SoCS: Collaborative Research: Focusing Attention to Improve the Performance of Citizen Science Systems: Beautiful Images and Perceptive Observers

该项目的目标是开发下一代社会计算公民科学平台，将人类分类器的努力与计算系统的努力结合起来，以最大限度地提高人类注意力的使用效率。处理研究人员面临的海量数字数据是21世纪研究的根本挑战。处理海量数据集的新技术、新工具和新战略，无论是由大量碱基对DNA序列组成的，还是由全天天文调查的TB级数据组成的，都为建立新的科学发现范式提供了机会，但这项任务并不容易。在许多研究领域，数据集的不断增长导致越来越多地采用自动化和无监督的分类方法。在许多情况下，这导致分类质量下降，机器学习和计算机视觉无法复制人类模式识别的成功。网络上公民科学的发展为这个问题提供了一个暂时的解决方案，表明有可能招募数十万名志愿者为结果做出真正的贡献，通过一群分类员的集体智慧促进人类的分析。然而，仅靠人类分类器将不能处理来自未来科学仪器的预期洪流数据。这项研究将由计算机和社会科学家之间的伙伴关系进行，通过系统实施解决自动化数据分析和社会科学方面的研究问题，同时与项目参与者进行实地研究和实验。这个项目的学术价值在于它对促进多个科学领域的知识和理解做出了贡献。首先，这项工作将有助于开发新的计算数据分析方法，最初是对天文图像的分析，后来扩展到其他领域。其次，该项目包括社会科学研究，以测试和应用在线环境中人类动机和学习的理论，然后这些理论可以应用于广泛的社会计算问题。通过将人和计算元素混合在一起，规划的系统有可能改变公民科学的应用及其数据分析方法。该项目将推动科学进步，同时促进教学、培训和学习。其公民科学活动最重要的更广泛的影响之一是使数十万志愿者社区能够参与研究，这是一种强大和迅速发展的非正式科学教育形式。通过选择从天文学开始但不限于该科学领域的相对一般性的图像分类专题，这笔赠款开发的技术将对今后类似研究领域的调查具有重要价值，从而加强研究和教育的基础设施。

The goal of this project is to develop a next-generation socio-computational citizen science platform that combines the efforts of human classifiers with those of computational systems to maximize the efficiency with which human attention can be used. Dealing with the flood of digital data that confronts researchers is the fundamental challenge of twenty-first century research. New techniques, tools and strategies for dealing with massive data sets, whether they consist of vast numbers of base-pair DNA sequences or terabytes of data from all-sky astronomical surveys, present an opportunity to establish a new paradigm of scientific discovery, but the task is not easy. In many areas of research, the relentless growth of data sets has led to the adoption of increasingly automated and unsupervised methods of classification. In many cases, this has led to degradation in classification quality, with machine learning and computer vision unable to replicate the successes of human pattern recognition. The growth of citizen science on the web has provided a temporary solution to this problem, demonstrating that it is possible to recruit hundreds of thousands of volunteers to make an authentic contribution to results, boosting human analysis through the collective wisdom of a crowd of classifiers. However, human classifiers alone will not be able to cope with expected flood of data from future scientific instruments. This research will be carried out by a partnership between computer and social scientists, addressing research problems both in automated data analysis and social science through systems implementation, alongside field research and experiments with project participants. The intellectual merit of this project lies in its contribution to advancing knowledge and understanding in multiple domains of science. First, the work will contribute to developing new methods of computational data analysis, initially with analysis of astronomical images, and later extending to additional fields. Second, the project includes social science research to test and apply theories of human motivation and learning in an online context, which can then be applied to a broad range of social-computational problems. By mixing human and computational elements, the planned system has the potential to transform the application of citizen science and its approach to data analysis. This project will advance science while promoting teaching, training and learning. One of the most significant broader impacts for its citizen science activities is enabling a community of hundreds of thousands of volunteers to participate in research, a powerful and rapidly developing form of informal science education. By choosing the relatively generic topic of image classification, beginning with astronomy but not limited to that field of science, the techniques developed under this grant will be of significant value to future investigations in similar research areas, thus enhancing the infrastructure for research and education.