Explainable AI for Obsolescence Using Weak Labels

This study examines the perception of global risks in Ukraine and Poland in the context of the ongoing Russian–Ukrainian conflict. Based on cross-sectional survey data, significant differences were found in how respondents from both countries perceive various threats. Ukrainians expressed higher levels of concern across all risk categories, particularly those directly related to the conflict, such as military operations and energy shortages. The results of the principal component analysis (PCA) revealed three distinct factors in Ukraine, indicating a diversified perception of risks, while in Poland, a single factor was identified, suggesting an integrated understanding of threats. These structures can be interpreted through the lens of polycrisis in Ukraine, where different dimensions of risks are seen as interconnected, and metacrisis in Poland, where various risks are viewed as manifestations of a single, overarching crisis structure. The study underscores how geopolitical contexts can shape societal perceptions of threats. It also suggests that the concepts of polycrisis and metacrisis are not only useful theoretical tools for explaining the complexity of contemporary global challenges but also practical frameworks through which individuals reflexively interpret and engage with global risks.

ISO-DeTr: A novel detection transformer for industrial small object detection

This study explored the connection between insurance pre-loss risk surveying and sustainability underwriting in the Nigerian non-life insurance companies. The research design adopted was quantitative based on Structural Equation Modelling (SEM) with data gathered from underwriting managers, risk surveyors, and underwriting officers in the Nigerian non-life insurance companies registered under the National Insurance Commission (NAICOM). The study tested four hypotheses with effect of hazard analysis and evaluation, loss prevention consultation, risk classification and rating, and pre-acceptance survey frequency on sustainability underwriting of non-life insurance companies. The results show that the four dimensions have a significant and positive impact on the sustainability underwriting of non-life insurance companies. The strongest predictor (β = 0.30, p < .001) was Risk Classification and Rating, then Hazard Analysis and Evaluation (β = 0.27, p <.001), Loss Prevention Consultation (β = 0.24, p = 0.003), and Pre-acceptance Survey Frequency (β = 0.20, p =.003). The model fit indices of (SRMR = 0.013, RMSEA = 0.000, CFI = 1.000, TLI = 1.000), proved the strength of the structural relationships. The research finds that sustainable underwriting requires excellence across dimensions of pre-loss risk surveying. The study suggests that the non-life insurance companies in Nigeria should focus on advanced systems of risk classification, institutionalize the critical investigation of hazards, incorporate the loss prevention consultation model into the business model, and initiate the dynamic frequency of surveys.

The aim of this study is to test whether financial risk disclosures required by IFRS 7 and Pillar 3 are value relevant for investors to support them in their investment decisions. The sample in the study consists of banks listed on the London, Paris, Frankfurt, Madrid, and Milan Stock Exchanges over an 8-year period, from 2007 to 2014. Based on the aforementioned standards, we built financial risk disclosure indexes and distinguished different risk categories, qualitative and quantitative, as well as credit, liquidity, and market risk. Our analyses confirm that there is a positive association between bank value and several categories of established risk disclosures. Furthermore, it suggests that disclosure adds value to more traditional risk value measures. Besides, our results suggest that investors pay attention to the strength of the bank authority when using risk disclosures.

Determinants of Economic and Accounting Components of Residual Income: An Application to the Pharmaceutical Industry

Insurance is one of the highest investments by the VIPs, which constitute the main source of profits in the financial industry. With the rapid growth of the insurance informatization, IT outsourcing service is growing up increasingly in the insurance informatization construction of position. In order to obtain better risk assessment, combing with the mechanism research of insurance information technology (IT) outsourcing, this paper develop a Framework Model of insurance IT outsourcing risk assessment based on the Control Objectives for Information and related Technology (COBIT) framework structure and the traditional model of IT outsourcing risk assessment. The Interview and Delphi Methods is to prove the high integration and Acquisition of available of the risk factors of the framework model, the Risk Matrix and Borda Rank Method is to prove the adaptability of the risk category and risk dimension. Finally, the result indicates that the structure of risk category, risk dimension and risk factor can do effective assessment for the insurance IT outsourcing risk.

This study evaluated the effects of sexual victimization history, rape myth acceptance, implicit attention, and recent learning on the cognitive processes underlying undergraduate women's explicit risk judgments. Participants were 194 undergraduate women between 18 and 24 years of age. The sample was ethnically diverse and composed primarily of freshman, heterosexual, and single women. Stimuli were written vignettes describing social situations that varied on dimensions of sexual victimization risk and potential impact on women's popularity. Participants completed cognitive tasks assessing relative attention to victimization risk versus popularity impact, learning about either risk or popularity impact, and explicit classification of victimization risk. Participants then completed the Sexual Experiences Survey (SES) and the Rape Myth Acceptance Scale; SES responses were used to quantify the severity of victimization experiences. More severe victimization history predicted use of higher thresholds for judging situations as risky, as well as lower sensitivity to risk and greater sensitivity to popularity impact when judging risk. Greater rape myth acceptance also predicted lower sensitivity to risk information. Higher relative attention to victimization risk predicted greater sensitivity to risk information when judging risk. Recent learning about either the risk or the popularity impact aspects of social situations modified sensitivity to risk versus popularity when making risk judgments. The study emphasizes the importance of distinguishing the threshold for judging situations as risky from sensitivity to risk-relevant information in understanding individual differences in women's risk judgments. Both processes may be important to consider when developing interventions to reduce women's risk for sexual victimization.

SUMMARYA major bottleneck in developing clinically impactful machine learning models is a lack of labeled training data for model supervision. Thus, medical researchers increasingly turn to weaker, noisier sources of supervision, such as leveraging extractions from unstructured text reports to supervise image classification. A key challenge in weak supervision is combining sources of information that may differ in quality and have correlated errors. Recently, a statistical theory of weak supervision called data programming has shown promise in addressing this challenge. Data programming now underpins many deployed machine-learning systems in the technology industry, even for critical applications. We propose a new technique for applying data programming to the problem of cross-modal weak supervision in medicine, wherein weak labels derived from an auxiliary modality (e.g., text) are used to train models over a different target modality (e.g., images). We evaluate our approach on diverse clinical tasks via direct comparison to institution-scale, hand-labeled datasets. We find that our supervision technique increases model performance by up to 6 points area under the receiver operating characteristic curve (ROC-AUC) over baseline methods by improving both coverage and quality of the weak labels. Our approach yields models that on average perform within 1.75 points ROC-AUC of those supervised with physician-years of hand labeling and outperform those supervised with physician-months of hand labeling by 10.25 points ROC-AUC, while using only person-days of developer time and clinician work—a time saving of 96%. Our results suggest that modern weak supervision techniques such as data programming may enable more rapid development and deployment of clinically useful machine-learning models.

LIDIAP

Purpose: The goal of this study is to map the current state of academic knowledge on Industry 5.0 by identifying key technologies, required competencies, and emerging thematic areas through a systematic literature review. Based on this analysis, the study proposes a conceptual framework that synthesizes these findings to support future research and enhance understanding of this evolving industrial paradigm—particularly by identifying critical capabilities for implementing human-centric, sustainable, and resilient strategies aligned with the Sustainable Development Goals (SDGs). Methodology: A systematic literature review was conducted by querying the Scopus and Web of Science databases. This comprised eight steps designed to comprehensively understand existing literature on I5.0. In total, 470 papers were assessed, 112 of which qualified for subsequent data extraction and analysis. Findings: The study organizes existing literature into five key thematic domains that structure current academic knowledge and guide future research on Industry 5.0: (1) Technologies and Digital Development; (2) Education, Skills, and Knowledge, alongside People, Ethics, Health, and Safety; (3) Society, Laws, and Government; (4) Benefits, Challenges, and General Factors, in conjunction with Organizational Strategy and Management; and (5) Sustainability. These domains offer a comprehensive perspective on the core components and interdependencies that characterize the Industry 5.0 paradigm. The findings also indicate that future research should prioritize empirical studies that examine how human-centricity, resilience, and sustainability are being operationalized within real-world industrial contexts. Implications: This study highlights specific gaps for future research, particularly the need to explore how human-centricity, sustainability, and resilience are integrated in practice. For organizations, the findings provide clarity on key technologies, competencies, and strategic priorities, helping guide transitions aligned with the SDGs. Originality: While grounded in a systematic literature review, this study offers originality by structuring dispersed academic knowledge into five thematic domains that reflect the evolving nature of Industry 5.0. The resulting framework does not claim to be a new theoretical model but adds value by connecting concepts, technologies, competencies, and implementation challenges in a coherent and actionable structure. This contribution helps scholars frame future research and offers organizations a clearer path to navigating the Industry 5.0 landscape.

In complex industrial environments, uncertainty is inherent in decision-making due to dynamic operating conditions, sensor variability, and the vast heterogeneity of data sources. Traditional deterministic models often fall short in capturing the probabilistic dependencies and hidden causal relationships that characterize these systems. As industries increasingly adopt data-driven strategies, probabilistic reasoning frameworks such as Bayesian Network (BN) modeling have gained prominence for their ability to encode domain knowledge, handle incomplete information, and support transparent inference under uncertainty. Bayesian Networks offer a graphical model-based approach to representing joint probability distributions over multiple interrelated variables. In large-scale industrial datasets—ranging from manufacturing process logs and predictive maintenance records to energy grid telemetry and supply chain metrics—BNs enable efficient reasoning by decomposing complex dependencies into directed acyclic graphs. These structures support not only diagnostic and prognostic tasks but also counterfactual analysis and real-time decision support. This paper explores the methodology and practical application of Bayesian Network Modeling for probabilistic reasoning and risk assessment in industrial contexts. Emphasis is placed on model construction from big data, structure learning from high-dimensional variables, and parameter estimation under noisy or partially missing data. Case studies from fault prediction in chemical processing plants and anomaly detection in smart grid infrastructure illustrate the scalability and interpretability of BNs in practice. The integration of expert knowledge with data-driven inference highlights the hybrid power of Bayesian models in enhancing industrial resilience, safety, and strategic planning.

When introducing new technologies, or dealing with uncertain situations in general, risk assessment is an established methodology to systematically and reliably consider whether intended benefits are gained or whether unwanted adverse effects are likely to occur. The German Advisory Council on Global Change (WBGU) distinguishes different risk categories according to the probability and extent of any possible damage. Building on that, the Gene-Risk Research Consortium elaborated a hierarchical risk assessment approach to analyze possible impact of the cultivation of genetically modified organisms (GMO). This approach is also adaptable to risks involved in the development of synthetic life forms. Since the use of GMO affects different levels of organization addressed by different scientific disciplines and stakeholders, the potential risks of GMO cultivation have to be denoted as being systemic and require interdisciplinary as well as transdisciplinary co-operation. Synthetic biology can build on risk management solutions which have been established for the use of GMO—at least to the extent that comparable risk dimensions have to be covered. For both technologies, risk assessment has to consider a wide spectrum of cause-and-effect chains and the potential impact over long time spans and large areas of space. It must also consider potential self-dispersal and subsequent evolutionary processes in the ecosphere after intended or accidental release into open ecosystems. A holistic risk assessment approach to GMO was settled upon by applying the concept of emergent properties structuring possible effects for different levels of organisation considering that interactions on lower levels (molecules, cells, organisms) in composition are supposed to bring up new interaction types on higher levels (populations, ecosystems, landscapes, biomes). In comparison with this structured approach, the current legal regulations, as established in the EU, can be improved in coherence and systematization by the proposed approach, in particular with regard to different ecological and economic implications of GMO (and in parallel potential releases of synthetic organisms). This is especially relevant on the landscape level; for instance, as a comprehensive systematization of region-specific adverse effects on non-target organisms, complex coexistence issues related to different production systems or some social ecological topics. In conclusion, human intervention involving self-reproducing entities by means of genetic engineering, as well as development of new synthetic life forms, should always evaluate the complete set of causal interactions on all levels of physical, biotic and social organisation in order to minimise the probability of unintended, undesirable and even harmful effects as far as feasible through anticipative assessment.KeywordsRisk AssessmentSynthetic BiologySystemic RiskEmergent PropertyGenetically Modify OrganismThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

The purpose of the article is to develop strategic priorities for the formation of competitive advantages of domestic sugar producers on the world market of agricultural products, based on the generalization of the theoretical foundations of competitive interactions of subjects of the world agricultural market and a comprehensive study of strategies for ensuring global leadership of sugar industry enterprises. Research methodology. To achieve the goal, a set of general scientific and special research methods was used using the methods of scientific abstraction, monographic analysis and synthesis, comparative and systematic analysis, abstract-logical analysis, which ensured the reliability, objectivity and thoroughness of the provisions and conclusions presented in the scientific article. Research results. Today, national economic development forms priority tasks, among which the international competitiveness of the domestic agro-industrial complex with the provision of food and energy security of our country occupies a prominent place. In this context, the state and development of the sugar industry of Ukraine is of particular interest. And that is why it is relevant to study the behavior of agro-industrial companies, and especially the use of competitive strategies on the world market of agro-industrial products. The general conclusions and recommendations of the study can be used in the activities of domestic agro-industrial companies and the formation of the state strategy for the development of the agro-industrial complex in order to strengthen competitive positions in the world market. Conclusions. Studying the practices of forming competitive strategies of the sugar industry in the world, the author indicated that the optimal strategy for increasing the competitiveness of domestic sugar producers should be the formation of a comprehensive strategy, which will be based on effective mechanisms, both on the part of producers and on the part of state regulation, which in in turn, it will contribute to the sustainable development of the industry, economic growth and strengthening of the position of sugar producers in foreign markets. Programs to ensure the competitiveness of sugar producers should be based on relationships between state authorities, domestic market participants and foreign partners.

Modern machine learning (ML) models are being used heavily in business domains to build effective decision support systems. As a primary requirement, supervised ML models need large labeled datasets. However, obtaining a high volume of labeled training data is both expensive and time-consuming. Researchers have proposed several labeling approaches to avoid manual labeling efforts. Active learning (AL) and Data Programming (DP) are two state-of-the-art techniques used to label datasets. Nevertheless, both approaches have their strengths and weaknesses. For example, AL is computationally expensive to apply on large industrial datasets; and labels generated by DP are often inaccurate and difficult to interpret. To address these challenges, in this paper, we propose a novel hybrid method that integrates the scalability of DP with the user engagement and accuracy of AL. The proposed approach aims at optimizing the labeling process by applying DP to generate initial noisy training data and then use AL to query the user to label only those points that maximize the accuracy of the final labels with a minimum annotation cost. To evaluate the proposed approach, we have used five open source datasets and a real-world business dataset of 1.5 million records. We use traditional active learning and data programming techniques as baselines to compare the performance and annotation cost of our proposed approach. The results show that the proposed method can achieve higher labeling accuracy than data programming. It also can minimize the labeling cost in real-world business scenarios, while delivering a comparable level of performance (accuracy) with active learning.

A spherical fuzzy objective weighting framework for factors affecting bidding decisions for construction projects: A case study of Taiwan