) systems are costly and require frequent sensor replacements. This paper proposes an innovative approach for continuous real-time blood glucose level estimation using non-invasive physiological signals and ensemble regression techniques. We observed significant differences in the predictive power of individual features and overall model performance between the diabetic and the non-diabetic groups. The best model for the diabetic group achieved better performance (R2= 0.50) compared to the best model for the non-diabetic group (R2= 0.27). However, when evaluated for clinical accuracy, the non-diabetic models outperformed the diabetic models. The best non -diabetic model had 100% of predictions in the clinically acceptable zones of the Clarke Error Grids (CEG), while the best diabetic model had only 91%. The between-group differences in model performance are likely due to the variations in underlying glucose dynamics between diabetic and non-diabetic individuals. Our findings highlight the necessity of collecting and sharing large multimodal datasets from both diabetic and non-diabetic populations, as well as evaluating glucose prediction models from both mathematical and clinical accuracy.

The prevelence of obstructive sleep apnea (OSA) has significantly increased around the world. The word “apnea” is equivalent to “no breath,” which is the main symptom of this condition. The challenge of detecting sleep disorders from consumer wearable sensors is attracting more and more researchers in the field. Some of them adopted deep learning approaches, while others combined feature engineering with conventional machine learning techniques. A study used at-home oximetry data to estimate the Apnea/Hypopnea Index (AHI) and achieved high intra-class correlation coefficients within 0.889-0.924 [1]. They then classified apnea severity using three typical AHI thresholds: 5 events/hour, 15 events/hour, and 30 events/hour, respectively. Another study using a CNN-based model to classify OSA severity and achieved a macro-accuracy of 80.51 %, a mean sensitivity of 79.95%, a mean specificity of 93.05% [2]. However, compared to clinical equipment, data obtained from wearable devices are low in frequency and resolution, making it challenging to capture all vital information for diagnosing OSA. The development and evaluation of a single model with the ability to reach high diagnostic performance using consumer trackers are still pending. In this study, we aimed to classify the apnea severity from consumer-grade low-granularity oximetry data.

Due to the complexity of traditional fine-tuning methods, prompt-based engineering methods, such as prompt tuning (PT), have recently received increasing attention for casting various downstream tasks into a large language model (LLM) format by prepending soft tunable embeddings to input sequences while keeping the majority of the LLM frozen [1]. On the other hand, the discourse analysis of text-based multi-party conversations (MPCs) has been used to obtain valuable insights like speaker emotion recognition [2]. Although most prior work considers indomain knowledge transfer using PT, much work remains to be done for PT-based out-of-domain (OOD) knowledge transfer such as adapting a LLM's knowledge about MPC modelling to a semantically dissimilar task such as depression detection. This study aims to detect human depression in MPCs by leveraging an LLM's knowledge about MPC modelling.

Parameter-efficient, soft, and prompt-based tuning methods have received increasing attention in various downstream tasks due to the high cost of traditional fine-tuning methods in pre-trained language models (PLM). Prompt tuning (PT) is one such effective mechanism which has achieved remarkable performance in transferring the acquired knowledge of a PLM to perform an unseen task within the same domain using task-specific prompts and informative instructions. Even though most prior work considers in-domain knowledge transfer using PT, much work remains to be done for PT-based out-of-domain knowledge transfer. In this study we propose ProDepDet, a novel framework specifically designed to use a PLM's knowledge about structure and semantic modelling in multi-party conversations to perform the unseen, out-of-domain task of depression detection. To our knowledge, this study is the first attempt to adapt the acquired knowledge of a PLM for out-of-domain task modelling using PT-based cross-task transferability. Experiments on few-shot and full data settings across multiple benchmark datasets demonstrate the superiority of our PT framework in two downstream tasks including depressed utterance classification and depressed speaker identification.

In response to the globalization of education, an increasing number of Japanese universities are adopting English-medium instruction in computing classes. However, instructors often encounter significant barriers, ranging from the students’ insufficient English language competency to the passive learner mentality, and research on EMI-based computing education remains limited. In this short paper, we share the strategies that we implemented both at the course level and at the department level to support students’ learning in EMI-based computing courses. We map our strategies to the three dimensions of transitions posed in the multiple and multidimensional transitions (MMT) theory. While previous studies and common EMI practices have focused primarily on supporting students’ academic and social-cultural transitions, some of our strategies also serve to support students’ psychological transition from high school to university in an EMI context. Our experience has relevance for other educational institutions, particularly those in Japan, where English-taught computing courses are being implemented or expanded.

In this short paper we present a study examining students’ experience using a gamification quiz platform, Kahoot!, in two engineering courses in a private Japanese university. We ran Kahoot! quiz game regularly at the end of each lecture session throughout a semester. Students were encouraged but not forced to engage in discussions with peers during gameplay. We evaluated our approach using an original questionnaire designed to probe students’ opinions of the usability and usefulness of the quiz games as well as their preferred play mode. The results showed that overall students had positive experience playing the Kahoot! quiz games in class, though many of them had no prior experience using Kahoot! or any kind of online quiz platforms before taking the courses we offered. Surprisingly, their interest in the quiz games did not wear off, and they enjoyed the fun and competitive atmosphere throughout the semester. Consistent with findings in prior studies, our analysis showed that many students considered the quiz games useful in helping them to concentrate in class and to develop a deeper understanding into the technical content. To a less degree, students found the quiz games motivating them to learn more about the subject. Students’ preference of the play mode split, which suggests the importance of diversifying the play modes to meet the needs of different types of learners. We reflected over the nature of our quiz questions using the revised version of the Bloom’s taxonomy. Our current design was centered on the ‘Remember’, ‘Understand’, and ‘Apply’ levels of factual and conceptual knowledge. Future studies are needed to expand the quiz game design to support the development of higher-order cognitive skills at the ‘Analyze’, ‘Evaluate’ and ‘Create’ levels, as well as adding additional objective measures for the assessment.

Abstract Sleep studies suggest that exams, jobs, and technologies keep teens up at night, but little research exists to engage teens in reflecting on their own sleep. We designed a set of cards and a web-based app ‘SleepBeta’ to support reflection by inviting teens to ask questions, explore, track, and experiment with sleep and related technology and lifestyle habits. Through card sorting, we invited teens to identify technology and lifestyle habits they wished to track. SleepBeta let teens track various habits and sleep whilst visualizing interrelationships between these data. Twelve teens and 11 parents participated in interviews before and after a 3-week field trial of SleepBeta. Our findings highlighted four distinct modes of reflection: reflection in preparation, reflection in action, reflection upon revisiting data, and reflection through social interaction. We discuss how our findings provide sensitizing concepts that reframe reflection from a post hoc activity with personal data, to an ongoing process that starts before technologies are used to generate data. We highlight design opportunities for scaffolding reflection in preparation, and we reflect on design choices that give teens control over their data.

It is well-known that stress affects sleep quality, suggesting abnormal brain activity during sleep when people are stressed. However, no study has examined where brain processes stress during sleep. This study aims to explore the associations between bedtime stress and the hemodynamics in the prefrontal cortex (PFC) during the first sleep cycle under free-living conditions. Stress biomarkers including salivary cortisol and secretory immunoglobulin A (sIgA) were measured using the SOMA Dual Analyte LFD test kits on the experiment nights between 22:00-23:00 to control the circadian oscillation of the stress-related hormones. Perceived stress level was rated on a 1-10 Likert scale right after the collection of the salivary samples. The hemodynamics of the pre-frontal cortex (PFC) was measured using a wearable functional near-infrared spectroscopy (fNIRS) device. Correlation analysis with statistical test was performed to examine the associations between different stress indicators and a set of time-domain and frequency-domain features derived from the hemodynamic responses. Significantly positive linear correlations were observed in the standard deviation, skewness, and kurtosis between the average concentration change of oxyhemoglobin and that of deoxyhemoglobin in the whole region of interest. Stress was found to correlate to the hemodynamics in the mid-DLPFC, the caudal-DLPFC, and the left RLPFC. The relationships between stress and these PFC subregions depends on the stress indicator adopted. Our finding provides supplementary support to the role of the PFC in processing stress. The preliminary results also shed light on the development of stress response markers in brain activity that can be measured with wearable brain-computer interface technologies.

A variety of self-tracking applications and devices have been developed in recent years to support users in tracking their weight, calories eaten, physical activities, sleep and productivity. The availability of all this data from multiple streams provides a rich environment for experimentation that allows users to improve certain aspects of their lives such as losing weight, getting better sleep or being more productive. In this paper we propose a framework that guides users to define, track, analyse, improve and control goals for better personal productivity. We present the outcome of a single-subject case study that was implemented over one year based on the proposed framework for academic productivity. This pilot study demonstrates how longitudinal multistream self-tracking data can be leveraged to gain actionable insights into personal productivity.

This study aimed to understand whether and how wearable technologies and mHealth services could be nurtured for self-care at the individual-level. We launched a self-care online survey in 2015, and a total of 188 participants (66% female; mean age = 30 years) completed the survey. Following the survey, we also conducted a qualitative study with 12 Fitbit users. In both studies, we focused on understanding three prerequisite elements for behavior change: mindset, tool set and skill set. The results showed that most people had the mindset that self-care was important for preventing chronic diseases, and their top health concerns were sleep quality, body weight, mood, skin conditions and chronic fatigue. As for tool set, users acknowledge the potentially positive impact and efficacy of the technologies in facilitating self-care. However, current technologies have several usability issues such as low accuracy, low technology transparency, and limited feedback. As for skill set, two major obstacles were identified: difficulty in sustaining the usage of the technologies, and lacking domain knowledge and data analysis skills to gain insights from personal data. Different from existing studies which mainly focused on understanding the tool set, i.e., the technologies and services per se, this study produced new insights on the current landscape of people's mindset and skill set on adopting the technologies for health behavior change. We also summarized the opportunities and challenges to guide researchers in designing new wearable technologies and mHealth services for self-care.

In this paper, we propose a relay recovery route maintenance protocol for ad hoc networks to combine the benefits of both proactive and reactive route recovery strategies and to minimize their drawbacks. In our proposal, one or more substitute routes usually become ready for the recovery of every link in a route before its break, while the route recovery process actually starts only when the upstream node of a link confirms the link break. Since this scheme does not broadcast any control packet, it can effectively recover a broken link without heavy control overhead traffic. Also, it helps reduce the time delay due to the recovery, since substitute routes are already available when the upstream node initiates the route recovery process. We further propose two automatic route shortening schemes to optimize the route during successful packet forwarding without causing extra control overhead. We have implemented our proposed schemes based on AODV and compared their performance with competitive schemes including original AODV. Simulation results demonstrate that our proposal definitely reduces the time delay and control overhead traffic in route repairing process, and that the route shortening schemes further leads to shorter time delay and average route length.