Communication links in vehicular ad hoc networks (VANETs) are frequently disrupted due to high vehicle mobility and rapid topology changes, resulting in packet losses, increased latency, and degraded communication reliability. To address this issue, this paper proposes a Prediction-Based Stable Path Routing (PSPR) scheme using Markov chains. PSPR predicts short-term path stability and selects routes expected to remain connected during the communication period. By incorporating predicted connectivity into routing decisions, the scheme improves packet delivery while reducing unnecessary retransmissions and route failures. The proposed method is evaluated using the Opportunistic Network Environment (ONE) simulator and compared with representative routing protocols concerning packet delivery ratio, latency, hop count, dropped packets, and transmission overhead. Simulation results show that PSPR provides improved routing performance, particularly under moderate-to-high vehicle densities, while maintaining lower communication overhead than replication-based routing schemes.
Trend-driven environments change rapidly, and consumers' decision responses may fluctuate. These fluctuations may be shaped by emotional directions underpinned by different motivational bases: social pressure, self-endorsed values, or both. Instead of focusing on these fluctuations, emotional tendencies—recurring orientations—may be a more stable basis. Such tendencies, grounded in different motivational bases, may shape decision-response tendencies within individual or across all decision contexts. This study conceptualises Fear of Missing Out on Trending Films (FOMO-TF) and Joy of Missing Out on Trending Films (JOMO-TF) as emotional tendencies. Consumer responses across five Consumer Decision-Making Process contexts are conceptualised as decision-response tendencies (DRTs). Data from 418 Thai consumers were analysed using a four-approach multi-method design structured by two conceptual models. It investigates whether each tendency influences individual DRTs and spans DRT pathways. It examines whether co-existing tendencies jointly maintain direction and co-occurring profiles retain their relative ordering across all DRTs. Both emotional tendencies positively influenced every individual DRT. However, FOMO-TF spanned all interconnected DRT pathways, while three positive JOMO-TF pathways fell just short of statistical significance. Thus, influence on each DRT did not necessarily span all DRT pathways. Co-occurring profiles differed mainly in intensity rather than in the mix of the two tendencies and retained relative ordering across all DRTs, despite their different motivational bases. The study introduces the Emotional Tendency Decision Framework and proposes the Process-Based Stage–Profile Strategy Guidelines (PSP) to help managers reduce complexity. Stage, process, and profile strategies target individual decision contexts, interconnected pathways, and consumer profiles, respectively.
AI-powered brand consultation is becoming an important digital touchpoint, yet limited research explains how consumers’ AI capabilities translate into brand-supportive recommendation responses. This study examines how AI literacy shapes willingness to recommend AI-powered online brand consultation through technology-related beliefs and self-determined motivation in Vietnam’s digital commerce context. Drawing on an integrated AI Literacy–UTAUT–Self-Determination Theory framework, AI literacy is conceptualized as a formative higher-order capability comprising awareness, usage, evaluation, and ethics. Based on survey data from 536 Vietnamese online consumers, the proposed model was tested using partial least squares structural equation modeling. The findings show that AI literacy significantly associated with performance expectancy, effort expectancy, social influence, and facilitating conditions, with AI evaluation emerging as the most influential literacy component. Autonomy, competence, and relatedness are positively associated with motivation, while behavioral intention is significantly associated with performance expectancy, facilitating conditions, and motivation, but not by effort expectancy or social influence. Behavioral intention is positively associated with willingness to recommend, suggesting that AI-powered consultation can generate recommendation-oriented brand-supportive responses. Theoretically, the study advances AI-mediated consumer research by positioning AI literacy as an upstream evaluative capability and identifying distinct instrumental and motivational pathways linking AI consultation to recommendation-oriented brand support.
Public-facing social robots are increasingly being introduced into environments such as airports, libraries, hospitals, and campuses to provide information, assistance, and engagement. Their operation in these open and dynamic settings introduces complex challenges that cut across privacy, cybersecurity, ethics, usability, and regulation. While existing studies have explored these dimensions separately, few have examined how they intersect in real-world deployments. This study draws on semi-structured interviews with twenty-one multidisciplinary experts to identify cross-cutting risks and practical design requirements for public-facing social robots. Thematic analysis identified interrelated challenges concerning technical capabilities, public interaction, and organizational conditions, including risks associated with third-party components. Experts emphasized privacy- and security-by-design principles, adaptive and inclusive interaction, resilient technical architectures, and certification mechanisms for responsible deployment. The study’s contribution is a context-specific synthesis that organizes these established engineering and governance principles around the conditions of public-facing deployment. The resulting requirements and indicators are presented as candidate guidance for subsequent operationalization and validation, rather than as validated engineering standards.
Cyber Threat Intelligence (CTI) knowledge graphs are used for threat detection, hunting, and attribution. However, static graphs accumulate outdated intelligence, thereby increasing false positives. In this study, we present TempCTI-KG, a knowledge graph maintenance and versioning framework that models temporal validity loss. Facts are represented by a validity interval and a time-dependent confidence score. Validity decay is learned through censoring-aware survival analysis conditioned on relationship type, source, and graph neighborhood attributes. The full dataset consists of 93,975 observations drawn from 40 Enterprise ATT&CK versions (2018–2026) and data from URLhaus, ThreatFox, and MalwareBazaar. Facts whose confidence falls below the threshold are transferred to a bitemporal archive. In the leak-free temporal evaluation, the relation-level Weibull variant achieved an AUC-ROC of 0.901 and a Brier score of 0.112. The graph-conditioned variant, with an AUC-ROC of 0.936 and a Brier score of 0.094, outperformed the supervised classifier (0.889). It was determined that more than 92% of the facts in the static graph lost their validity during the replay of the URLhaus history. While the best TTL achieved an F1 score of 0.29, the graph-conditioned TempCTI-KG achieved an F1 score of 0.45. It was observed that network indicators have a median survival time of less than one day, while ATT&CK technique relationships have a median lifespan of several decades.
