Nowadays, adolescents’ mental health issues are becoming more and more severe. In Japan, a survey carried out in 2021 showed that about 10% of 11-12 year-old students and about 20% of 13-15 year-old students were in a state of depression. Globally, it is estimated that 1 in 7 of 10-19 year-olds experience mental health issues. However, because mental health is prone to countless factors, many of those who have mental health issues remain untreated. Studies report that most mental disorders manifest before age 25, typically during the pre-teen and teenage years. Given that adolescents spend the greatest portion of their lives at school during this period, research should focus on the promotion of adolescents’ mental wellbeing within the school environment. A study regarding the relations between school climate and adolescents’ mental health showed that the students’ mental state varied among individuals rather than between schools. Other research indicates that students’ perceived connection to social relations, teacher-student relations, and commitment to school are associated with better mental health. Considering these results together, I seek to highlight the importance of building good interpersonal relationships for students’ mental health.

The research question is “To what extent have US government’s policies to reduce algorithmic bias been successful?”. The aim of this study is to utilise both geographical and economical perspectives to understand whether legislation is able to reduce the economic impacts and inequalities caused by algorithmic bias. This World Studies EE belongs to the theme of Equality and Inequality, since it discusses the global
issue of algorithmic bias, which causes unfair treatment to an individual or group, leading to inequality. This research highlights solutions to algorithmic bias, linking directly to the tenth Sustainable Development Goal, which is to Reduce Inequality.

Self-dehumanization is a consequence of immoral behavior. While it has been under-explored in moral psychology, existing studies by Bastian et al. (“Losing our humanity”)and Kouchaki et al. have reported contradictory findings on self-dehumanization’s implications on morality. This paper aims to consolidate the current literature and present a dual model to explain the psychological processes of this phenomenon. The model hypothesizes that moral self-regulation moderates the effects of self-dehumanization on morality. This makes the success of the regulatory pathway the prime predictor of whether self-dehumanization leads to moral or immoral behavior. The following sections explain the individual processes involved in those moral and immoral pathways. The main arguments are that 1) successful moral self-regulation appeals to our innate desire for self-completion, thus motivating future reparative actions, and 2)unsuccessful moral self-regulation enables disengagement, which leads to future immoral behavior. This can initiate a self-fueling cycle as more failed self-regulation occurs. Together,these hypotheses produce a nuanced, dynamic model that highlights the importance of understanding the role of self-dehumanization in moral psychology.

Cancer has been a prevalent medical concern among many scientists, and within cancer, the specific causes and treatment methods still have a comparatively low recovery rate. MicroRNAs (miRNAs) are endogenous non-coding functional RNAs that regulate gene expression by inhibiting/promoting certain signaling pathways.5 They could be a potential indicator of cancer and can be detected from miRNA screening of patients’ blood samples. This indicator could allow scientists to determine potential cancer victims at a very early stage and begin targeted therapy, or early treatment, which could be what makes the difference between a full recovery and no recovery. In this project, we aim to improve the understanding of gene expression in relation to cancer, using machine learning to identify miRNAs with a high relatedness to cancer and find pathways connected to this relatively novel field.

In this paper, we compare the performances of traditional machine learning models using feature engineering and word vectors and the state-of-the-art language modelBERT using word embeddings on three datasets. We also consider the time and cost efficiency of feature engineering compared to BERT. From our results we conclude that the use of the BERT model was only worth the time and cost trade-off for one of the three datasets we used for comparison, where the BERT model significantly outperformed any kind of traditional classifier that uses feature vectors, instead of embeddings. Using the BERT model for the other datasets only achieved an increase of 0.03 and 0.05 of accuracy and F1 score respectively, which could be argued makes its use not worth the time and cost of GPU.

This study investigates the forecasting accuracies of Long Short-Term Memory models with different architectures and sheds light on the optimal combination of parameters when forecasting stock prices for the S&P500 fund.