Artificial Intelligence and Inclusive Education

2020 Springer Nature China New Development Awards

As the world begins to emerge from the first wave of the COVID-19 pandemic, significant patterns of inequality in educational provision are beginning to surface. As various countries imposed social distancing measures, including the closing of schools, a generation of students across the world have been impacted. However, as accounts of ‘education in lockdown’ are revealing, this impact has been experienced very differently, not just as a result of national measures, but also, crucially, in relation to divergent socio-economic conditions within individual countries, as well as diverse approaches by individual schools and teachers. A picture is emerging of vastly dissimilar educational experiences, often following rather tired divisions between the already-privileged and the already-disadvantaged. In the UK, for example, there are calls to prevent the existing ‘disadvantage gap’ from being exacerbated by COVID-19 (EPI 2020). Internationally, the OECD has also proposed a framework to help governments structure their education responses, which particularly stressed the importance of ensuring equity and well-being (Gouëdard et al. 2020).

As governments seek to hastily breathe life back into devastated financial systems across the globe, education will no doubt once again feature centrally in calls for reform that seek to align the sector ever more directly with the economy. As The Economist has recently reported, the UK government is insisting that ‘universities that get covid loans will need to focus more on subjects that either deliver higher wages (such as engineering) or are judged to be particularly important to the country (such as teaching)’, as well as promising ‘reforms to boost vocational education’ (2020, p17). Further, the Australian government has been reported as planning to ‘double the cost of humanities courses while lowering fees for subjects it reckons are in areas of future employment growth (The Economist 2020, p17). However, in the race for skills and competitiveness, the wider project of education is at risk of being overlooked. As Gert Biesta has long argued, the purpose of education is not just a matter of qualification (which one might straightforwardly align with the transmission of the knowledge and skills required to enter employment, in areas assumed to invigorate the economy), but also socialisation and subjectification (Biesta 2009; and for a recent discussion, see Biesta 2020). In other words, while education transfers knowledge, in doing so, whether overtly or not, it also teaches students about, and introduces them into, particular kinds of societies and cultures, as well as forming them as individuals. Clearly there is value in providing people with skills and knowledge that can lead to productive employment, and if this process also contributes to the wider stability of country and its economy, many would view this as an unquestionable success. However, we would do well to acknowledge that the domains of socialisation and subjectification are inescapable; not matter how much we think we are focusing solely on apt qualifications, we are also (re)producing our society and its subjects. As Biesta suggests: ‘those involved in the design and enactment of education — including policymakers and teachers — should always engage with the question of what their efforts seek to bring about in each domain’ (2020, p92), and this seems particularly pressing as we consider a ‘post-pandemic’ education (see Carrigan 2020 for a discussion of this in the context of the university). This is where, despite an imperative to recover quickly from the pandemic, we need to be conscious of the values that we embed (often unconsciously) into our education systems, and of whether the outcomes of education contribute to the sustainable development of our planet and our human society.

Where education is framed as a ‘quick fix’ for national economic woes, the application of artificial intelligence (AI) will likely be continually mooted as an obvious route to enhancement and efficiency in the sector. One of the likely reasons for this is the so-called online ‘pivot’, through which universities and schools are already scrambling to shift their existing teaching programmes onto digital platforms, and in the process creating conditions for the collection of unprecedented volumes of educational data. This data will be seen by many, particularly the private educational companies already involved in the development of data-driven educational technology, as the fuel for a new era of AI-infused education. Indeed, it is the influx of private enterprise into the public education system that is a key issue here. The so-called ‘Ed Tech’ sector is one of the few areas that appear to have benefitted from the pandemic, as the software they supply is now perhaps seen as indispensable for the continuation of teaching and learning where campuses and classrooms remain closed. Venture capital funding in the ‘Ed Tech’ sector this year has been reported at 4.1 billion dollars, representing a five year high for the period January to July (Kunthara 2020). It is also notable that in this period, the AI-based homework and tutoring platform Yuanfudao claimed the largest ever venture capital investment for an ‘Ed Tech’ start-up, totally $1 billion (Dai 2020). The dominant view of AI, particularly so in the commercial sector, is that adoption will be straightforwardly beneficial for both education and private enterprise. As Williamson puts it, the typical view is that ‘AIEd [Artificial Intelligence for Education] will solve complex educational problems while accruing profitable advantage for companies and investors’ (Williamson 2020). 

However, where data-driven technologies, such as AIEd, begin to become more mainstream in education, serious attention needs to be given to the ways such systems shape the everyday practices of teaching, learning, administration, and governance in the sector. As has been argued in education technology research, the simplistic narrative of ‘enhancement’:

carries with it a set of discursive limitations and deeply conservative assumptions which actively limit our capacity to be critical about education and its relation to technology. At the same time, it fails to do justice equally to the disruptive, disturbing and generative dimensions of the academy’s enmeshment with the digital. (Bayne 2015, p7)

The kinds of data collection and processing, as well as the forms of direct and intensive intervention, involved in AIEd, usher in a number of profound shifts in our relationships with technology, that cannot be reduced to a notion of mere ‘enhancement’. Firstly, at the level of data collection, such systems require ever-more widespread and fine-grained documenting of student behaviours, rendering learners ‘visible’ to institutions and their technologies in new, and often misunderstood ways, to which many are oblivious. In this sense, AIEd introduces unprecedented forms of surveillance, which will alter the ways both students and teachers behave under the ‘gaze’ of the institution. Secondly, the processing of this data is often presumed to introduce more objective and precise insights about educational activity. However, as recent criticism of the failed use of algorithmic decision-making to predict student A-level grades in the UK demonstrates (see Amoore 2020), such systems have in-built biases that frequently result in unfair calculations that discriminate and marginalise particular populations. Importantly, both of these issues – the effects of increased surveillance and in-built discrimination - are the kind of crucial insights that educational and sociological perspectives can bring to the technical development of AI.

It is the capacity to automatically intervene in student learning, however, that perhaps distinguishes AIEd specifically from other related kinds of educational technologies. As we discuss in the introduction to our book, Artificial intelligence and Inclusive Education: Speculative Future and Emerging Practices (Knox et al. 2019), one of the most profound potential directions for this technology is in the area of so-called ‘personalisation’; software which supposedly automatically tailors teaching resources and assessments to an individual student. While there may be benefits to students in this individualising approach, particularly for success in assessments, there are also significant concerns about how ‘personalising’ AIEd will radically change educational experiences. As Friesen discusses, despite being a desired ideal in the West, from as far back as the classical era, personalised, one-to-one tuition has been a recurring myth about pedagogical relationships, only ever true for a very privileged minority (2020). Rather, for the vast majority of learners, learning has always been, to some extent at least, a social affair, involving connections to others and shared experiences. Further, teaching, in our contemporary sense of the term, has always had a fundamental communal element, as teachers delicately balance individual needs with collective aims, universal standards, and the group dynamics of classrooms or cohorts. In this sense, an authentic one-to-one pedagogical relationship is incredibly unusual, and certainly untested ground for mainstream education. Returning to Biesta’s three purposes for education (2009; 2020), it is difficult to see how a fully ‘personalised’ educational environment would foster any kind of recognisable or desired socialisation, as students would appear to have little in the way of shared experiences. As for subjectification, it is notable that ‘personalisation’, despite its allusions to the ‘personal’, seems to have little to do with the forming of an individual but very specific ‘learning objectives’ tied to particular subject matter. 

As we argue in the introduction to our book, research into inclusive education offers useful insights for challenging some of the assumptions associated with ‘personalisation’, and also for creatively and critically considering new forms of AIEd. Inclusive education, as an approach to promote all learners’ participation regardless of individual differences, might be seen as method of developing ‘common ground’ within (often increasingly) diverse student groups, and in this sense, it offers a valuable set of principles for thinking about how we might design and develop future AIEd. Where educational institutions, notably universities, begin to offer online and distance provision in the wake of continued social distancing, bringing students together to form important learning communities remains a significant challenge without the comforts of the usual campus or classroom environment. In such a scenario, it seems likely that students may have less opportunity for shared experiences with others, and the incursion of automated teaching in the form of AIEd stands to further entrench the isolation of learners in the post-COVID educational landscape, particularly where individuals do not conform to assumptions about ‘normal’ access to technology at home.

Three key questions for the near future of AIEd development, therefore, are:

•    Where AIEd becomes more mainstream due to increased online learning, how do we ensure that all students, particularly those in need of the support and care that institutions usually provide, benefit from the advantages of the technology?
•    How can AIEd foster inclusive community, social connections, and shared learning experiences, as well as the benefits of ‘personalised’ content delivery, particularly in times of potential isolation?
•    How to ensure meaningful participation of practitioners, students and families in AIEd policy and practice development that supports a more sustainable and socially just future?

We hope that our book might encourage the kind of conversations, from both fields of Artificial Intelligence and Inclusive Education, that will address these questions in context, and we thank our authors for contributing such timely, critical, creative, and engaging chapters that are sure to stimulate further debates in this emerging field of research and practice. Our book is divided into three thematic sections: 1) Artificial Intelligence and Inclusion – opening a dialogue; 2) Emerging Practices; and 3) Critical Perspectives and Speculative Futures. Through these themes, our authors provide an in-depth analysis of current issues, and offer a number of innovative ideas that, we all hope, will tangibly influence a new generation of AIEd development. We also are immensely grateful to the Advanced Innovation Centre for Future Education (AICFE) at Beijing Normal University for supporting our work, as well as the editors Zhongying Shi, Shenquan Yu, Xudong Zhu, and Mang Li for including our work in the excellent Springer series Perspectives on Rethinking and Reforming Education. 

>> Back to the top page

CNPF-HSS-SDG20-Author-Image © Springer Nature

Jeremy Knox

Lecturer in Digital Education at the University of Edinburgh, UK, and Co-Director of the Centre for Research in Digital Education

CNPF-HSS-SDG20-Author-Image © Springer Nature

Yuchen Wang

Research Associate & Tutor at the Institute for Education, Community and Society at the University of Edinburgh, UK

CNPF-HSS-SDG20-Author-Image © Springer Nature

Michael Gallagher

Research Associate at the Centre for Research in Digital Education at the University of Edinburgh, UK, and Director and Co-Founder of a digital education and ICT4D consultancy