An AI Tutoring System for Pollinator Conservation Community Science Training

用于传粉媒介保护社区科学培训的人工智能辅导系统

• 批准号: 2303019
• 负责人: Sarath Sreedharan
• 金额: $ 84.99万
• 依托单位: Colorado State University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2023
• 资助国家: 美国
• 起止时间: 2023-09-15 至 2026-08-31
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2303019&HistoricalAwards=false
• 关键词: AI; Tutoring; System; Pollinator; Conservation

Increased access to mobile phones and other internet capable devices has created new opportunities for adult learner audiences to engage in community science programs, which rely on volunteers to provide a large amount of data for analysis. But the lack of scalability of volunteer training is a limitation on community science programs. Volunteers may come to the activity with varying degrees of expertise and different personal goals with respect to the activity. This project seeks to apply explainable artificial intelligence to the challenge of personalizing training for adult citizen scientists. The approach will be developed in the context of the Native Bee Watch (NBW) biodiversity monitoring project that began in 2016 at Colorado State University. NBW trains volunteers to identify and monitor native bees and other pollinators, and educates volunteers about principles and practices from entomology and other fields of ecology and biology. This project will (1) create an online learning environment that will provide adaptive feedback to volunteers to support their self-directed learning, (2) perform controlled user studies to refine the performance of the system, and (3) perform a longitudinal study on the effectiveness of the system for helping volunteers. As a result of helping the volunteers acquire skills and STEM knowledge, the volunteers should, in turn, produce higher-quality observations that can improve the scientific analyses based on the data, and be more likely to continue contributing over time. The project's research questions explore how to structure curricula that can support learners with a wide variety of prior expertise, how to develop algorithms that can estimate the current expertise of a learner while minimizing intrusive assessment tests, how to extend explainable AI to provide feedback tailored to individual learners, how to provide customized suggestions to help learners make better use of the tutoring system to manage and support their learning and citizen science participation, and how the learner expertise estimates can be used to identify areas where the tutoring system may be in need of improved accuracy. Data will be collected via surveys, interviews, observations, focus groups, usage metrics, and performance data. Analyses include thematic analysis of qualitative data and descriptive and comparative statistics. The intellectual merit of the project lies in the exploration of both the pedagogical and algorithmic aspects of supporting adult informal science learning - adults are an under-studied population in informal learning, requiring different approaches to learner modeling and the provision of different kinds of automated feedback than is common in traditional classroom-based intelligent tutoring systems. Because the system is designed to embrace self-driven informal learning, it contains several novel approaches: it will model learners' use of software features to help them manage their mastery of using the software to attain their own (possibly idiosyncratic) learning objectives; and it will be transparent about how its estimate of the learner's skill level was reached and provide learners with the opportunity to better calibrate the estimate - a novel form of dialogue-based evaluation that both respects the agency of the informal learner and improves the system's functionality (both its estimates of learner skill, and the accuracy of the vision model). With respect to broader impacts, while some elements of the developed system will be highly specific to the current domain (pollinator identification), by creating a scalable training system, the user base of the NBW community science project can be radically extended, expanding both the number of adult learners acquiring STEM skills and the amount of quality data being gathered across a wider geographic region. The general approach can in turn be adapted to new citizen science programs, providing a template for how such programs can refactor the fundamental way the program interacts with volunteers, potentially changing the landscape for informal adult science education. Results from this work will be disseminated to a range of educational research, citizen science practitioner, and artificial intelligence venues, enhancing the literatures in all areas.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

移动的电话和其他互联网设备的普及为成年学习者观众参与社区科学项目创造了新的机会，这些项目依靠志愿者提供大量数据进行分析。但是，志愿者培训缺乏可扩展性是社区科学项目的一个限制。参加活动的志愿者可能具有不同程度的专业知识和与活动有关的不同个人目标。该项目旨在将可解释的人工智能应用于成人公民科学家个性化培训的挑战。该方法将在2016年在科罗拉多州立大学开始的土著蜜蜂观察（NBW）生物多样性监测项目的背景下开发。NBW培训志愿者识别和监测本地蜜蜂和其他授粉者，并教育志愿者了解昆虫学和其他生态学和生物学领域的原则和实践。该项目将（1）创建一个在线学习环境，为志愿者提供自适应反馈，以支持他们的自主学习，（2）进行受控用户研究，以改进系统的性能，以及（3）对系统帮助志愿者的有效性进行纵向研究。作为帮助志愿者获得技能和STEM知识的结果，志愿者应该反过来产生更高质量的观察结果，可以改善基于数据的科学分析，并更有可能随着时间的推移继续做出贡献。该项目的研究问题探讨了如何构建课程，以支持具有各种先前专业知识的学习者，如何开发可以估计学习者当前专业知识的算法，同时最大限度地减少侵入性评估测试，如何扩展可解释的人工智能以提供针对个人学习者的反馈，如何提供定制建议，帮助学习者更好地利用辅导系统来管理和支持他们的学习和公民科学参与，以及如何使用学习者专业知识估计来识别辅导系统可能需要提高准确性的领域。将通过调查、访谈、观察、焦点小组、使用指标和性能数据收集数据。分析包括对定性数据的专题分析以及描述性和比较性统计。 该项目的智力价值在于探索支持成人非正式科学学习的教学和算法方面-成人是非正式学习中研究不足的人群，需要不同的学习者建模方法和提供不同类型的自动反馈，而不是传统的基于课堂的智能辅导系统。由于该系统的设计是为了拥抱自我驱动的非正式学习，它包含了几个新的方法：它将模拟学习者使用软件功能，以帮助他们管理他们的掌握使用软件，以实现自己的（可能是特殊的）学习目标;它将是透明的，它对学习者技能水平的估计是如何达到的，并为学习者提供更好地校准估计的机会-一种新形式的基于对话的评价，既尊重非正式学习者的代理，又提高了系统的功能（既评估学习者的技能，又提高了视觉模型的准确性）。关于更广泛的影响，虽然所开发系统的某些元素将高度特定于当前领域（传粉者识别），但通过创建可扩展的培训系统，NBW社区科学项目的用户群可以从根本上扩展，扩大获得STEM技能的成人学习者的数量以及在更广泛的地理区域收集的高质量数据的数量。一般的方法可以反过来适应新的公民科学计划，提供了一个模板，这样的计划可以重构的基本方式与志愿者的程序交互，可能会改变景观的非正式成人科学教育。这项工作的结果将传播到一系列的教育研究，公民科学从业者，和人工智能场所，加强在所有领域的文献。这个奖项反映了NSF的法定使命，并已被认为是值得通过使用基金会的智力价值和更广泛的影响审查标准进行评估的支持。