This study investigates the influence of overconfidence among Chief Executive Officers (CEOs) and Chief Financial Officers (CFOs) on accounting conservatism through the lens of the false consensus effect. It differentiates between non-discretionary conservatism, which is not attributable to managerial opportunism, and discretionary conservatism, which is subject to managerial judgment. Drawing on a sample of publicly held firms in the United States, the findings indicate that the false consensus effect associated with managerial overconfidence intensifies systematic cognitive bias in non-discretionary conservatism, while diminishing the degree of conservatism in discretionary financial reporting. Collectively, these results underscore the significance of the false consensus effect as a determinant of managerial decision-making processes.
This study explores school-aged children's preferences and biases towards second language (L2) English accents in Taiwan, where exposure to diverse English accents is increasing due to the integration of international non-native English-speaking teachers. Two experiments examined how accent familiarity and ethnicity cues influence children's preferences when listening to pairs of accented voices. A teacher preference task was used, featuring a verbal guise in a neutral condition and two ethnicity-cued conditions (congruent and incongruent). Experiment 1 found that third-grade children (N = 235) preferred the familiar Mandarin-accented English over native American-accented English in the neutral condition, highlighting the role of accent familiarity. Experiment 2 showed a preference for American-accented English over other less familiar non-Mandarin accents, further reinforcing the effect of familiarity. Across both experiments, when ethnicity cues were introduced, children preferred American-accented English more strongly when it was paired with a Caucasian face (congruent), but not with an Asian face (incongruent). Additionally, children with higher English vocabulary levels showed stronger preferences for American-accented English paired with Caucasian faces. These results suggest that children's accent preferences are shaped by both familiarity and ethnicity cues, with biases potentially becoming more pronounced as L2 vocabulary increases.
The limited efficacy of immunotherapy in colorectal cancer (CRC) underscores the need to better define immunosuppressive mechanisms within the tumor microenvironment (TME). We applied single-cell RNA sequencing (scRNA-seq) to characterize stromal–immune interactions shaping the CRC TME. Paired tumor and adjacent normal mucosa samples from eight patients with CRC were analyzed by scRNA-seq. Integrated bioinformatic approaches, including trajectory inference, transcriptional regulon analysis, and cell–cell communication modeling, were used to define cellular differentiation and intercellular signaling. Key findings were validated using The Cancer Genome Atlas datasets, multiplex immunofluorescence staining, and in vitro functional assays. Trajectory analysis demonstrated a progressive increase in triggering receptor expressed on myeloid cells 1 (TREM1) expression during tumor-associated myeloid differentiation. TREM1-positive myeloid cells exhibited enriched M2-like signatures and were preferentially associated with consensus molecular subtype 4 tumors, characterized by stromal activation and poor clinical outcomes. Stromal profiling identified an α-smooth muscle actin (ACTA2)-positive population associated with CRC progression. Spatial analyses revealed close localization of TREM1-positive myeloid cells and ACTA2-positive cancer-associated fibroblasts (CAFs) in tumor tissues. Mechanistically, integrated bioinformatic analyses indicated that TREM1-positive myeloid cells engage CAFs primarily through secreted phosphoprotein 1 (SPP1) signaling, while CAFs reinforce immunosuppressive myeloid phenotypes via TGF-β-associated extracellular matrix pathways. Functional assays showed that TREM1 inhibition reduced SPP1 expression and attenuated M2 macrophage polarization. These findings identify a bidirectional interaction between TREM1-positive myeloid cells and CAFs that contributes to a profibrotic and immunosuppressive CRC microenvironment, highlighting the TREM1–SPP1 axis as a pathway of potential translational relevance.
The integration of artificial intelligence (AI) tools like ChatGPT into education presents significant opportunities to foster interactive and student-centered learning. This study surveyed 61 educators during an online AI workshop to examine their baseline prompt-design skills, explore their perceptions of student use of ChatGPT, and identify associated challenges and strategic opportunities. Importantly, the six-element prompt design framework was deliberately shared only after data collection, ensuring that the study captured prompt-design abilities at a true baseline without prior intervention, which constitutes a novel contribution of this research. Findings reveal that although most educators expressed enthusiasm for ChatGPT's potential, there was notable variability in their ability to design effective prompts. Specifically, over 77% of the initial prompts were vague and lacked essential elements such as context, role, or format, indicating a substantial gap in prompt-engineering competence. Some respondents faced difficulties in crafting contextual and specific prompts, while others demonstrated creative uses of ChatGPT to inspire classroom activities and support student discussions. Additionally, ethical concerns, such as plagiarism, overreliance, and the reliability of ChatGPT-generated answers, emerged as critical issues. The study highlights the urgent need for targeted digital literacy training, prompt-design strategies, and institutional policies that promote the ethical and pedagogically sound use of ChatGPT in both physical and online classrooms. These findings provide a unique and timely foundation for guiding AI adoption in education by underscoring the necessity of equipping educators with tailored digital literacy training and clear ethical guidelines, which are critical for fostering innovative, responsible, and context-sensitive teaching practices.
This study employs a data-driven approach to evaluate player performance in Taiwan's professional basketball league, P. LEAGUE+, and constructs an AI-integrated performance assessment model. Initially, web scraping techniques are used to extract structured and unstructured data - including basic statistics, advanced metrics, and textual content - from the official P. LEAGUE+ website, covering guards, forwards, and centers. Subsequently, a domain-specific AI vector database is established using Retrieval-Augmented Generation (RAG) for data cleansing. The model then integrates Entropy and TOPSIS decision analysis methods to develop an AI-Driven Comprehensive Performance Index for automated performance scoring. TAIDE-LX-7B is further applied for data inference and decision-making to identify the league's Annual First Team. Using the 2021-2022 season data from six teams as a case study, the accuracy and validity of the proposed AI-driven model are verified. Results show that the AI-Driven Comprehensive Performance Index aligns with official postseason selections for guards and centers, while discrepancies in forward selection are attributed to league policies favoring domestic players.
