94500 research outputs found
Sort by
Trade-offs between fairness and performance in educational AI : Analyzing post-processing bias mitigation on the OULAD
No full textContext:
AI-driven educational tools often face a trade-off between fairness and performance, particularly when addressing biases across sensitive demographic attributes. While fairness metrics have been developed to monitor and mitigate bias, optimizing all of these metrics simultaneously is mathematically infeasible, and adjustments to fairness often result in a decrease in overall system performance.
Objective:
This study investigates the trade-off between predictive performance and fairness in educational AI systems, focusing on gender and disability as sensitive attributes. We evaluate whether post-processing fairness interventions can mitigate group-level disparities while preserving model usability.
Method:
Using the Open University Learning Analytics Dataset, we trained four machine learning models to predict student outcomes. We applied the equalized odds post-processing technique to mitigate bias and assessed model performance with accuracy, F1-score, and AUC, alongside fairness metrics including statistical parity difference (SPD) and equal opportunity difference (EOD). Statistical significance of changes was tested using the Wilcoxon signed-rank test.
Results:
All models achieved strong baseline predictive performance, with RF performing best overall. However, systematic disparities were evident, particularly for students with disabilities, showing that high accuracy does not necessarily ensure equitable outcomes. Post-processing reduced group-level disparities substantially, with SPD and EOD values approaching zero, though accuracy and F1-scores decreased slightly but significantly. RF and ANN were more resilient to fairness adjustments.
Conclusion:
This study highlights the importance of fairness-aware machine learning, such as post-processing interventions, and suggests that appropriate mitigation methods should be used to ensure benefits are distributed equitably across diverse learners, without favoring any particular fairness metric.peerReviewe
General introduction to the special issue on resilience in learning
No full textIn this introduction to the special issue ‘Resilience in Learning: In Search of Protective Factors and Compensatory Mechanisms’ we provide an overview of the seven articles featured in this special issue and address important themes in this relatively new area of research. These seven articles each have their own foci and form an introduction into the emerging field of academic resilience in the context of education. They cover a theoretical paper, scoping review, several original empirical studies, and an in-depth reflection on the emerging evidence for protective factors and compensatory mechanisms in the context of learning. As this special issue should only be considered as a starting point, we end with suggesting future directions to advance the field.nonPeerReviewe
Rethinking social media and mental health : The role of emotion regulation difficulties
No full textResearch, on the whole, does not suggest that time spent on social media is associated with risks to mental health, although it is possible there are more nuances about how people use social media. Further, evidence suggests that individuals with emotion regulation difficulties may be drawn to certain social media behaviours as a means of coping with distress. The present study aimed to examine whether emotion regulation difficulties predict patterns of social media use and, in turn, symptoms of depression and anxiety. We examined four distinct types of social media use: (1) image management-based, (2) social comparison-based, (3) negative engagement-based, and (4) passive consumption-based. Sampling 548 adults aged 18–84 years (Mage = 33.16, SD = 17.37; 401 (73.2 %) female; 128 (23.2 %) male), we tested a structural equation model to examine whether the four distinct types of social media use mediated links between difficulties in emotion regulation at Time 1 and depression and anxiety symptomology at Time 2, one week later. Results suggested that, when controlling for age, difficulties in emotion regulation significantly predicted all types of social media use and symptoms of depression and anxiety over one week. Only comparison-based social media use predicted anxiety symptoms over time. The model explained 50.1 % and 52.1 % of the variance in depression and anxiety symptoms, respectively. Taken together, these findings suggest the critical importance of emotion regulation in predicting mental health. By contrast, with the exception of social comparison and anxiety, no form of social media use predicted mental health outcomes.peerReviewe
Artificial Intelligence : Using Machine Learning to Predict Students’ Performance
No full textBeing able to predict students’ performance has been a primary driver for the adoption of learning analytics and has attracted many scientists to the field. Predictive modeling focuses on using students’ data to forecast outcomes such as student grades, enabling teachers and administrators to offer just-in-time support to students at risk. This chapter uses advanced predictive methods, namely machine learning, where the goal is to predict continuous variables like grades. The chapter uses advanced and popular AI/machine learning algorithms like Random Forest, K-Nearest Neighbor, Linear Regression, Neural Networks, and Support Vector Machines. The chapter provides a practical guide to building and evaluating predictive models with R using two approaches: one is the classic approach for predictive modeling with R, and the other more modern approach using the tidymodels suite.peerReviewe
Does AI Democratize Commercial Content Creation? Changes to the Strategic Communication Profession
No full textGenerative AI (GenAI) is suggested to transform and democratize knowledge work such as strategic communication significantly. This chapter reviews changes in strategic communication due to GenAI, examining how past professional experience affects outcomes using the GPT-4o. It tests the AI democratization hypothesis: Does prior expertise matter, or can novices perform as more advanced professionals with the help of AI? The data was gathered from Finnish media company professionals, a creative agency, and pre-professional university students. The study has three parts: (1) self-assessment of experience and GenAI use ( N = 12), (2) GenAI-augmented commercial content creation/strategic communication tasks ( N = 121), and (3) qualitative analysis of outcomes by an experienced content creator. Findings indicate a slight democratization effect of GenAI, enhancing the quality of novices to advanced levels while posing a possibility for negative standardization. Unsupervised GenAI use may mildly harm organizational communication and strategic pursuits. Based on the findings, suggestions for future research are provided.peerReviewe
Detection of Histidine Kinase Activities of Bacteriophytochromes with Phos-tag™ Acrylamide Gel Electrophoresis
No full textPhosphorylation is a key regulatory mechanism of cellular functions, and a majority of known bacteriophytochromes (BphPs) act as histidine kinases (HKs) in two-component signaling. These red/far-red light-sensing HKs play a role in diverse light responses in bacteria. In this chapter, we introduce protocols to easily track the net kinase and phosphatase activities of BphP HKs. The protocols provided here apply Phos-tag™ acrylamide gel electrophoresis for protein samples under controlled illumination conditions. They can be applied to BphP HKs and with a few modifications to photoreceptor HKs in general.peerReviewe
Gender-based cognitive bias and design thinking in the work of Finnish IT professionals
No full textContext
Cognitive bias is a concern in artificial intelligence (AI) development. Research shows the prominence of cognitive bias within algorithms. We argue that cognitive bias is more than training data, but rather development team composition. Design Thinking (DT) is an approach used to reduce bias via multidisciplinary expertise. The article presents a study examining DT in addressing gender-based cognitive bias in the Finnish information technology industry.
Objective
The aim was to examine how the gender of IT professionals influences familiarity with and use of DT, coupled with awareness and addressing of cognitive bias in IT development processes.
Method
A mixed method questionnaire was used to collect data from N = 93 participants. Questions probed familiarity with DT, use of DT, and cognitive bias handling in participants’ organizations. Non-parametric tests were used to analyze quantitative data, due to abnormal distributions. Atlas.ti was used to code and analyze the qualitative data. Categorization determined whether participants recognized bias in their work, and the importance they attributed towards dealing with gender-based bias in IT.
Results
Women were more likely to view gender-based cognitive bias as relevant. Women were significantly more familiar with DT as a methodology (p = 0.028), men were significantly more likely to engage in user studies (p = 0.018). Older participants showed a tendency to emphasize the importance of open discussion more than other participant groups, with some analyses indicating a trend-level difference (p = 0.085). Qualitative responses indicated the importance of discussion in development teams to avoid or mitigate bias, suggesting the need for organizational psychological safety.
Conclusion
The paper provides novel contributions to the human dimension of bias in AI and IT in general. Results show that men and women IT professionals were aware of DT, yet men professionals were more likely to mitigate bias through collecting insight from end-users.peerReviewe
Politicians in Representative Politics : An Introduction to an Ambiguous Conceptual History
No full textpeerReviewe
Mining Patterns and Clusters with Transition Network Analysis : A Heterogeneity Approach
No full textIn this chapter, we demonstrate how we can identify and study clusters within Transition Network Analysis (TNA) to reveal the underlying heterogeneity in learners’ behavioral patterns. Specifically, we rely on mixture Markov models (MMM) to identify latent subgroups characterized by unique transition probabilities, a method that can also incorporate covariates to explain the identified clusters. We employ the tna R package to understand the distinct transition dynamics between states or events in each cluster through the study of centrality measures, communities and cliques. Lastly, we exemplify how to implement other forms of clustering (e.g., distance based) and grouping, as well as other types of transition networks (e.g., frequency-based transition networks).peerReviewe
Synthesis, X-ray diffraction, computational, biological, and molecular docking studies of Cu(II) and Ni(II) complexes containing N, O-donor ligands
No full textSeven metal(II) complexes of furfural (HL1) and thiophene (HL2) hydrazone ligands were synthesized and characterized as [Cu(L1)Cl(H2O)2] (1), [Cu(L1)2] (2), [Cu(HL2)(L2)Cl] (3), [Cu(L2)2].nH2O (4a; n=3 and 4b; n=0), [Ni(L1)2] (5) and [Ni(L2)2] (6). The supramolecular structures of HL2 and 5 were elucidated by single crystal X-ray diffraction, Hirshfeld and DFT calculations. For HL2, the most significant contacts are O‧‧‧H (11.6%) and S‧‧‧H (11.3%) while for 5; the prominent significant contacts are H‧‧‧H (37.1%) and Ni‧‧‧C (2.7 %). XPS confirmed the assembly of Cu(Ⅱ) and Ni(Ⅱ) complexes, where the binding energy differences (∆BE) due to spin orbital coupling of the central metal are 20 and 17 eV, respectively. DFT-calculated molecular structure of 5 aligns with experimental X-ray geometry, facilitating the 3D prediction of other complexes. All studied complexes exhibited enhanced anticancer activity against the HepG-2 and A-549 cancer cell lines compared to their corresponding ligands. 1 and 3 exhibited the highest cytotoxic activity with IC50 values of 11.34±0.23 and 8.45±0.23 µM against the HepG-2 cell line, and 17.43±0.34 and 12.08±0.38 µM against the A-549 cell line, respectively. Moreover, 1 showed improved antibacterial activity, especially against S. aureus and B. subtilis, outperforming gentamycin. Furthermore, 5 showed comparable activity against A. fumigatus to ketoconazole. Molecular docking studies were conducted to explore the inhibitory mechanism of ligands and their synthesized complexes against human topoisomerase II alpha, vascular endothelial growth factor receptor 2, and microbial DNA gyrase. Complex (3) showed higher affinity towards 3WZE and 4FM9 proteins with binding energies of -8.68 and -7.78 kcal/mol, respectively.peerReviewe