arXiv - CS - Computers and Society最新文献

{"title":"Beyond One-Time Validation: A Framework for Adaptive Validation of Prognostic and Diagnostic AI-based Medical Devices","authors":"Florian Hellmeier, Kay Brosien, Carsten Eickhoff, Alexander Meyer","doi":"arxiv-2409.04794","DOIUrl":"https://doi.org/arxiv-2409.04794","url":null,"abstract":"Prognostic and diagnostic AI-based medical devices hold immense promise for\u0000advancing healthcare, yet their rapid development has outpaced the\u0000establishment of appropriate validation methods. Existing approaches often fall\u0000short in addressing the complexity of practically deploying these devices and\u0000ensuring their effective, continued operation in real-world settings. Building\u0000on recent discussions around the validation of AI models in medicine and\u0000drawing from validation practices in other fields, a framework to address this\u0000gap is presented. It offers a structured, robust approach to validation that\u0000helps ensure device reliability across differing clinical environments. The\u0000primary challenges to device performance upon deployment are discussed while\u0000highlighting the impact of changes related to individual healthcare\u0000institutions and operational processes. The presented framework emphasizes the\u0000importance of repeating validation and fine-tuning during deployment, aiming to\u0000mitigate these issues while being adaptable to challenges unforeseen during\u0000device development. The framework is also positioned within the current US and\u0000EU regulatory landscapes, underscoring its practical viability and relevance\u0000considering regulatory requirements. Additionally, a practical example\u0000demonstrating potential benefits of the framework is presented. Lastly,\u0000guidance on assessing model performance is offered and the importance of\u0000involving clinical stakeholders in the validation and fine-tuning process is\u0000discussed.","PeriodicalId":501112,"journal":{"name":"arXiv - CS - Computers and Society","volume":"11 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142184061","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Secure Traffic Sign Recognition: An Attention-Enabled Universal Image Inpainting Mechanism against Light Patch Attacks","authors":"Hangcheng Cao, Longzhi Yuan, Guowen Xu, Ziyang He, Zhengru Fang, Yuguang Fang","doi":"arxiv-2409.04133","DOIUrl":"https://doi.org/arxiv-2409.04133","url":null,"abstract":"Traffic sign recognition systems play a crucial role in assisting drivers to\u0000make informed decisions while driving. However, due to the heavy reliance on\u0000deep learning technologies, particularly for future connected and autonomous\u0000driving, these systems are susceptible to adversarial attacks that pose\u0000significant safety risks to both personal and public transportation. Notably,\u0000researchers recently identified a new attack vector to deceive sign recognition\u0000systems: projecting well-designed adversarial light patches onto traffic signs.\u0000In comparison with traditional adversarial stickers or graffiti, these emerging\u0000light patches exhibit heightened aggression due to their ease of implementation\u0000and outstanding stealthiness. To effectively counter this security threat, we\u0000propose a universal image inpainting mechanism, namely, SafeSign. It relies on\u0000attention-enabled multi-view image fusion to repair traffic signs contaminated\u0000by adversarial light patches, thereby ensuring the accurate sign recognition.\u0000Here, we initially explore the fundamental impact of malicious light patches on\u0000the local and global feature spaces of authentic traffic signs. Then, we design\u0000a binary mask-based U-Net image generation pipeline outputting diverse\u0000contaminated sign patterns, to provide our image inpainting model with needed\u0000training data. Following this, we develop an attention mechanism-enabled neural\u0000network to jointly utilize the complementary information from multi-view images\u0000to repair contaminated signs. Finally, extensive experiments are conducted to\u0000evaluate SafeSign's effectiveness in resisting potential light patch-based\u0000attacks, bringing an average accuracy improvement of 54.8% in three widely-used\u0000sign recognition models","PeriodicalId":501112,"journal":{"name":"arXiv - CS - Computers and Society","volume":"26 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142184062","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Automated Journalism","authors":"Wang Ngai Yeung, Tomás Dodds","doi":"arxiv-2409.03462","DOIUrl":"https://doi.org/arxiv-2409.03462","url":null,"abstract":"Developed as a response to the increasing popularity of data-driven\u0000journalism, automated journalism refers to the process of automating the\u0000collection, production, and distribution of news content and other data with\u0000the assistance of computer programs. Although the algorithmic technologies\u0000associated with automated journalism remain in the initial stage of\u0000development, early adopters have already praised the usefulness of automated\u0000journalism for generating routine news based on clean, structured data. Most\u0000noticeably, the Associated Press and The New York Times have been automating\u0000news content to cover financial and sports issues for over a decade.\u0000Nevertheless, research on automated journalism is also alerting to the dangers\u0000of using algorithms for news creation and distribution, including the possible\u0000bias behind AI systems or the human bias of those who develop computer\u0000programs. The popularization of automated news content also has important\u0000implications for the infrastructure of the newsroom, the role performance of\u0000journalists and other non-journalistic professionals, and the distribution of\u0000news content to a datafied audience.","PeriodicalId":501112,"journal":{"name":"arXiv - CS - Computers and Society","volume":"21 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142183889","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Disclosure of AI-Generated News Increases Engagement but Does Not Reduce Aversion, Despite Positive Quality Ratings","authors":"Fabrizio Gilardi, Sabrina Di Lorenzo, Juri Ezzaini, Beryl Santa, Benjamin Streiff, Eric Zurfluh, Emma Hoes","doi":"arxiv-2409.03500","DOIUrl":"https://doi.org/arxiv-2409.03500","url":null,"abstract":"The advancement of artificial intelligence (AI) has led to its application in\u0000many areas, including journalism. One key issue is the public's perception of\u0000AI-generated content. This preregistered study investigates (i) the perceived\u0000quality of AI-assisted and AI-generated versus human-generated news articles,\u0000(ii) whether disclosure of AI's involvement in generating these news articles\u0000influences engagement with them, and (iii) whether such awareness affects the\u0000willingness to read AI-generated articles in the future. We employed a\u0000between-subjects survey experiment with 599 participants from the\u0000German-speaking part of Switzerland, who evaluated the credibility,\u0000readability, and expertise of news articles. These articles were either written\u0000by journalists (control group), rewritten by AI (AI-assisted group), or\u0000entirely generated by AI (AI-generated group). Our results indicate that all\u0000news articles, regardless of whether they were written by journalists or AI,\u0000were perceived to be of equal quality. When participants in the treatment\u0000groups were subsequently made aware of AI's involvement in generating the\u0000articles, they expressed a higher willingness to engage with (i.e., continue\u0000reading) the articles than participants in the control group. However, they\u0000were not more willing to read AI-generated news in the future. These results\u0000suggest that aversion to AI usage in news media is not primarily rooted in a\u0000perceived lack of quality, and that by disclosing using AI, journalists could\u0000attract more immediate engagement with their content, at least in the short\u0000term.","PeriodicalId":501112,"journal":{"name":"arXiv - CS - Computers and Society","volume":"62 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142227784","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"From MOOC to MAIC: Reshaping Online Teaching and Learning through LLM-driven Agents","authors":"Jifan Yu, Zheyuan Zhang, Daniel Zhang-li, Shangqing Tu, Zhanxin Hao, Rui Miao Li, Haoxuan Li, Yuanchun Wang, Hanming Li, Linlu Gong, Jie Cao, Jiayin Lin, Jinchang Zhou, Fei Qin, Haohua Wang, Jianxiao Jiang, Lijun Deng, Yisi Zhan, Chaojun Xiao, Xusheng Dai, Xuan Yan, Nianyi Lin, Nan Zhang, Ruixin Ni, Yang Dang, Lei Hou, Yu Zhang, Xu Han, Manli Li, Juanzi Li, Zhiyuan Liu, Huiqin Liu, Maosong Sun","doi":"arxiv-2409.03512","DOIUrl":"https://doi.org/arxiv-2409.03512","url":null,"abstract":"Since the first instances of online education, where courses were uploaded to\u0000accessible and shared online platforms, this form of scaling the dissemination\u0000of human knowledge to reach a broader audience has sparked extensive discussion\u0000and widespread adoption. Recognizing that personalized learning still holds\u0000significant potential for improvement, new AI technologies have been\u0000continuously integrated into this learning format, resulting in a variety of\u0000educational AI applications such as educational recommendation and intelligent\u0000tutoring. The emergence of intelligence in large language models (LLMs) has\u0000allowed for these educational enhancements to be built upon a unified\u0000foundational model, enabling deeper integration. In this context, we propose\u0000MAIC (Massive AI-empowered Course), a new form of online education that\u0000leverages LLM-driven multi-agent systems to construct an AI-augmented\u0000classroom, balancing scalability with adaptivity. Beyond exploring the\u0000conceptual framework and technical innovations, we conduct preliminary\u0000experiments at Tsinghua University, one of China's leading universities.\u0000Drawing from over 100,000 learning records of more than 500 students, we obtain\u0000a series of valuable observations and initial analyses. This project will\u0000continue to evolve, ultimately aiming to establish a comprehensive open\u0000platform that supports and unifies research, technology, and applications in\u0000exploring the possibilities of online education in the era of large model AI.\u0000We envision this platform as a collaborative hub, bringing together educators,\u0000researchers, and innovators to collectively explore the future of AI-driven\u0000online education.","PeriodicalId":501112,"journal":{"name":"arXiv - CS - Computers and Society","volume":"62 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142223986","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Innovation in Education: Developing and Assessing Gamification in the University of the Philippines Open University Massive Open Online Courses","authors":"Cecille Moldez, Mari Anjeli Crisanto, Ma Gian Rose Cerdeña, Diego S. Maranan, Roberto Figueroa","doi":"arxiv-2409.03309","DOIUrl":"https://doi.org/arxiv-2409.03309","url":null,"abstract":"The University of the Philippines Open University has been at the forefront\u0000of providing Massive Open Online Courses to address knowledge and skill gaps,\u0000aiming to make education accessible and contributing to societal goals.\u0000Recognising challenges in student engagement and completion rates within\u0000Massive Open Online Courses, the authors conducted a study by incorporating\u0000gamification into one of the University of the Philippines Open University's\u0000Massive Open Online Courses to assess its impact on these aspects. Gamification\u0000involves integrating game elements to motivate and engage users. This study\u0000explored the incorporation of Moodle elements such as badges, leaderboards, and\u0000progress bars. Using Moodle analytics, the study also tracked student\u0000engagement, views, and posts throughout the course, offering valuable insights\u0000into the influence of gamification on user behaviour. Furthermore, the study\u0000delved into participant feedback gathered through post-evaluation surveys,\u0000providing a comprehensive understanding of their experiences with the gamified\u0000course design. With a 28.86% completion rate and positive participant\u0000reception, the study concluded that gamification can enhance learner\u0000motivation, participation, and overall satisfaction. This research contributes\u0000to the ongoing discourse on innovative educational methods, positioning\u0000gamification as a promising avenue for creating interactive and impactful\u0000online learning experiences in the Philippines and beyond.","PeriodicalId":501112,"journal":{"name":"arXiv - CS - Computers and Society","volume":"66 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142183891","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Flat-earth communities on Brazilian Telegram: when faith is used to question the existence of gravity as a physics phenomenon","authors":"Ergon Cugler de Moraes Silva","doi":"arxiv-2409.03800","DOIUrl":"https://doi.org/arxiv-2409.03800","url":null,"abstract":"Conspiracy theories related to flat-earthism have gained traction on\u0000Brazilian Telegram, especially in times of global crisis, such as the COVID-19\u0000pandemic, when distrust in scientific and governmental institutions has\u0000intensified. Therefore, this study aims to address the research question: how\u0000are Brazilian conspiracy theory communities on flat earth topics characterized\u0000and articulated on Telegram? It is worth noting that this study is part of a\u0000series of seven studies whose main objective is to understand and characterize\u0000Brazilian conspiracy theory communities on Telegram. This series of seven\u0000studies is openly and originally available on arXiv at Cornell University,\u0000applying a mirrored method across the seven studies, changing only the thematic\u0000object of analysis and providing investigation replicability, including with\u0000proprietary and authored codes, adding to the culture of free and open-source\u0000software. Regarding the main findings of this study, the following were\u0000observed: During the Pandemic, flat-earthist discussions increased by 400%,\u0000driven by distrust in scientific institutions; Flat-Earther communities act as\u0000portals for other conspiracy theories, such as the New World Order; Although\u0000smaller, the flat-Earther network has influential groups that disseminate\u0000content and perpetuate narratives; Religious themes such as God and the Bible\u0000are central, combining religious elements with distrust in science;\u0000Flat-Earther communities use themes such as gravity to challenge established\u0000scientific concepts, reinforcing an alternative view of the world.","PeriodicalId":501112,"journal":{"name":"arXiv - CS - Computers and Society","volume":"12 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142223987","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"AI data transparency: an exploration through the lens of AI incidents","authors":"Sophia Worth, Ben Snaith, Arunav Das, Gefion Thuermer, Elena Simperl","doi":"arxiv-2409.03307","DOIUrl":"https://doi.org/arxiv-2409.03307","url":null,"abstract":"Knowing more about the data used to build AI systems is critical for allowing\u0000different stakeholders to play their part in ensuring responsible and\u0000appropriate deployment and use. Meanwhile, a 2023 report shows that data\u0000transparency lags significantly behind other areas of AI transparency in\u0000popular foundation models. In this research, we sought to build on these\u0000findings, exploring the status of public documentation about data practices\u0000within AI systems generating public concern. Our findings demonstrate that low data transparency persists across a wide\u0000range of systems, and further that issues of transparency and explainability at\u0000model- and system- level create barriers for investigating data transparency\u0000information to address public concerns about AI systems. We highlight a need to\u0000develop systematic ways of monitoring AI data transparency that account for the\u0000diversity of AI system types, and for such efforts to build on further\u0000understanding of the needs of those both supplying and using data transparency\u0000information.","PeriodicalId":501112,"journal":{"name":"arXiv - CS - Computers and Society","volume":"3 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142183892","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Journalists are most likely to receive abuse: Analysing online abuse of UK public figures across sport, politics, and journalism on Twitter","authors":"Liam Burke-Moore, Angus R. Williams, Jonathan Bright","doi":"arxiv-2409.03376","DOIUrl":"https://doi.org/arxiv-2409.03376","url":null,"abstract":"Engaging with online social media platforms is an important part of life as a\u0000public figure in modern society, enabling connection with broad audiences and\u0000providing a platform for spreading ideas. However, public figures are often\u0000disproportionate recipients of hate and abuse on these platforms, degrading\u0000public discourse. While significant research on abuse received by groups such\u0000as politicians and journalists exists, little has been done to understand the\u0000differences in the dynamics of abuse across different groups of public figures,\u0000systematically and at scale. To address this, we present analysis of a novel\u0000dataset of 45.5M tweets targeted at 4,602 UK public figures across 3 domains\u0000(members of parliament, footballers, journalists), labelled using fine-tuned\u0000transformer-based language models. We find that MPs receive more abuse in\u0000absolute terms, but that journalists are most likely to receive abuse after\u0000controlling for other factors. We show that abuse is unevenly distributed in\u0000all groups, with a small number of individuals receiving the majority of abuse,\u0000and that for some groups, abuse is more temporally uneven, being driven by\u0000specific events, particularly for footballers. We also find that a more\u0000prominent online presence and being male are indicative of higher levels of\u0000abuse across all 3 domains.","PeriodicalId":501112,"journal":{"name":"arXiv - CS - Computers and Society","volume":"390 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142183890","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Evaluating the Effects of Digital Privacy Regulations on User Trust","authors":"Mehmet Berk Cetin","doi":"arxiv-2409.02614","DOIUrl":"https://doi.org/arxiv-2409.02614","url":null,"abstract":"In today's digital society, issues related to digital privacy have become\u0000increasingly important. Issues such as data breaches result in misuse of data,\u0000financial loss, and cyberbullying, which leads to less user trust in digital\u0000services. This research investigates the impact of digital privacy laws on user\u0000trust by comparing the regulations in the Netherlands, Ghana, and Malaysia. The\u0000study employs a comparative case study method, involving interviews with\u0000digital privacy law experts, IT educators, and consumers from each country. The\u0000main findings reveal that while the General Data Protection Regulation (GDPR)\u0000in the Netherlands is strict, its practical impact is limited by enforcement\u0000challenges. In Ghana, the Data Protection Act is underutilized due to low\u0000public awareness and insufficient enforcement, leading to reliance on personal\u0000protective measures. In Malaysia, trust in digital services is largely\u0000dependent on the security practices of individual platforms rather than the\u0000Personal Data Protection Act. The study highlights the importance of public\u0000awareness, effective enforcement, and cultural considerations in shaping the\u0000effectiveness of digital privacy laws. Based on these insights, a\u0000recommendation framework is proposed to enhance digital privacy practices, also\u0000aiming to provide valuable guidance for policymakers, businesses, and citizens\u0000in navigating the challenges of digitalization.","PeriodicalId":501112,"journal":{"name":"arXiv - CS - Computers and Society","volume":"3 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142183893","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}