Frontiers in Artificial Intelligence最新文献

{"title":"Multimodal driver emotion recognition using motor activity and facial expressions.","authors":"Carlos H Espino-Salinas, Huizilopoztli Luna-García, José M Celaya-Padilla, Cristian Barría-Huidobro, Nadia Karina Gamboa Rosales, David Rondon, Klinge Orlando Villalba-Condori","doi":"10.3389/frai.2024.1467051","DOIUrl":"10.3389/frai.2024.1467051","url":null,"abstract":"<p><p>Driving performance can be significantly impacted when a person experiences intense emotions behind the wheel. Research shows that emotions such as anger, sadness, agitation, and joy can increase the risk of traffic accidents. This study introduces a methodology to recognize four specific emotions using an intelligent model that processes and analyzes signals from motor activity and driver behavior, which are generated by interactions with basic driving elements, along with facial geometry images captured during emotion induction. The research applies machine learning to identify the most relevant motor activity signals for emotion recognition. Furthermore, a pre-trained Convolutional Neural Network (CNN) model is employed to extract probability vectors from images corresponding to the four emotions under investigation. These data sources are integrated through a unidimensional network for emotion classification. The main proposal of this research was to develop a multimodal intelligent model that combines motor activity signals and facial geometry images to accurately recognize four specific emotions (anger, sadness, agitation, and joy) in drivers, achieving a 96.0% accuracy in a simulated environment. The study confirmed a significant relationship between drivers' motor activity, behavior, facial geometry, and the induced emotions.</p>","PeriodicalId":33315,"journal":{"name":"Frontiers in Artificial Intelligence","volume":"7 ","pages":"1467051"},"PeriodicalIF":3.0,"publicationDate":"2024-11-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11631879/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142813449","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Toward explainable deep learning in healthcare through transition matrix and user-friendly features.","authors":"Oleksander Barmak, Iurii Krak, Sergiy Yakovlev, Eduard Manziuk, Pavlo Radiuk, Vladislav Kuznetsov","doi":"10.3389/frai.2024.1482141","DOIUrl":"10.3389/frai.2024.1482141","url":null,"abstract":"<p><p>Modern artificial intelligence (AI) solutions often face challenges due to the \"black box\" nature of deep learning (DL) models, which limits their transparency and trustworthiness in critical medical applications. In this study, we propose and evaluate a scalable approach based on a transition matrix to enhance the interpretability of DL models in medical signal and image processing by translating complex model decisions into user-friendly and justifiable features for healthcare professionals. The criteria for choosing interpretable features were clearly defined, incorporating clinical guidelines and expert rules to align model outputs with established medical standards. The proposed approach was tested on two medical datasets: electrocardiography (ECG) for arrhythmia detection and magnetic resonance imaging (MRI) for heart disease classification. The performance of the DL models was compared with expert annotations using Cohen's Kappa coefficient to assess agreement, achieving coefficients of 0.89 for the ECG dataset and 0.80 for the MRI dataset. These results demonstrate strong agreement, underscoring the reliability of the approach in providing accurate, understandable, and justifiable explanations of DL model decisions. The scalability of the approach suggests its potential applicability across various medical domains, enhancing the generalizability and utility of DL models in healthcare while addressing practical challenges and ethical considerations.</p>","PeriodicalId":33315,"journal":{"name":"Frontiers in Artificial Intelligence","volume":"7 ","pages":"1482141"},"PeriodicalIF":3.0,"publicationDate":"2024-11-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11625760/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142802415","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Accuracy improvement in financial sanction screening: is natural language processing the solution?","authors":"Seihee Kim, ShengYun Yang","doi":"10.3389/frai.2024.1374323","DOIUrl":"10.3389/frai.2024.1374323","url":null,"abstract":"<p><p>Sanction screening is a crucial banking compliance process that protects financial institutions from inadvertently engaging with internationally sanctioned individuals or organizations. Given the severe consequences, including financial crime risks and potential loss of banking licenses, effective execution is essential. One of the major challenges in this process is balancing the high rate of false positives, which exceed 90% and lead to inefficiencies due to increased human oversight, with the more critical issue of false negatives, which pose severe regulatory and financial risks by allowing sanctioned entities to go undetected. This study explores the use of Natural Language Processing (NLP) to enhance the accuracy of sanction screening, with a particular focus on reducing false negatives. Using an experimental approach, we evaluated a prototype NLP program on a dataset of sanctioned entities and transactions, assessing its performance in minimising false negatives and understanding its effect on false positives. Our findings demonstrate that while NLP significantly improves sensitivity by detecting more true positives, it also increases false positives, resulting in a trade-off between improved detection and reduced overall accuracy. Given the heightened risks associated with false negatives, this research emphasizes the importance of prioritizing their reduction. The study provides practical insights into how NLP can enhance sanction screening, while recognizing the need for ongoing adaptation to the dynamic nature of the field.</p>","PeriodicalId":33315,"journal":{"name":"Frontiers in Artificial Intelligence","volume":"7 ","pages":"1374323"},"PeriodicalIF":3.0,"publicationDate":"2024-11-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11621073/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142802299","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Active learning with human heuristics: an algorithm robust to labeling bias.","authors":"Sriram Ravichandran, Nandan Sudarsanam, Balaraman Ravindran, Konstantinos V Katsikopoulos","doi":"10.3389/frai.2024.1491932","DOIUrl":"10.3389/frai.2024.1491932","url":null,"abstract":"<p><p>Active learning enables prediction models to achieve better performance faster by adaptively querying an oracle for the labels of data points. Sometimes the oracle is a human, for example when a medical diagnosis is provided by a doctor. According to the behavioral sciences, people, because they employ heuristics, might sometimes exhibit biases in labeling. How does modeling the oracle as a human heuristic affect the performance of active learning algorithms? If there is a drop in performance, can one design active learning algorithms robust to labeling bias? The present article provides answers. We investigate two established human heuristics (fast-and-frugal tree, tallying model) combined with four active learning algorithms (entropy sampling, multi-view learning, conventional information density, and, our proposal, inverse information density) and three standard classifiers (logistic regression, random forests, support vector machines), and apply their combinations to 15 datasets where people routinely provide labels, such as health and other domains like marketing and transportation. There are two main results. First, we show that if a heuristic provides labels, the performance of active learning algorithms significantly drops, sometimes below random. Hence, it is key to design active learning algorithms that are robust to labeling bias. Our second contribution is to provide such a robust algorithm. The proposed inverse information density algorithm, which is inspired by human psychology, achieves an overall improvement of 87% over the best of the other algorithms. In conclusion, designing and benchmarking active learning algorithms can benefit from incorporating the modeling of human heuristics.</p>","PeriodicalId":33315,"journal":{"name":"Frontiers in Artificial Intelligence","volume":"7 ","pages":"1491932"},"PeriodicalIF":3.0,"publicationDate":"2024-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11611880/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142772822","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Vision-language models for medical report generation and visual question answering: a review.","authors":"Iryna Hartsock, Ghulam Rasool","doi":"10.3389/frai.2024.1430984","DOIUrl":"10.3389/frai.2024.1430984","url":null,"abstract":"<p><p>Medical vision-language models (VLMs) combine computer vision (CV) and natural language processing (NLP) to analyze visual and textual medical data. Our paper reviews recent advancements in developing VLMs specialized for healthcare, focusing on publicly available models designed for medical report generation and visual question answering (VQA). We provide background on NLP and CV, explaining how techniques from both fields are integrated into VLMs, with visual and language data often fused using Transformer-based architectures to enable effective learning from multimodal data. Key areas we address include the exploration of 18 public medical vision-language datasets, in-depth analyses of the architectures and pre-training strategies of 16 recent noteworthy medical VLMs, and comprehensive discussion on evaluation metrics for assessing VLMs' performance in medical report generation and VQA. We also highlight current challenges facing medical VLM development, including limited data availability, concerns with data privacy, and lack of proper evaluation metrics, among others, while also proposing future directions to address these obstacles. Overall, our review summarizes the recent progress in developing VLMs to harness multimodal medical data for improved healthcare applications.</p>","PeriodicalId":33315,"journal":{"name":"Frontiers in Artificial Intelligence","volume":"7 ","pages":"1430984"},"PeriodicalIF":3.0,"publicationDate":"2024-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11611889/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142773002","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Investigating generative AI models and detection techniques: impacts of tokenization and dataset size on identification of AI-generated text.","authors":"Haowei Hua, Co-Jiayu Yao","doi":"10.3389/frai.2024.1469197","DOIUrl":"10.3389/frai.2024.1469197","url":null,"abstract":"<p><p>Generative AI models, including ChatGPT, Gemini, and Claude, are increasingly significant in enhancing K-12 education, offering support across various disciplines. These models provide sample answers for humanities prompts, solve mathematical equations, and brainstorm novel ideas. Despite their educational value, ethical concerns have emerged regarding their potential to mislead students into copying answers directly from AI when completing assignments, assessments, or research papers. Current detectors, such as GPT-Zero, struggle to identify modified AI-generated texts and show reduced reliability for English as a Second Language learners. This study investigates detection of academic cheating by use of generative AI in high-stakes writing assessments. Classical machine learning models, including logistic regression, XGBoost, and support vector machine, are used to distinguish between AI-generated and student-written essays. Additionally, large language models including BERT, RoBERTa, and Electra are examined and compared to traditional machine learning models. The analysis focuses on prompt 1 from the ASAP Kaggle competition. To evaluate the effectiveness of various detection methods and generative AI models, we include ChatGPT, Claude, and Gemini in their base, pro, and latest versions. Furthermore, we examine the impact of paraphrasing tools such as GPT-Humanizer and QuillBot and introduce a new method of using synonym information to detect humanized AI texts. Additionally, the relationship between dataset size and model performance is explored to inform data collection in future research.</p>","PeriodicalId":33315,"journal":{"name":"Frontiers in Artificial Intelligence","volume":"7 ","pages":"1469197"},"PeriodicalIF":3.0,"publicationDate":"2024-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11611853/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142772991","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Can AI teach me employability? A multi-national study in three countries.","authors":"Dev Aditya, Krizia Silvestri, Pauldy Cj Otermans","doi":"10.3389/frai.2024.1461158","DOIUrl":"https://doi.org/10.3389/frai.2024.1461158","url":null,"abstract":"<p><p>This paper examines the impact of using an Artificial Intelligence (AI) teacher for current Higher Education (HE) students from three countries. The study utilized an AI avatar powered by a fine-tuned Large Language Model (LLM), OIMISA, which is trained solely for teaching and learning applications. The AI teacher provided a 9-lesson course on employability and transferable skills. In total 207 students across the three institutions enrolled in the programme. The results demonstrate a noteworthy completion rate of over 47%, along with high levels of engagement across all student cohorts and high satisfaction rates from the students. These show the potential for AI-based virtual teachers across countries for students of HE compared to the use of MOOC platforms.</p>","PeriodicalId":33315,"journal":{"name":"Frontiers in Artificial Intelligence","volume":"7 ","pages":"1461158"},"PeriodicalIF":3.0,"publicationDate":"2024-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11609157/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142772990","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The effect of AI on pink marketing: the case of women's purchasing behavior using mobile applications.","authors":"Hasan Beyari","doi":"10.3389/frai.2024.1502580","DOIUrl":"https://doi.org/10.3389/frai.2024.1502580","url":null,"abstract":"<p><p>This research looks in detail at the dynamics of pink marketing and its effect on the purchase behavior of Saudi women through mobile applications, with an emphasis on Artificial Intelligence (AI) as a moderator. Furthermore, this study assesses the effects of customized pink marketing strategies - product, price, promotion, and place - on buying intentions and behaviors. A closed-ended questionnaire was adopted to measure constructs associated with women's mobile app purchase behavior influenced by pink marketing and AI elements. Structural Equation Modeling (SEM) was the study tool used to examine how AI affects women's consumer behavior and how it influences pink marketing. The results suggest that each component of the pink marketing mix significantly influences buying behavior, especially price and promotion. Additionally, AI has a significant moderating effect, improving the personalization and effectiveness of marketing activities. The results of this study highlight the essential role of AI in forming consumer engagement in the digital market, providing useful input for marketers who intend to target women in Saudi Arabia. This study complements the understanding of gender marketing in the digital era and provides a vision for the possibility of AI fundamentally changing traditional approaches.</p>","PeriodicalId":33315,"journal":{"name":"Frontiers in Artificial Intelligence","volume":"7 ","pages":"1502580"},"PeriodicalIF":3.0,"publicationDate":"2024-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11609155/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142772993","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Sequence labeling via reinforcement learning with aggregate labels.","authors":"Marcel Geromel, Philipp Cimiano","doi":"10.3389/frai.2024.1463164","DOIUrl":"https://doi.org/10.3389/frai.2024.1463164","url":null,"abstract":"<p><p>Sequence labeling is pervasive in natural language processing, encompassing tasks such as Named Entity Recognition, Question Answering, and Information Extraction. Traditionally, these tasks are addressed via supervised machine learning approaches. However, despite their success, these approaches are constrained by two key limitations: a common mismatch between the training and evaluation objective, and the resource-intensive acquisition of ground-truth token-level annotations. In this work, we introduce a novel reinforcement learning approach to sequence labeling that leverages aggregate annotations by counting entity mentions to generate feedback for training, thereby addressing the aforementioned limitations. We conduct experiments using various combinations of aggregate feedback and reward functions for comparison, focusing on Named Entity Recognition to validate our approach. The results suggest that sequence labeling can be learned from purely count-based labels, even at the sequence-level. Overall, this count-based method has the potential to significantly reduce annotation costs and variances, as counting entity mentions is more straightforward than determining exact boundaries.</p>","PeriodicalId":33315,"journal":{"name":"Frontiers in Artificial Intelligence","volume":"7 ","pages":"1463164"},"PeriodicalIF":3.0,"publicationDate":"2024-11-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11604619/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142772992","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The impact of AI on education and careers: What do students think?","authors":"Sarah R Thomson, Beverley Ann Pickard-Jones, Stephanie Baines, Pauldy C J Otermans","doi":"10.3389/frai.2024.1457299","DOIUrl":"10.3389/frai.2024.1457299","url":null,"abstract":"<p><strong>Introduction: </strong>Providing one-on-one support to large cohorts is challenging, yet emerging AI technologies show promise in bridging the gap between the support students want and what educators can provide. They offer students a way to engage with their course material in a way that feels fluent and instinctive. Whilst educators may have views on the appropriates for AI, the tools themselves, as well as the novel ways in which they can be used, are continually changing.</p><p><strong>Methods: </strong>The aim of this study was to probe students' familiarity with AI tools, their views on its current uses, their understanding of universities' AI policies, and finally their impressions of its importance, both to their degree and their future careers. We surveyed 453 psychology and sport science students across two institutions in the UK, predominantly those in the first and second year of undergraduate study, and conducted a series of five focus groups to explore the emerging themes of the survey in more detail.</p><p><strong>Results: </strong>Our results showed a wide range of responses in terms of students' familiarity with the tools and what they believe AI tools could and should not be used for. Most students emphasized the importance of understanding how AI tools function and their potential applications in both their academic studies and future careers. The results indicated a strong desire among students to learn more about AI technologies. Furthermore, there was a significant interest in receiving dedicated support for integrating these tools into their coursework, driven by the belief that such skills will be sought after by future employers. However, most students were not familiar with their university's published AI policies.</p><p><strong>Discussion: </strong>This research on pedagogical methods supports a broader long-term ambition to better understand and improve our teaching, learning, and student engagement through the adoption of AI and the effective use of technology and suggests a need for a more comprehensive approach to communicating these important guidelines on an on-going basis, especially as the tools and guidelines evolve.</p>","PeriodicalId":33315,"journal":{"name":"Frontiers in Artificial Intelligence","volume":"7 ","pages":"1457299"},"PeriodicalIF":3.0,"publicationDate":"2024-11-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11602497/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142751809","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}