Fair-AIEd: Participatory design approaches to creating ethical AI for international education and development

世界银行最近警告说，教育领域将出现全球性的“学习危机”，没有学习，教育将无法兑现其消除极端贫困、为所有人创造机会和繁荣的承诺。对此，人工智能(AI)被认为具有通过减少获取教育的障碍、自动化管理流程和优化改善学习结果的方法来加速实现全球教育目标的进程的潜力。为了实现这一愿景，正在建立公共和私人伙伴关系(P3)，在发展中国家推出教育中的人工智能(AIED)倡议，以努力推动教育领域的创新数字转型。然而，虽然人工智能在教育中和为教育带来许多好处，但它也可能带来同样多的挑战，包括社会、政治、经济和伦理后果。虽然人工智能解决方案有相当大的优点，但政策制定者、从业者、行业和其他相关利益攸关方面临的一个重大问题是，没有切实的迹象表明教育中的人工智能将促进所需的转变，也没有证据表明引入这类系统可能会对全球南方学校社区的社会生活产生影响。有鉴于此，公平-艾德项目将作为新兴市场经济体(加纳和南非)的例子，研究在两个非洲背景下P3倡议对在教育中使用人工智能的影响。除了人工智能系统可能对教学和学习产生的影响外，该项目还将调查与企业在发展中国家教育实践中设计和使用人工智能所扮演的领导角色相关的潜在好处、危害和风险。该项目的一个关键成功成果将在于为进一步的研究确定一个基线，以期了解人工智能对全球南方教育的影响。该项目由以下问题指导：RQ1：将人工智能系统嵌入国际教育和发展背景的社会、政治、经济、教育和伦理影响是什么？RQ2：P3伙伴关系如何最有效地引导机器学习，以推动公平的人工智能创新促进国际教育和发展？RQ3：政府如何促进为发展目标创建伦理Aed政策？将使用人种学案例研究来探索潜在的影响。根据现场获得的数据，来自行业、政府、学术界和民间团体的利益相关者将共同设计一种能够对不同人群做出反应的算法影响评估工具。该框架将以FIRE-ML(机器学习)为参照点，确定关键的跨文化价值观和社会问题，根据这些价值观和社会问题，可以确定和评估人工智能在教育中的影响。该工具的设计将结合当地的调整资源，以适应从实地工作中出现的文化、宗教或其他价值差异来源。影响评估将在加纳和南非的六所K-12学校试行，以确定该工具的有效性。将列出、绘制和分析人工智能技术的影响，以说明这一新兴环境中的关键问题和关切，并确定积极教育变革的潜力。在确定了适用于AED的益处、危害和风险后，利益相关者将共同设计和开发公平的信任标志。在发展中国家为其学校选择企业合作伙伴和技术系统时，该标志将对发展中国家的教育领导有用。该工具的开发和使用还将为愿意承诺遵守道德原则的公司提供指导，根据这些原则，将根据本地化软件和了解监管合规性来评判其教育技术提案。

Warning of a global "learning crisis" in education, the World Bank recently claimed that without learning, education will fail to deliver on its promise to eliminate extreme poverty and create opportunity and prosperity for all. In response, Artificial Intelligence (AI) has been proposed as having the potential to accelerate the process of achieving global education goals by reducing barriers to access education, automating management processes, and optimising methods for improvement of learning outcomes. In order to realise this vision, public and private partnerships (P3s) are being established where artificial intelligence in education (AIEd) initiatives are being rolled out in developing countries in an effort to spur innovative digital transformations in education. However, while AI in and for education is claimed to bring many benefits, it also potentially brings as many challenges, including social, political, economic, and ethical consequences. While the AI solution has considerable merit, a significant problem facing policy-makers, practitioners, industry and other relevant stakeholders is that there are no tangible indications that AI in education will promote the desired shifts or evidence of the impact that introducing such systems might have on the social life of Global South school communities.In light of this, the Fair-AIEd project will examine the impact of P3 initiatives on the use of AI in education in two African contexts as examples of emerging market economies (Ghana and South Africa). As well as the implications that AI systems might have for teaching and learning, the project will investigate potential benefits, harms, and risks associated with the leadership roles corporations play in design and use of AI within the educational practices of developing nations. A key successful outcome of this project will lie in establishing a baseline for further research which seeks to understand the impact of AI in education in the Global South. The project is guided by the following questions:RQ1: What are the social, political, economic, pedagogical, and ethical implications of embedding AI systems into international education and development contexts?RQ2: How can P3 partnerships most effectively channel machine learning to drive fair-AI innovation for international education and development?RQ3: How can governments facilitate the creation of ethical AIEd policies for development goals?Potential impacts will be explored using ethnographic case-studies. Informed by data obtained in the field, stakeholders from industry, government, academia, and civic groups will co-design an Algorithmic Impact Assessment tool that can be responsive to diverse populations. Using fair-ML (Machine Learning) as a point of reference, the framework will identify key cross-cultural values and social issues against which the implications of AI in education can be identified and evaluated. Design of the tool will incorporate a local adjustment resource to accommodate cultural, religious, or other sources of value differences that emerge from field work. The Impact Assessment will be piloted in six K-12 schools across Ghana and South Africa to determine the effectiveness of the tool. The impacts of AI technologies will be listed, mapped, and analysed to illustrate key issues and concerns in this emerging landscape and to identify potential for positive educational change. Upon identification of benefits, harms and risks as they apply to AIEd, stakeholders will then co-design and develop a Fair-AIEd Trust Mark. The Mark will be useful for the educational leadership of developing nations as they choose corporate partners and technical systems for their schools. Development and use of this tool will also provide guidance to companies who are willing to commit to the ethical principles upon which their educational technology proposals will be judged in terms of localising software and understanding regulatory compliance.