Organisation of work and application of tech­nology in production and logistics is likely to change within the forthcoming decade. This paper presents first results of a scenario analysis with a time horizon of 2035 focusing on the exploration of future competence transfer, work culture und learning-conducive environments. The analysis identified the relevance and importance of key stakeholders in this area and, further on, is anticipating the spheres of influence and containing key factors which influence the development of the areas under consideration.

Urban mobility infrastructure preservation and resilience arise from interdependencies across infrastructure, social systems, and institutions, rather than from any single sector in isolation. This paper reframes mobility infrastructure preservation for ICE 2026 by treating cross-domain cascades as a collaboration and governance challenge that shapes competitiveness through reliable labour-market access and urban logistics, sustainability through multi-benefit and equity-sensitive outcomes, and resilience through well-sequenced portfolios that combine engineering, operational, and institutional levers. We present a PRISMA-guided evidence synthesis of 130 Q1 studies published between 2015 and March 2026, drawing on Scopus, Web of Science Core Collection, and IEEE Xplore. We classify couplings using a five-pathway taxonomy (physical, informational/cyber-physical, behavioural/spatial, economic/financial, and institutional/governance), and we group responses into engineering, adaptive, and transformative mechanisms using a codification-and-durability rule to distinguish durable change from temporary responses. Equity is treated explicitly through exposure, access, and agency, and macro-context conditions are interpreted through a PESTLE lens. The synthesis highlights three recurring cascade nexuses, transport-energy-ICT, built environment-health-social, and governance-finance-legal, which repeatedly shape recovery trajectories and distributional effects. We consolidate practitioner-facing benchmark concepts, including recovery parameters such as T₉₀/T_rec, network criticality or percolation thresholds, and accessibility deltas. We identify persistent gaps in causal evidence, metric comparability, and coverage of underrepresented regions. The paper contributes a practical sequencing heuristic (E to A to T) and a minimum indicator kit to support inter-agency monitoring and investment appraisal.

Rural areas lag behind urban areas with respect to digital transformation. Forestry is an industry predominantly operating in rural areas facing increasing demands for efficiency, safety and sustainability. In this paper, we present a new collaborative business model supporting shared risk-taking with local stakeholders to mitigate uncertain return on investment for 5G network providers operating in rural regions. These 5G network capabilities are necessary to implement new digital solutions for the forestry industry. Two innovative solutions are developed, tested and validated in collaboration with corporations, SMEs and research institutes during multiple project stages in a Living lab setting as part of the Horizon Europe funded project COMMECT, aimed at empowering rural industries and public communities with digital innovations built on advanced 5G/IoT network technologies. The collaborative business model developed supports the deployment of these solutions in practice. The validation of the business models involves a combined technology and business readiness assessment (TRL and BRL) expanding the focus on technological maturity frequently applied in digital innovation projects and at the same time supporting the acceleration of the projected maturity and impacts from the innovations in the exploitation and commercialization stage. The developed business model and insights are relevant for multiple actors and industries involved in the funding, investment, technical development and commercialisation of digital innovations

Detecting cyberbullying on social media platforms remains a challenging task due to linguistic ambiguity, severe class imbalance, and evaluation leakage arising from repeated users and interaction structures. To address these challenges, we propose a leakage-aware fusion approach that combines tweet text with lightweight metadata features.
To mitigate user-level memorization and better simulate real-world deployment, we adopt user-disjoint data splits. We further introduce a dual-input neural architecture that integrates a RoBERTa-based text encoder with a multilayer perceptron (MLP) for metadata representation.
Our experimental framework evaluates multiple baselines, including TF-IDF with logistic regression, metadata-only models, and a RoBERTa text-only model, alongside the proposed deep fusion approach. The fusion model employs a partially fine-tuned RoBERTa encoder jointly trained with a metadata MLP.
Experiments conducted on a subset of the Bullying Traces v3.0 dataset show that the deep fusion model improves positive-class F1 from 0.698 to 0.754 under the user-disjoint protocol. Additionally, a structure-only metadata ablation achieves an F1 score of 0.759 while reducing feature complexity.
We further analyze deployment trade-offs through threshold sweeps, highlighting control over false positives and false negatives across different operating points.
Overall, the results emphasize the importance of leakage-aware evaluation and demonstrate that lightweight structural signals can substantially enhance cyberbullying detection performance.