譯題：教育中的生成式人工智能——教育者與專家觀點(diǎn)（一）

Product by The Open Innovation Team and Department for Education" 英國政府開放創(chuàng)新團(tuán)隊(duì)及英國教育部

Introduction

Over the last year， interest in and use of generative artificial intelligence （GenAI） has rapidly increased. GenAI uses foundation models， including large language models （LLMs）， trained on large volumes of data. Notable GenAI foundation models are OpenAI’s GPT-3.5 and GPT-4， which underpin the chatbots ChatGPT and Bing Chat.These tools can be used to produce artificially generated content such as text， audio， code， images and videos. Other examples of GenAI tools include Google Bard， Claude and Midjourney. This technology is also increasingly being integrated within other digital tools.

Although GenAI is not new， recent advances in the underlying technology and greater accessibility mean that the public can now use it more easily. This poses opportunities and challenges for the education sector.

The Digital Strategy Division in the Department for Education （DfE） asked HM Government’s Open Innovation Team （OIT） to explore the opportunities and risks for GenAI in education.

This report contains insights from interviews with teachers and educators at 23 educational institutions， 14 interviews with experts from academia and the education technology （EdTech） industry， a range of quantitative data sources， and key themes from academic and grey literature. We have also drawn on the DfE’s Call for Evidence summary of responses where relevant.

The report covers：

？誗How the sector has responded to and adopted GenAI technology.

？誗Applications and opportunities for GenAI in education.

？誗Reported impact and benefits of GenAI use in education.

？誗Barriers to adoption and risks that GenAI presents for education.

？誗Support the sector would like to receive from the DfE and government.

Key findings

Teachers and experts acknowledge that GenAI could have a transformative impact on education. From helping teachers save time by automating tasks， to improving teaching effectiveness by personalising learning for students， there is significant potential for GenAI to benefit the sector. At the same time， there is considerable concern about the risks it presents， as well as scepticism about whether these can be mitigated.

Use of GenAI among teachers and students has rapidly increased over the last year. By November 2023， 42% of primary and secondary teachers had used GenAI in their role （an increase from 17% in April）. Pupils and students may be using GenAI more than their teachers. 74% percent of online 16-24 year olds in the UK have used a GenAI tool.

Evidence on GenAI use in an educational context shows figures ranging from 14% to 67% of students having used GenAI for schoolwork and studies.

Educators are already realising the benefits of using GenAI， namely in helping them save time by creating lesson resources， plan lessons and streamline administrative processes. Reported benefits for students include increased engagement and better support for students with special educational needs and disabilities （SEND） by personalising learning materials.

There is widespread recognition of the risks GenAI presents for education. There is considerable concern around GenAI-enabled academic malpractice， student overreliance on GenAI， as well as ethical， safety and data privacy risks of use. The potential for GenAI to widen educational inequalities was raised. There is also concern that the benefits of GenAI for education will never be fully realised due to barriers to adoption. Lack of knowledge about how to use GenAI， and poor digital skills and infrastructure are limiting further use among teachers. Negative media coverage affects some teachers’ perceptions of GenAI， as well as raises concern about the threat of AI to teacher job security.

There is appetite for government support to ensure GenAI adoption in education is safe， effective， and aligns with good pedagogy. There is little robust evidence on the impact of GenAI in education， with experts emphasising the need to build an impartial evidence base to better understand the impact of GenAI tools on education. Suggestions for the role of government included sharing guidance and best practice and ensuring student data protection and privacy. As the sector adapts to GenAI， experts and educators also highlight the need for a longer-term strategy for artificial intelligence （AI） in education that is future-proofed to keep pace with the evolving nature of this technology.

譯文

簡介

在過去一年里，人們對生成式人工智能的興趣和使用頻次迅速增加。生成式人工智能使用基礎(chǔ)模型，包括大型語言模型，對大量數(shù)據(jù)進(jìn)行了訓(xùn)練。最讓人耳熟能詳?shù)纳墒饺斯ぶ悄芑A(chǔ)模型是OpenAI設(shè)計(jì)的GPT-3.5和GPT-4，它們是聊天機(jī)器人ChatGPT和“必應(yīng)聊天”的基礎(chǔ)模型。生成式人工智能工具如谷歌Bard、 Claude和Midjourney，可以生成各種內(nèi)容，如文本、音頻、代碼、圖像和視頻。生成式人工智能技術(shù)也被越來越多地集成到其他數(shù)字工具中。

雖然生成式人工智能并不是什么新鮮事物，但最近基礎(chǔ)技術(shù)的進(jìn)步和更大的可訪問性意味著公眾現(xiàn)在可以更容易接觸和使用它。這也給教育部門帶來了機(jī)遇和挑戰(zhàn)。

英國教育部的數(shù)字戰(zhàn)略部門要求英國政府的開放創(chuàng)新團(tuán)隊(duì)探索生成式人工智能在教育中的機(jī)遇和風(fēng)險(xiǎn)。

本報(bào)告包含了對23個(gè)教育機(jī)構(gòu)的教師和教育工作者以及來自學(xué)術(shù)界和教育技術(shù)行業(yè)的14個(gè)專家的訪談結(jié)果，一系列定量數(shù)據(jù)來源，來自學(xué)術(shù)和灰色文獻(xiàn)的關(guān)鍵主題，等等。我們還借鑒了英國教育部門的相關(guān)回應(yīng)證據(jù)摘要。

該報(bào)告主要涵蓋了：

教育界如何應(yīng)對和采用生成式人工智能技術(shù)。

生成式人工智能在教育中的應(yīng)用和機(jī)會(huì)。

生成式人工智能在教育中使用的影響和好處。

生成式人工智能給教育帶來的適應(yīng)障礙和風(fēng)險(xiǎn)。

該行業(yè)希望得到教育部和政府的支持。

主要調(diào)查結(jié)果

教育者和相關(guān)專家們都認(rèn)為，生成式人工智能可能會(huì)對教育產(chǎn)生變革性的影響。從通過自動(dòng)化任務(wù)幫助教師節(jié)省時(shí)間，到通過為學(xué)生提供個(gè)性化學(xué)習(xí)服務(wù)來提高學(xué)習(xí)效率，可以看出生成式人工智能對教育有巨大的應(yīng)用潛力和益處。與此同時(shí)，人們對它所帶來的風(fēng)險(xiǎn)表示了相當(dāng)大的擔(dān)憂，以及對能否減輕這些風(fēng)險(xiǎn)表示懷疑。

在過去的一年中，使用生成式人工智能的教師和學(xué)生人數(shù)迅速增加。到2023年11月，42%的中小學(xué)教師使用了生成式人工智能課程（較同年4月份的17%來說有所增加）。學(xué)生可能比教師更多地使用生成式人工智能。在英國，74%的16～24歲在線用戶使用過生成式人工智能工具。

有關(guān)在教育環(huán)境中使用生成式人工智能的數(shù)據(jù)顯示，14%～67%的學(xué)生在完成作業(yè)和日常學(xué)習(xí)中使用生成式人工智能。

教育工作者們已經(jīng)意識(shí)到使用生成式人工智能的好處，如生成式人工智能能夠通過創(chuàng)建課程資源、對課程進(jìn)行計(jì)劃和安排、簡化行政流程來幫助教師節(jié)省工作時(shí)間。據(jù)報(bào)道，生成式人工智能對學(xué)生的好處包括，通過個(gè)性化的學(xué)習(xí)材料提高了學(xué)生的參與度，并為有特殊教育需求和殘疾的學(xué)生提供了更好的支持。

人們普遍認(rèn)識(shí)到生成式人工智能可能會(huì)對教育帶來風(fēng)險(xiǎn)，如學(xué)術(shù)不端、學(xué)生過度依賴生成式人工智能以及倫理、安全和數(shù)據(jù)隱私等方面的風(fēng)險(xiǎn)，并為此感到相當(dāng)擔(dān)憂。有人擔(dān)心生成式人工智能擴(kuò)大了教育不平等的可能性。還有人擔(dān)心由于在教育中采用方面的障礙，生成式人工智能對教育的好處永遠(yuǎn)無法充分實(shí)現(xiàn)。另外，缺乏關(guān)于如何正確使用生成式人工智能的知識(shí)，以及糟糕的數(shù)字技能和基礎(chǔ)設(shè)施，限制了教師進(jìn)一步使用生成式人工智能。還有一些負(fù)面的媒體報(bào)道影響了一些教師對生成式人工智能的看法，同時(shí)也引發(fā)了人們對人工智能會(huì)取代教師工作的擔(dān)憂。

人們希望得到政府的支持，以確保在教育中采用的生成式人工智能是安全的、有效的，并與良好的教育方法相一致。關(guān)于生成式人工智能在教育中的影響，幾乎沒有強(qiáng)有力的證據(jù)，因此專家們強(qiáng)調(diào)需要建立一個(gè)公正的證據(jù)基礎(chǔ)，以便更好地了解生成式人工智能工具對教育的影響。專家們還建議政府能夠?qū)處熯M(jìn)行指導(dǎo)并分享最佳實(shí)踐的成果，以及確保學(xué)生的數(shù)據(jù)安全和隱私。隨著教育行業(yè)適應(yīng)生成式人工智能，專家和教育工作者認(rèn)為，需要制定一項(xiàng)面向未來的人工智能教育長期戰(zhàn)略，以便跟上這項(xiàng)技術(shù)不斷發(fā)展的步伐。