340880 research outputs found
Sort by
When Do Individuals Believe in Themselves Rather Than in Artificial Intelligence? Insights from Longitudinal Investigations in Corporate Credit-Rating Contexts
No full textIndividuals often prioritize their own judgements rather than heeding the advice of artificial intelligence (AI). This study draws on the literature on anchoring theory and cognitive biases to explore the theoretical mechanisms underlying individuals' reliance on AI advice and how this reliance affects decision performance. Specifically, we examined situations in which (1) individuals' knowledge accumulated over time, (2) multiple information sources were available, and (3) AI could emulate users' decisions. We developed a 'corporate credit-rating' AI system that could provide more accurate advice than users. We then conducted two main longitudinal studies and four supplementary ones - six in total - with each study comprising three sessions. Our findings demonstrated that individuals' initial estimates became more similar to AI advice over time. As the difference between individuals' initial estimates and AI advice increased, individuals were more inclined to revise their initial judgements but showed lower relative dependence on AI. This effect, however, depended on the individuals' experience in decision-making. Additionally, introducing additional information reduced the similarity between the initial estimate and AI advice, but the proximity of additional information to AI advice facilitated individuals' adjustment to the advice. We discuss the theoretical and practical implications of these results
Inclusive Language and Privacy Policies:A Rights-based Approach
No full textThis article explores the role of inclusive language in privacy policies, emphasizing its legal significance and impact on user trust and regulatory compliance. By analysing privacy policies from Amazon, Uber, and Meta, the study highlights the prevalent use of masculine as a default linguistic form, reinforcing gender biases and excluding diverse identities. The research argues that the right to personal data protection should inherently include non-discriminatory language as part of the principle of transparency under EU data protection law. The study further underscores how privacy policies, as fundamental tools for informing users about their rights, should reflect corporate commitment to gender equity and non-discrimination. It proposes that supervisory authorities should consider language inclusivity when interpreting GDPR requirements and suggests AI-driven solutions to assist companies in implementing inclusive language in legal texts. Ultimately, the findings call for a shift toward gender-fair language in privacy policies as a necessary step toward fostering a more equitable digital environment aligned with EU fundamental rights
Malignant pleural mesothelioma classification and survival prediction with CT imaging using ResNet
No full textObjectives: This study aims to achieve accurate differentiation of malignant pleural mesothelioma (MPM) from metastatic pleural disease (MPD) and to predict the overall survival of MPM. Materials and methods: This IRB-approved retrospective study included 385 subjects in total (85 patients with malignant mesothelioma and 290 with MPD secondary to lung adenocarcinoma). A ResNet-3D-18 model was trained on annotated pretreatment CT scans to distinguish MPM from MPD. Using chronological segregation, the training cohort included 70 histologically confirmed mesothelioma and 258 MPD cases, with an independent test cohort of 15 MPM and 32 MPD cases for validation. A multivariate logistic regression model served as the clinical benchmark for comparison. Deep learning features extracted from the trained ResNet model were then assessed for their prognostic utility in MPM patients using a random forest classifier. Model performance was evaluated at both lesion- and patient-levels, with metrics including the area under the ROC curve (AUC), accuracy, sensitivity, specificity, positive predictive value, and negative predictive value. Results: The ResNet-3D-18 model demonstrated excellent discriminative performance in differentiating MPM from MPD, with mean AUCs of 0.972 (95% CI 0.947–0.990) and 0.840 (95% CI 0.757–0.929) in the training and independent test cohorts. Compared to the clinical model, the deep learning approach showed higher sensitivity (0.867 vs. 0.533) in the independent test dataset. For overall survival prediction in MPM patients, the random forest classifier achieved an AUC of 0.829 (95% CI 0.663–0.943) in 5-fold cross-validation. Conclusions: ResNet-3D-18 classification model has excellent abilities in differentiating MPM from MPD, and morphological distinctions between MPM and MPD also contain prognostic information. Key Points: Question The rising global incidence of malignant pleural mesothelioma contrasts with persistent diagnostic challenges. Findings Deep learning-derived discriminative features simultaneously contain prognostic information. Clinical relevance This study bridges the gap between radiological findings and clinical decision-making in MPM, offering a reproducible tool for early diagnosis and personalized prognosis prediction based on CT imaging alone.</p
AI in Early and Primary Education:Societal, Classroom, and Teacher Perspectives on Ethical and Pedagogical Integration
No full textThe integration of artificial intelligence (AI) in education (AIEd) comes along with both opportunities and challenges, particularly in ensuring its alignment with pedagogical principles and the teachers’ needs. This chapter explores the role of AIEd from the perspective of particular teacher needs, classroom dynamics, and broader societal implications. Employing the digital divide theory, we reflect upon the potential inequalities of AI and its impact in education. Furthermore, we discuss the findings of our systematic literature review on the current use of AIEd with particular emphasis on pre-school and primary education. Our results indicated that while AIEd applications in pre-school and primary education promise efficiencies in personalized learning and administrative tasks, their development and implementation often overlook critical pedagogical considerations and teacher guidance. Last, the chapter argues that teachers play an essential role in bridging the gap between technology and effective teaching, ensuring that the AIEd applications will not just be technologically advanced but also aligned with the learning goals and course design needs. Collaboration between different stakeholders (researchers, teachers, developers) is essential to create AI tools that are user-friendly, ethically sound, and tailored to meet diverse student needs
Private vs. Public Schooling:The role of school composition
No full textPublicly funded private schooling is a common feature of many education systems, yet its implications for educational equity and effectiveness remain contested. While private schools often exhibit higher student achievement, the sources of this advantage are not well understood. In particular, differences in student composition-especially in terms of socioeconomic status (SES)-are likely to play a key role. This paper examines how school-level SES composition contributes to achievement differences between public and private schools. Using propensity score matching (PSM) on data from 22,441 French ninth-grade students, we find that private school students outperform their public school peers in mathematics and French, with especially large effects for low-SES students, an underrepresented group in private schools. While school composition explains only part of these effects, it accounts for a substantial share of the performance gap among high-SES students, rendering the adjusted effect statistically indistinguishable from zero. These findings highlight which students benefit most from private schooling and point to the need for further research into the mechanisms underlying performance differences across school sectors
Patient-Centred Explainability in IVF Outcome Prediction
No full textThis paper evaluates the user interface of an in vitro fertility (IVF) outcome prediction tool, focussing on its understandability for patients or potential patients. We analyse four years of anonymous patient feedback, followed by a user survey and interviews to quantify trust and understandability. Results highlight a lay user's need for prediction model explainability beyond the model feature space. We identify user concerns about data shifts and model exclusions that impact trust. The results call attention to the shortcomings of current practices in explainable AI research and design and the need for explainability beyond model feature space and epistemic assumptions, particularly in high-stakes healthcare contexts where users gather extensive information and develop complex mental models. To address these challenges, we propose a dialogue-based interface and explore user expectations for personalised explanations
Functional near-infrared spectroscopy as a biomarker of TMS efficacy in treatment-resistant depression
No full textBackground: Reliable biomarkers for predicting treatment response and suicide risk in treatment-resistant depression (TRD) are limited. Functional near-infrared spectroscopy (fNIRS) offers a noninvasive means to assess prefrontal cortical activation linked to therapeutic outcomes. Methods: In a double-blind, randomized, sham-controlled trial, 100 inpatients with TRD received either active or sham prolonged intermittent theta-burst stimulation (aiTBS) over the left dorsolateral prefrontal cortex (DLPFC) across 2 weeks. fNIRS measured oxyhemoglobin (oxy-Hb) levels at rest and during a verbal fluency task (VFT) and two-back working memory task, both before and after aiTBS. Clinical outcomes included Montgomery-Åsberg Depression Rating Scale (MADRS), Hamilton Depression Rating Scale item 3 (HAMD-3), and Beck Scale for Suicide Ideation (BSS). Results: Baseline BSS, HAMD-3, and MADRS scores did not differ between groups (all P > 0.05). Post-treatment, the active group showed significant improvements in BSS, HAMD-3, and MADRS (all P < 0.05). Active aiTBS increased oxy-Hb in the left DLPFC and right orbitofrontal cortex (OFC) during the two-back task, and in the left DLPFC, OFC, and frontopolar cortex (FPC) during the VFT. Greater left DLPFC activation during the VFT correlated with MADRS improvement, and baseline OFC activation predicted antidepressant response. No fNIRS measures predicted changes in suicidality. Conclusions: Task-evoked prefrontal activation—especially in the left DLPFC and OFC—may serve as a biomarker for antidepressant efficacy in TRD, though fNIRS did not predict suicide risk reduction.</p
Dynamic Conformal Prediction for Multi-Target Regression:Optimising Informational Efficiency under Joint Validity
No full textInductive conformal prediction equips point regressors with finite-sample prediction sets that provably contain the unknown label with prescribed probability. For multi-target regression, joint coverage across all output dimensions can be guaranteed by combining one-dimensional conformal predictors, one for each output dimension, resulting in an axis-aligned hyperrectangular prediction region. The validity and informational efficiency of these hyperrectangular prediction regions depend on the choice of the targeted error rate for the individual one-dimensional conformal predictors. We cast this choice as an error-budget allocation problem and introduce Dynamic Conformal Prediction for Multi-Target Regression (DCP-MT), a method that finds the budget allocation which minimises the hyperrectangles’ volumes while retaining joint coverage under exchangeability. Experiments on synthetic and real-world data sets demonstrate that DCP-MT reduces hyperrectangle volumes compared to state-of-the-art methods when nonconformity scores’ correlations across target dimensions are weak or heterogeneous, while maintaining the nominal coverage. The proposed method thus offers a simple, drop-in solution for existing multi-target regression pipelines
Learning asymmetry as a predictor of mood and behavior dynamics:A network analysis
No full textWhile studying appetitive and aversive conditioning is common in psychopathology research, studies that measure both types of learning simultaneously are rare. To gain insight into the role of appetitive and aversive learning in the complex interaction of positive mood, negative mood, worry, craving, avoidance and impulsive behavior, this study used a relative measure of the strength of appetitive versus aversive learning – the learning asymmetry – as a predictor of network dynamics of mood states and behavior. 100 healthy volunteers performed an appetitive and aversive conditioning task and completed an ecological momentary assessment study, where they were surveyed six times per day for 21 days. Groups were defined based on higher sensitivity to appetitive learning (positive learning asymmetry) or aversive learning (negative learning asymmetry). The positive asymmetry group was hypothesized to be more sensitive to positive mood changes, and the negative asymmetry group was hypothesized to be more sensitive to negative mood changes. Contrary to our hypothesis, results show that impulsive behavior was more likely to follow negative mood, specifically anger, in the positive but not the negative asymmetry group. These results demonstrate the potential for network analysis to elucidate complex interactions between mood and behavior associated with individual differences in learning