2444402 research outputs found
Sort by
Gibberellic acid and light effects on seed germination in the seagrass <i>Zostera marina</i>
No full textAbstractSeagrass meadows have been heavily affected by human activities, with Zostera marina L. (Zosteraceae) being one of the most impacted species. Seed‐based methods are currently the preferred approach for their restoration, yet low germination rates and poor seedling establishment remain significant challenges. This study explored the combined effects of light spectra (white, red, and darkness), photoperiod, and gibberellic acid (GA3−0, 50, 500, and 1000 mg L−1) on Z. marina seed germination using a fully crossed incubation experiment. Penalised logistic regression and Cox proportional hazards analysis were chosen to account for low germination events and to analyse the temporal dynamics of germination. We found that light conditions, particularly red light and darkness, when combined with GA3, significantly enhanced germination probability. Furthermore, mid (50 mg L−1) and high (500 mg L−1) GA3 concentrations reduced time‐to‐germination. Morphometric analysis of the cotyledonary and leaf tissue development indicates no adverse effects of the treatments on seedling development. Our findings suggest that light and GA3 treatments effectively improve germination success and reduce dormancy in Z. marina seeds. Seed treatments can mitigate stress‐ or manipulation‐induced dormancy and can represent a viable strategy for on‐demand germination, such as in the context of seed‐based restoration efforts.AbstractSeagrass meadows have been heavily affected by human activities, with Zostera marina L. (Zosteraceae) being one of the most impacted species. Seed‐based methods are currently the preferred approach for their restoration, yet low germination rates and poor seedling establishment remain significant challenges. This study explored the combined effects of light spectra (white, red, and darkness), photoperiod, and gibberellic acid (GA3−0, 50, 500, and 1000 mg L−1) on Z. marina seed germination using a fully crossed incubation experiment. Penalised logistic regression and Cox proportional hazards analysis were chosen to account for low germination events and to analyse the temporal dynamics of germination. We found that light conditions, particularly red light and darkness, when combined with GA3, significantly enhanced germination probability. Furthermore, mid (50 mg L−1) and high (500 mg L−1) GA3 concentrations reduced time‐to‐germination. Morphometric analysis of the cotyledonary and leaf tissue development indicates no adverse effects of the treatments on seedling development. Our findings suggest that light and GA3 treatments effectively improve germination success and reduce dormancy in Z. marina seeds. Seed treatments can mitigate stress‐ or manipulation‐induced dormancy and can represent a viable strategy for on‐demand germination, such as in the context of seed‐based restoration efforts.
Preschool teachers' beliefs and perspectives on gender in education : a qualitative study in Flanders
No full textPreschool teachers are crucial actors in the socialization process and the construction of gender ideas of young children. Their gender ideas influence interactions with children. Education has a profound gendered nature in which gender stereotypes are often reproduced and enacted. This qualitative study is a critical exploration of preschool teachers' beliefs and perspectives on sex and gender in the classroom. Semi-structured interviews (n = 22) with Flemish preschool teachers were conducted and analysed using thematic analysis. The findings show that preschool teachers have an open-minded but ambivalent vision towards gender (stereotypes), which, as they report, is reflected in the classroom. The vast majority of preschool teachers act intuitively rather than consciously or explicitly addressing gender in the classroom. Most teachers say not to ban stereotypical toys or activities and aim to make sure what they offer in their classroom allows the stimulation of children's agency in the gender socialization process.Preschool teachers are crucial actors in the socialization process and the construction of gender ideas of young children. Their gender ideas influence interactions with children. Education has a profound gendered nature in which gender stereotypes are often reproduced and enacted. This qualitative study is a critical exploration of preschool teachers' beliefs and perspectives on sex and gender in the classroom. Semi-structured interviews (n = 22) with Flemish preschool teachers were conducted and analysed using thematic analysis. The findings show that preschool teachers have an open-minded but ambivalent vision towards gender (stereotypes), which, as they report, is reflected in the classroom. The vast majority of preschool teachers act intuitively rather than consciously or explicitly addressing gender in the classroom. Most teachers say not to ban stereotypical toys or activities and aim to make sure what they offer in their classroom allows the stimulation of children's agency in the gender socialization process.A
Diepzeemijnbouw op een kruispunt: wordt 2025 een sleuteljaar?
No full textHoewel het niet de eerste keer zou zijn dat men spreekt van een sleuteljaar voor de diepzeemijnbouwsector en het daaraan gekoppelde juridisch regime, blijkt 2025 toch opnieuw zeer bepalend te worden. Tegengestelde trends hebben zich de voorbije jaren gemanifesteerd en een (al dan niet frontale) botsing lijkt stilaan onafwendbaar.Hoewel het niet de eerste keer zou zijn dat men spreekt van een sleuteljaar voor de diepzeemijnbouwsector en het daaraan gekoppelde juridisch regime, blijkt 2025 toch opnieuw zeer bepalend te worden. Tegengestelde trends hebben zich de voorbije jaren gemanifesteerd en een (al dan niet frontale) botsing lijkt stilaan onafwendbaar.
Below the tip of the iceberg : exploring understudied feedback interactions in organizations
No full textPublic defense: 2025-03-11D
Empowering teachers for interdisciplinary collaborative design : exploring Teacher Design Teams for general subjects in secondary vocational education
No full textPublic defense: 2025-04-01D
Evaluating feature importance biases in logistic regression: Recommendations for Robust Statistical Methods. Author's reply
No full textOver seksuele gezondheid bij ouderen heerst nog steeds een taboe
No full textThis article comprises a narrative literature review of sexual health in older adults, a topic that remains taboo. According to the World Health Organization (WHO), sexuality is an essential aspect of human existence throughout one’s entire life. It is a common misconception that older adults are no longer interested in intimacy and sex. However, research shows that individuals over the age of 65 remain sexually active. For instance, one in three Belgians over 70 years is still sexually active, comparable to their Dutch and American peers. In addition to sexual intercourse, other forms of sexual activity, such as masturbation, oral sex, and especially tenderness (kissing, cuddling etc.) are very common. Moreover, sexually active older adults seem to be satisfied with their sex life. Sexual satisfaction is associated with better physical health, fewer depressive symptoms and a better quality of life. Unfortunately, caregivers often see older adults as ‘asexual’ or the opposite ‘oversexed’ if they show their interest in sex. Moreover, sexuality in older adults is insufficiently addressed in the training of health-care professionals and these negative stereotypes or ‘sexual ageism’ are rarely discussed. As a result, caregivers pay too little attention to the sexual health of their older patients and don’t take the initiative to discuss this topic. This allows sexual problems, sexually transmitted infections (STI’s) and sexual violence among older adults to go unnoticed.To promote sexual health in the older population, there is an urgent need for increased awareness and research about sexuality in older adults, fewer prejudices and better training of health care professionals.This article comprises a narrative literature review of sexual health in older adults, a topic that remains taboo. According to the World Health Organization (WHO), sexuality is an essential aspect of human existence throughout one’s entire life. It is a common misconception that older adults are no longer interested in intimacy and sex. However, research shows that individuals over the age of 65 remain sexually active. For instance, one in three Belgians over 70 years is still sexually active, comparable to their Dutch and American peers. In addition to sexual intercourse, other forms of sexual activity, such as masturbation, oral sex, and especially tenderness (kissing, cuddling etc.) are very common. Moreover, sexually active older adults seem to be satisfied with their sex life. Sexual satisfaction is associated with better physical health, fewer depressive symptoms and a better quality of life. Unfortunately, caregivers often see older adults as ‘asexual’ or the opposite ‘oversexed’ if they show their interest in sex. Moreover, sexuality in older adults is insufficiently addressed in the training of health-care professionals and these negative stereotypes or ‘sexual ageism’ are rarely discussed. As a result, caregivers pay too little attention to the sexual health of their older patients and don’t take the initiative to discuss this topic. This allows sexual problems, sexually transmitted infections (STI’s) and sexual violence among older adults to go unnoticed.To promote sexual health in the older population, there is an urgent need for increased awareness and research about sexuality in older adults, fewer prejudices and better training of health care professionals.
Evaluating the stability of model explanations in instance-dependent cost-sensitive credit scoring
No full textIn credit scoring, the use of machine learning techniques to support or automate decision-making has become increasingly prevalent. While many advanced models have been proposed to improve discriminatory power, they often overlook the financial consequences of misclassifying applicants, particularly when the costs of such errors are asymmetrical and vary by loan. To address this, several instance-dependent cost-sensitive (IDCS) classifiers have been proposed, incorporating individual loan costs directly into the model's loss function (Bahnsen et al. 2014, Höppner et. al 2020, Vanderschueren et al. 2022). However, these models are often "black-box" algorithms, making their predictions hard to interpret, raising concerns amid increasing regulatory scrutiny, such as the GDPR and the proposed EU AI Act.In response, eXplainable AI (XAI) techniques, such as SHAP and LIME, have emerged to provide post-hoc explanations of opaque models' decision-making processes while maintaining predictive performance. For a model to be truly interpretable, however, its explanations must not only clarify its decisions but also exhibit robustness, ensuring that similar inputs yield consistent explanations (Alvarez-Melis and Jaakkola, 2018). Unstable explanations may undermine the utility of these models, exposing financial institutions to legal and reputational risks. While existing studies in credit scoring have explored the stability of SHAP and LIME explanations (Visani et al., 2022), as well as the impact of class imbalance on these explanations (Chen et al., 2024), the behaviour of these methods when applied to IDCS classifiers remains unexplored. This gap is critical because XAI techniques such as SHAP and LIME generate explanations based on the raw predictions of a model. Consequently, alterations in the learning process---such as the integration of instance-dependent costs into the loss function---may profoundly impact both the nature and stability of these explanations.Our research proposes a two-faceted methodology: first, it evaluates prevalent IDCS classifiers in terms of their cost-efficiency and discriminatory power, and second, it assesses the stability of SHAP and LIME explanations when applied to these classifiers. We conduct a comparative analysis of traditional machine learning models and their IDCS counterparts using four open-source credit scoring datasets. To evaluate model performance, we use a combination of traditional and cost-sensitive metrics, introducing relative Average Expected Cost (relAEC) as a novel, dimensionless adaptation of the AEC metric designed for cross-dataset comparisons. To assess explanation stability, we measure the Coefficient of Variation (CV) and Sequential Rank Agreement (SRA) (Ekstrøm et al., 2018) of feature importances for both traditional and IDCS models. We also examine the additional impact of class imbalance through a controlled resampling procedure, following Chen et al. (2024).The results show that while IDCS classifiers improve cost-efficiency, they produce significantly less stable explanations compared to traditional models, especially as class imbalance increases. This highlights a critical trade-off between cost optimization and interpretability in credit scoring. Given the growing regulatory emphasis on explainability, this research underscores the urgent need to address the stability issues in IDCS classifiers to ensure that their cost advantages are not overruled by unreliable or untrustworthy explanations.In credit scoring, the use of machine learning techniques to support or automate decision-making has become increasingly prevalent. While many advanced models have been proposed to improve discriminatory power, they often overlook the financial consequences of misclassifying applicants, particularly when the costs of such errors are asymmetrical and vary by loan. To address this, several instance-dependent cost-sensitive (IDCS) classifiers have been proposed, incorporating individual loan costs directly into the model's loss function (Bahnsen et al. 2014, Höppner et. al 2020, Vanderschueren et al. 2022). However, these models are often "black-box" algorithms, making their predictions hard to interpret, raising concerns amid increasing regulatory scrutiny, such as the GDPR and the proposed EU AI Act.In response, eXplainable AI (XAI) techniques, such as SHAP and LIME, have emerged to provide post-hoc explanations of opaque models' decision-making processes while maintaining predictive performance. For a model to be truly interpretable, however, its explanations must not only clarify its decisions but also exhibit robustness, ensuring that similar inputs yield consistent explanations (Alvarez-Melis and Jaakkola, 2018). Unstable explanations may undermine the utility of these models, exposing financial institutions to legal and reputational risks. While existing studies in credit scoring have explored the stability of SHAP and LIME explanations (Visani et al., 2022), as well as the impact of class imbalance on these explanations (Chen et al., 2024), the behaviour of these methods when applied to IDCS classifiers remains unexplored. This gap is critical because XAI techniques such as SHAP and LIME generate explanations based on the raw predictions of a model. Consequently, alterations in the learning process---such as the integration of instance-dependent costs into the loss function---may profoundly impact both the nature and stability of these explanations.Our research proposes a two-faceted methodology: first, it evaluates prevalent IDCS classifiers in terms of their cost-efficiency and discriminatory power, and second, it assesses the stability of SHAP and LIME explanations when applied to these classifiers. We conduct a comparative analysis of traditional machine learning models and their IDCS counterparts using four open-source credit scoring datasets. To evaluate model performance, we use a combination of traditional and cost-sensitive metrics, introducing relative Average Expected Cost (relAEC) as a novel, dimensionless adaptation of the AEC metric designed for cross-dataset comparisons. To assess explanation stability, we measure the Coefficient of Variation (CV) and Sequential Rank Agreement (SRA) (Ekstrøm et al., 2018) of feature importances for both traditional and IDCS models. We also examine the additional impact of class imbalance through a controlled resampling procedure, following Chen et al. (2024).The results show that while IDCS classifiers improve cost-efficiency, they produce significantly less stable explanations compared to traditional models, especially as class imbalance increases. This highlights a critical trade-off between cost optimization and interpretability in credit scoring. Given the growing regulatory emphasis on explainability, this research underscores the urgent need to address the stability issues in IDCS classifiers to ensure that their cost advantages are not overruled by unreliable or untrustworthy explanations.C
The impact of undermining belief on traumatic memory /
No full textDissertation presented in partial fulfillment of the requirements for the degree of Doctor in criminological sciences and the degree of Doctor in pscholog