arXiv - CS - Computers and Society最新文献

{"title":"Aligning XAI with EU Regulations for Smart Biomedical Devices: A Methodology for Compliance Analysis","authors":"Francesco Sovrano, Michael Lognoul, Giulia Vilone","doi":"arxiv-2408.15121","DOIUrl":"https://doi.org/arxiv-2408.15121","url":null,"abstract":"Significant investment and development have gone into integrating Artificial\u0000Intelligence (AI) in medical and healthcare applications, leading to advanced\u0000control systems in medical technology. However, the opacity of AI systems\u0000raises concerns about essential characteristics needed in such sensitive\u0000applications, like transparency and trustworthiness. Our study addresses these\u0000concerns by investigating a process for selecting the most adequate Explainable\u0000AI (XAI) methods to comply with the explanation requirements of key EU\u0000regulations in the context of smart bioelectronics for medical devices. The\u0000adopted methodology starts with categorising smart devices by their control\u0000mechanisms (open-loop, closed-loop, and semi-closed-loop systems) and delving\u0000into their technology. Then, we analyse these regulations to define their\u0000explainability requirements for the various devices and related goals.\u0000Simultaneously, we classify XAI methods by their explanatory objectives. This\u0000allows for matching legal explainability requirements with XAI explanatory\u0000goals and determining the suitable XAI algorithms for achieving them. Our\u0000findings provide a nuanced understanding of which XAI algorithms align better\u0000with EU regulations for different types of medical devices. We demonstrate this\u0000through practical case studies on different neural implants, from chronic\u0000disease management to advanced prosthetics. This study fills a crucial gap in\u0000aligning XAI applications in bioelectronics with stringent provisions of EU\u0000regulations. It provides a practical framework for developers and researchers,\u0000ensuring their AI innovations advance healthcare technology and adhere to legal\u0000and ethical standards.","PeriodicalId":501112,"journal":{"name":"arXiv - CS - Computers and Society","volume":"141 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142183918","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":"How will advanced AI systems impact democracy?","authors":"Christopher Summerfield, Lisa Argyle, Michiel Bakker, Teddy Collins, Esin Durmus, Tyna Eloundou, Iason Gabriel, Deep Ganguli, Kobi Hackenburg, Gillian Hadfield, Luke Hewitt, Saffron Huang, Helene Landemore, Nahema Marchal, Aviv Ovadya, Ariel Procaccia, Mathias Risse, Bruce Schneier, Elizabeth Seger, Divya Siddarth, Henrik Skaug Sætra, MH Tessler, Matthew Botvinick","doi":"arxiv-2409.06729","DOIUrl":"https://doi.org/arxiv-2409.06729","url":null,"abstract":"Advanced AI systems capable of generating humanlike text and multimodal\u0000content are now widely available. In this paper, we discuss the impacts that\u0000generative artificial intelligence may have on democratic processes. We\u0000consider the consequences of AI for citizens' ability to make informed choices\u0000about political representatives and issues (epistemic impacts). We ask how AI\u0000might be used to destabilise or support democratic mechanisms like elections\u0000(material impacts). Finally, we discuss whether AI will strengthen or weaken\u0000democratic principles (foundational impacts). It is widely acknowledged that\u0000new AI systems could pose significant challenges for democracy. However, it has\u0000also been argued that generative AI offers new opportunities to educate and\u0000learn from citizens, strengthen public discourse, help people find common\u0000ground, and to reimagine how democracies might work better.","PeriodicalId":501112,"journal":{"name":"arXiv - CS - Computers and Society","volume":"1 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142183927","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":"An evidence-based and critical analysis of the Fediverse decentralization promises","authors":"Henrique S. Xavier","doi":"arxiv-2408.15383","DOIUrl":"https://doi.org/arxiv-2408.15383","url":null,"abstract":"This paper examines the potential of the Fediverse, a federated network of\u0000social media and content platforms, to counter the centralization and dominance\u0000of commercial platforms on the social Web. We gather evidence from the\u0000technology powering the Fediverse (especially the ActivityPub protocol),\u0000current statistical data regarding Fediverse user distribution over instances,\u0000and the status of two older, similar, decentralized technologies: e-mail and\u0000the Web. Our findings suggest that Fediverse will face significant challenges\u0000in fulfilling its decentralization promises, potentially hindering its ability\u0000to positively impact the social Web on a large scale.","PeriodicalId":501112,"journal":{"name":"arXiv - CS - Computers and Society","volume":"56 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142183916","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":"Urban context and delivery performance: Modelling service time for cargo bikes and vans across diverse urban environments","authors":"Maxwell Schrader, Navish Kumar, Esben Sørig, Soonmyeong Yoon, Akash Srivastava, Kai Xu, Maria Astefanoaei, Nicolas Collignon","doi":"arxiv-2409.06730","DOIUrl":"https://doi.org/arxiv-2409.06730","url":null,"abstract":"Light goods vehicles (LGV) used extensively in the last mile of delivery are\u0000one of the leading polluters in cities. Cargo-bike logistics and Light Electric\u0000Vehicles (LEVs) have been put forward as a high impact candidate for replacing\u0000LGVs. Studies have estimated over half of urban van deliveries being\u0000replaceable by cargo-bikes, due to their faster speeds, shorter parking times\u0000and more efficient routes across cities. However, the logistics sector suffers\u0000from a lack of publicly available data, particularly pertaining to cargo-bike\u0000deliveries, thus limiting the understanding of their potential benefits.\u0000Specifically, service time (which includes cruising for parking, and walking to\u0000destination) is a major, but often overlooked component of delivery time\u0000modelling. The aim of this study is to establish a framework for measuring the\u0000performance of delivery vehicles, with an initial focus on modelling service\u0000times of vans and cargo-bikes across diverse urban environments. We introduce\u0000two datasets that allow for in-depth analysis and modelling of service times of\u0000cargo bikes and use existing datasets to reason about differences in delivery\u0000performance across vehicle types. We introduce a modelling framework to predict\u0000the service times of deliveries based on urban context. We employ Uber's H3\u0000index to divide cities into hexagonal cells and aggregate OpenStreetMap tags\u0000for each cell, providing a detailed assessment of urban context. Leveraging\u0000this spatial grid, we use GeoVex to represent micro-regions as points in a\u0000continuous vector space, which then serve as input for predicting vehicle\u0000service times. We show that geospatial embeddings can effectively capture urban\u0000contexts and facilitate generalizations to new contexts and cities. Our\u0000methodology addresses the challenge of limited comparative data available for\u0000different vehicle types within the same urban settings.","PeriodicalId":501112,"journal":{"name":"arXiv - CS - Computers and Society","volume":"23 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142183926","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":"Measuring Human Contribution in AI-Assisted Content Generation","authors":"Yueqi Xie, Tao Qi, Jingwei Yi, Ryan Whalen, Junming Huang, Qian Ding, Yu Xie, Xing Xie, Fangzhao Wu","doi":"arxiv-2408.14792","DOIUrl":"https://doi.org/arxiv-2408.14792","url":null,"abstract":"With the growing prevalence of generative artificial intelligence (AI), an\u0000increasing amount of content is no longer exclusively generated by humans but\u0000by generative AI models with human guidance. This shift presents notable\u0000challenges for the delineation of originality due to the varying degrees of\u0000human contribution in AI-assisted works. This study raises the research\u0000question of measuring human contribution in AI-assisted content generation and\u0000introduces a framework to address this question that is grounded in information\u0000theory. By calculating mutual information between human input and AI-assisted\u0000output relative to self-information of AI-assisted output, we quantify the\u0000proportional information contribution of humans in content generation. Our\u0000experimental results demonstrate that the proposed measure effectively\u0000discriminates between varying degrees of human contribution across multiple\u0000creative domains. We hope that this work lays a foundation for measuring human\u0000contributions in AI-assisted content generation in the era of generative AI.","PeriodicalId":501112,"journal":{"name":"arXiv - CS - Computers and Society","volume":"41 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142223992","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":"Properties of Effective Information Anonymity Regulations","authors":"Aloni Cohen, Micah Altman, Francesca Falzon, Evangelina Anna Markatou, Kobbi Nissim","doi":"arxiv-2408.14740","DOIUrl":"https://doi.org/arxiv-2408.14740","url":null,"abstract":"A firm seeks to analyze a dataset and to release the results. The dataset\u0000contains information about individual people, and the firm is subject to some\u0000regulation that forbids the release of the dataset itself. The regulation also\u0000imposes conditions on the release of the results. What properties should the\u0000regulation satisfy? We restrict our attention to regulations tailored to\u0000controlling the downstream effects of the release specifically on the\u0000individuals to whom the data relate. A particular example of interest is an\u0000anonymization rule, where a data protection regulation limiting the disclosure\u0000of personally identifiable information does not restrict the distribution of\u0000data that has been sufficiently anonymized. In this paper, we develop a set of technical requirements for anonymization\u0000rules and related regulations. The requirements are derived by situating within\u0000a simple abstract model of data processing a set of guiding general principles\u0000put forth in prior work. We describe an approach to evaluating such regulations\u0000using these requirements -- thus enabling the application of the general\u0000principles for the design of mechanisms. As an exemplar, we evaluate competing\u0000interpretations of regulatory requirements from the EU's General Data\u0000Protection Regulation.","PeriodicalId":501112,"journal":{"name":"arXiv - CS - Computers and Society","volume":"18 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142183919","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":"Multi-Class Plant Leaf Disease Detection: A CNN-based Approach with Mobile App Integration","authors":"Md Aziz Hosen Foysal, Foyez Ahmed, Md Zahurul Haque","doi":"arxiv-2408.15289","DOIUrl":"https://doi.org/arxiv-2408.15289","url":null,"abstract":"Plant diseases significantly impact agricultural productivity, resulting in\u0000economic losses and food insecurity. Prompt and accurate detection is crucial\u0000for the efficient management and mitigation of plant diseases. This study\u0000investigates advanced techniques in plant disease detection, emphasizing the\u0000integration of image processing, machine learning, deep learning methods, and\u0000mobile technologies. High-resolution images of plant leaves were captured and\u0000analyzed using convolutional neural networks (CNNs) to detect symptoms of\u0000various diseases, such as blight, mildew, and rust. This study explores 14\u0000classes of plants and diagnoses 26 unique plant diseases. We focus on common\u0000diseases affecting various crops. The model was trained on a diverse dataset\u0000encompassing multiple crops and disease types, achieving 98.14% accuracy in\u0000disease diagnosis. Finally integrated this model into mobile apps for real-time\u0000disease diagnosis.","PeriodicalId":501112,"journal":{"name":"arXiv - CS - Computers and Society","volume":"4 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142183917","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 Large Language Models on Spatial Tasks: A Multi-Task Benchmarking Study","authors":"Liuchang Xu Shuo Zhao, Qingming Lin, Luyao Chen, Qianqian Luo, Sensen Wu, Xinyue Ye, Hailin Feng, Zhenhong Du","doi":"arxiv-2408.14438","DOIUrl":"https://doi.org/arxiv-2408.14438","url":null,"abstract":"The advent of large language models such as ChatGPT, Gemini, and others has\u0000underscored the importance of evaluating their diverse capabilities, ranging\u0000from natural language understanding to code generation. However, their\u0000performance on spatial tasks has not been comprehensively assessed. This study\u0000addresses this gap by introducing a novel multi-task spatial evaluation\u0000dataset, designed to systematically explore and compare the performance of\u0000several advanced models on spatial tasks. The dataset encompasses twelve\u0000distinct task types, including spatial understanding and path planning, each\u0000with verified, accurate answers. We evaluated multiple models, including\u0000OpenAI's gpt-3.5-turbo, gpt-4o, and ZhipuAI's glm-4, through a two-phase\u0000testing approach. Initially, we conducted zero-shot testing, followed by\u0000categorizing the dataset by difficulty and performing prompt tuning tests.\u0000Results indicate that gpt-4o achieved the highest overall accuracy in the first\u0000phase, with an average of 71.3%. Although moonshot-v1-8k slightly\u0000underperformed overall, it surpassed gpt-4o in place name recognition tasks.\u0000The study also highlights the impact of prompt strategies on model performance\u0000in specific tasks. For example, the Chain-of-Thought (COT) strategy increased\u0000gpt-4o's accuracy in path planning from 12.4% to 87.5%, while a one-shot\u0000strategy enhanced moonshot-v1-8k's accuracy in mapping tasks from 10.1% to\u000076.3%.","PeriodicalId":501112,"journal":{"name":"arXiv - CS - Computers and Society","volume":"416 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142183922","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":"Social perception of faces in a vision-language model","authors":"Carina I. Hausladen, Manuel Knott, Colin F. Camerer, Pietro Perona","doi":"arxiv-2408.14435","DOIUrl":"https://doi.org/arxiv-2408.14435","url":null,"abstract":"We explore social perception of human faces in CLIP, a widely used\u0000open-source vision-language model. To this end, we compare the similarity in\u0000CLIP embeddings between different textual prompts and a set of face images. Our\u0000textual prompts are constructed from well-validated social psychology terms\u0000denoting social perception. The face images are synthetic and are\u0000systematically and independently varied along six dimensions: the legally\u0000protected attributes of age, gender, and race, as well as facial expression,\u0000lighting, and pose. Independently and systematically manipulating face\u0000attributes allows us to study the effect of each on social perception and\u0000avoids confounds that can occur in wild-collected data due to uncontrolled\u0000systematic correlations between attributes. Thus, our findings are experimental\u0000rather than observational. Our main findings are three. First, while CLIP is\u0000trained on the widest variety of images and texts, it is able to make\u0000fine-grained human-like social judgments on face images. Second, age, gender,\u0000and race do systematically impact CLIP's social perception of faces, suggesting\u0000an undesirable bias in CLIP vis-a-vis legally protected attributes. Most\u0000strikingly, we find a strong pattern of bias concerning the faces of Black\u0000women, where CLIP produces extreme values of social perception across different\u0000ages and facial expressions. Third, facial expression impacts social perception\u0000more than age and lighting as much as age. The last finding predicts that\u0000studies that do not control for unprotected visual attributes may reach the\u0000wrong conclusions on bias. Our novel method of investigation, which is founded\u0000on the social psychology literature and on the experiments involving the\u0000manipulation of individual attributes, yields sharper and more reliable\u0000observations than previous observational methods and may be applied to study\u0000biases in any vision-language model.","PeriodicalId":501112,"journal":{"name":"arXiv - CS - Computers and Society","volume":"67 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142183924","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":"The Impact of Group Discussion and Formation on Student Performance: An Experience Report in a Large CS1 Course","authors":"Tong Wu, Xiaohang Tang, Sam Wong, Xi Chen, Clifford A. Shaffer, Yan Chen","doi":"arxiv-2408.14610","DOIUrl":"https://doi.org/arxiv-2408.14610","url":null,"abstract":"Programming instructors often conduct collaborative learning activities, such\u0000as Peer Instruction (PI), to enhance student motivation, engagement, and\u0000learning gains. However, the impact of group discussion and formation\u0000mechanisms on student performance remains unclear. To investigate this, we\u0000conducted an 11-session experiment in a large, in-person CS1 course. We\u0000employed both random and expertise-balanced grouping methods to examine the\u0000efficacy of different group mechanisms and the impact of expert students'\u0000presence on collaborative learning. Our observations revealed complex dynamics\u0000within the collaborative learning environment. Among 255 groups, 146 actively\u0000engaged in discussions, with 96 of these groups demonstrating improvement for\u0000poor-performing students. Interestingly, our analysis revealed that different\u0000grouping methods (expertise-balanced or random) did not significantly influence\u0000discussion engagement or poor-performing students' improvement. In our deeper\u0000qualitative analysis, we found that struggling students often derived benefits\u0000from interactions with expert peers, but this positive effect was not\u0000consistent across all groups. We identified challenges that expert students\u0000face in peer instruction interactions, highlighting the complexity of\u0000leveraging expertise within group discussions.","PeriodicalId":501112,"journal":{"name":"arXiv - CS - Computers and Society","volume":"49 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142183969","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}