1 Introduction

Epistemic insight seeks to develop metacognition in order to develop the quality of self-awareness and knowledge of knowledge necessary to be able to bring different subjects into conversation with one another (Billingsley & Ramos Arias, 2017). One goal of teaching for epistemic insight is to create understanding of knowledge production (Billingsley & Fraser, 2018). The manner of producing new knowledge will be influenced by the presence of AI tools and thus teaching for epistemic insight will need to respond. Our aim in this paper is to begin a conversation with those involved with education for epistemic insight to make visible the importance of ‘knower awareness’ (Blackie, 2022). The advent of generative AI has brought into focus the problem of the separation of knowledge and the knower. Herein we focus on the opportunity afforded by the emergence of AI tools to consider the implications for higher education.

Generative artificial intelligence toolsFootnote 1 (herein AI tools) are here to stay (Celik, 2023). They will have substantial impact on society in the coming years. Although many of these tools have been around in some form for several years, higher education had remained relatively impervious to them (Luckin & Cukurova, 2019) until the release of ChatGPT-3 to the public in November 2022. There are various different kinds of AI tools. ChatGPT is part of a group of AI tools called Large Language Models (LLMs) that has created an flurry of activity in higher education in the last year. There is simply no way that higher education institutions can ignore their presence. There are two important reasons for this. Firstly, because many professional industries will be deeply impacted by the presence of AI tools—medicine and law to name just two. The role of a primary health care provider or general practitioner could largely be taken over by a fine-tuned language model (Yang et al., 2022). As diagnosis is essentially a form of pattern recognition, this function of a medical practitioner can potentially be achieved by an algorithm. Likewise, there are already AI chatbots generating legal documents in the United States (Surden, 2019). Jobs that have been considered highly skilled professions in which the primary task can be outsourced to an algorithm, will no longer hold their current cachet.

If we accept Stiegler’s (2019) argument that technological innovation drives economic change and disrupts social and cultural practices which thereafter are obliged to adapt until the new technology becomes a taken-for-granted cultural fact (Stiegler, 2019), it follows that higher education will have to adapt to the AI innovation. Now that generative AI is accessible in the public domain, those who work in higher education are obliged to reconsider the value, purposes and practices of higher education.

At one level, we need to recognise that higher education is training students for an unknown future (Kramm & McKenna, 2023). We have not yet, and cannot yet, grasp the ways in which knowledge building will change when AI tools are fully integrated into society. At a second level, there is a strong narrative of the value of higher education in terms of human capital development as mostly a private but also a public good (McArthur, 2023). In a globalised digital economy, supported by neo-liberal ideology, social institutions, including universities, are increasingly run on business models underpinned by market values. Higher education is now valued to the extent that it serves extrinsic purposes such as the production of employable graduates, especially professionals, and the development of national economies through scientific and technological innovation (Ashwin, 2020). These extrinsic drivers are crowding out old intrinsic purposes such as the generation and transmission of knowledge in order to transform students into critical, capable social actors. At a third level, higher education involves the production of new knowledge in a way that is not nearly as visible in either primary or secondary education. So the impact of AI tools on knowledge creation affects higher education to a much greater degree.

A more immediate issue created by AI is that traditional assessment tasks, such as writing an essay, can no longer uncritically be assumed to be evidence of learning. It is this issue that has caught the attention of most academics. Swiecki et al. (2022) point out the ‘standard assessment paradigm’ has serious limitations. Most significantly they assert the problem of a single measure or threshold being applied uniformly across a cohort. In higher education, we have taken the capacity to produce a particular artefact of reasonable quality as evidence for learning. With the availability of AI tools, the assumption that the measurement of standardised tasks is an ‘objective’ and valid means of assessing individual learning can no longer hold (Swiecki et al., 2022). The capacity to produce an artefact cannot be taken as proxy of evidence that students have changed ‘their understanding of themselves and their place in the world’ (Ashwin, 2020, 3). Whilst there is more to be said on this topic, our point here is simply to note that rethinking assessment methods in higher education is long overdue and, in this regard, the ‘threat’ of AI is to be welcomed (McKenna et al., 2023).

Whilst we fully acknowledge the challenge of valid assessments, our purpose here is to argue that there is in fact a deeper issue that is brought into view with the emergence of generative AI tools. Over the past century how we understand the purpose and value of education has been deeply shaped by the use of mechanical metaphors derived from machines and computers to describe human functioning; for example, the measurement of teaching and research via counts of graduates and outputs. When these metaphors are used, the human is likened, and thereby reduced to, an information processing organism. The problem we now face is that generative AI is significantly faster than the human brain and less prone to distraction and confirmation bias (Fry, 2018).Footnote 2 In other words, for functions such as information retrieval and pattern recognition, generative AI significantly outperforms humans. Whilst the mechanical metaphors long predate AI tools, the problem with these metaphors is made sharply visible by the emergence thereof. We need therefore to seriously (re)consider the source and consequence of using such metaphors for our understanding of human cognition, knowledge practices and the purposes of higher education. This line of inquiry is congruent with the notion of ‘epistemic insight’.

Below we develop and justify our argument that higher education will only meet the challenges posed to it by the emergence of generative AI tools if we rethink the purposes and values of higher education and correct the assumptions and practices currently based on the ‘human as machine’ metaphor. In the first section, we support our argument by citing thinkers in neuroscience, education and science whose works suggest that our current models of cognition, education and scientific practice are reductive and inadequate. In section two, we introduce the work of McGilchrist to argue that we need to recover holistic and relational ways of thinking that have been neglected by modernity’s emphasis on analytic forms of thought. The importance of the foundational metaphors we use to conceptualise education is supported by the work of Stein (2019) who contrasts ‘mind-as-computer’ with ‘mind-as-ecosystem’. Further, we refer to the insights of Prigogine and Stengers (1985) and Stengers (1997) to suggest that this more holistic view of cognition is in keeping with their critique of the universalised, abstracted discourse of modern science. In section three, we introduce the reader to the debate on ‘powerful knowledge’ from sociology of education and higher education studies. Whilst supporting a realist view of knowledge, we suggest that the advocates of this position may be working with an idealised and out-dated view of scientific practice. In section four, we discuss the importance of re-inserting the knower into knowledge claims. We argue that the pretention to objectivity and universalism in modern scientific rationality is problematic because it obscures the political and ethical consequences of knowledge production. Finally in section five, we summarise the key premises and principles of our argument for future higher education practices that might address the challenges presented by AI tools. We argue that we should welcome these challenges as an opportunity to rethink and update our founding metaphors and ideas about how education, and science education in particular, should be understood and practised going forward.

2 The Way in Which We Talk About the Human Matters

In this section we introduce the work of McGilchrist (2019, 2021) on cognition to show that the modern emphasis on analytic forms of thinking ignores the brain’s capacity for holistic and relational forms of cognition. In The Master and his Emissary, psychiatrist Iain McGilchrist (2019) argues that the two hemispheres of our brains give rise to two qualities of attention. This distinction is a consequence of the fact that we are embodied beings who, through evolutionary history, have had to develop two different kinds of attention. The left hemisphere is analytical—breaking things down into patterns and parts. If we consider a bird looking for seeds on the ground, it is the left hemisphere that enables the bird to distinguish the seeds from the pebbles. The right hemisphere is integrative and sees relationships in the whole. It is the right hemisphere that allows the same bird to keep an eye on its environment so that it can spot a predator. McGilchrist argues that, in the West, we are dominated by left hemisphere ways of viewing the world. That which can be measured (atomised) is viewed as more true and more real than the picture of the whole, which is quietly ignored and increasingly taken as irrelevant. The emergence of the two forms of attention is a direct result of the conscious being having a physical body. The highly focused analytical left hemisphere would not survive long without the holistic relational attention of the right hemisphere. But the atomised thinking of the left hemisphere is not wired to recognise the value of the holistic contribution of the right hemisphere.

The reduction of human cognition to a specialised information processing machine, valourising left hemisphere attention and neglecting right hemisphere insight, is highly problematic. Yet this worldview is entirely compatible with the human capital narrative of education referred to in the Introduction above. If higher education is valued, especially by governments and businesses, only for delivering on instrumental, extrinsic purposes such as economic development, then AI tools do indeed present an existential threat to the system. We can expect that in the next decade or so, the combination of processing power and the refinement of these toolsFootnote 3 to apply to particular knowledge fields will result in AI tools surpassing the analytic capacity of human beings. Surden (2019) notes ‘This is a fascinating fact—that machines can use detected patterns to make useful decisions about certain complex things without understanding their underlying meaning or significance in the way a human might’ (p.1315). But as McGilchrist (2021) argues, in the West, we have overemphasised left hemisphere cognition – atomisation and pattern recognition—which suggests that we cannot make Surden’s assumption that the human completing a similar task will in fact make more significant connections by seeing the whole. This is consistent with the goals of epistemic insight (Billingsley & Ramos Arias, 2017). Our current higher education system may be undermining this capacity precisely because the left hemisphere view of world that serves the needs of global capital is valued more highly than the holistic view. When we consider contemporary complex problems such as climate change, it is clear that if higher education is to contribute to their resolution, a major critique and reorientation of higher education is long overdue.

Whilst McGilchrist builds his argument from the medical literature, neuroscience and philosophy, Stein (2019) draws on a different intellectual stream. Stein has a background in education and draws primarily on Wilber’s (Walsh & Wilber, 2010; Wilber, 1998) Integral Theory. He offers a strong juxtaposition of ‘mind-as-computer’ and ‘mind-as-ecosystem’.

Whilst Stein’s juxtaposition of these two metaphors (Table 1) gives a strong visceral ‘sense’ of the possibilities of another vision for higher education, there are two underpinning ideas which are worth pointing out. Firstly, in the eco-system metaphor, higher education is relational. There are relationships between the student, the teacher and peers as well as with their lifeworlds and contexts which impact how they interact both socially and intellectually in the educational community and environment. Secondly, higher education is not simply about accessing and acquiring knowledge. The attainment of knowledge should have a transformational effect on the student. This aligns with Ashwin’s (2020) argument that the intrinsic purpose of higher education is the transformation of the student through an engagement with disciplinary knowledge. Stein’s (2019) vision leads us back to McGilchrist (2019) and the later development of his position into a full scale philosophy in The Matter with Things (McGilchrist, 2021). McGilchrist (2021) argues that any truth claim must draw on four elements—science, reason, intuition, and imagination. Science and scientific methodology work by isolating parts of systems in order to understand them (Bhaskar, 1978; Stengers, 1997) and thus can be considered primarily a function of the left hemisphere. Reason functions in both hemispheres. Intuition and imagination require the use of the right hemisphere. Intuition and imagination are a product of the relational. Imagination here is much more broadly understood than our fictional constructions. For example, imagination is the function we use to evoke empathy. We can sit with someone in grief and have some kind of feeling for their pain because we ourselves know what it is to have lost someone and we can extrapolate from our own experience to theirs. Surprisingly, McGilchrist (2021) argues that in any intellectual endeavour we need to hold all four dimensions in balance. This position is unusual in higher education. The dominant educational paradigm strongly aligns with an emphasis on science and reason but largely overlooks the value of intuition and imagination. However, it is the holding together of all four aspects which is the ‘superpower’ of the human. The right hemisphere’s capacity to see how the parts relate to a larger whole is a necessary correction to reductionist machine models of human cognition and to the work we get machines to do for us.

Table 1 The implications of different guiding metaphors for education (Stein, 2019)
Full size table

McGilchrist’s (2021) position is supported by Prigogine and Stengers (1985) and Stengers (1997) who have challenged the classical view of science. The key argument in Order out of chaos: Man’s new dialogue with nature (1985) is that the static view of classical science no longer serves us well. This is because science now shows us that the material world is open, complex, unstable, dynamic and irreversible in time. Prigogine and Stengers (1985) describe how the shift in science away from the concrete to the abstract is not a function of trying to grasp the nature of the material world, but rather a result of the limitations of classical science. As scientists tried to ‘purify’ their practices, science was compelled to isolate itself from its social and environmental contexts in order to achieve autonomy and a universal conceptualisation of knowledge. In practice, modern scientists have learned to prepare and isolate aspects of material reality so that their findings conform to ideal conceptual schemes. Stengers (1997) argues that scientific rationality and discourse conceals the contingent singularity involved in the invention of an experiment by a practising scientist (implying the use of imagination and intuition). In order to communicate an experimental event to other scientists, the scientist must construct their results as simplified abstract representations of reality that leave out much of the contextual and personal details involved in setting up and conducting an experiment. For example, when reporting their results in the universal discourse of modern science, the scientist will typically use the passive voice that hides the subjects of the verbs. Furthermore, the verbs themselves are often abstracted into complex nominalisations—a grammatical form that removes both subject and process. One effect of these discursive devices is to foreground things as facts whereby the subject-object relation and the context of the procedure are obscured. Stengers calls on scientists to take ethical and political responsibility for the selective and reductionist nature of their discourse.

Returning to McGilchrist’s (2019, 2021) position—it is the fact that our brains are embodied—which gives us the dual capacity to engage in such abstract scientific reasoning and also to draw on intuition and imagination (for example in designing experiments). To fail to attend to the embodiment of the brain—that is refuse to recognise that it is an embodied human being who must enact knowledge in the world—is to operate with a reductive understanding of the human and of how science is practised. To develop this point further, we turn to a critique of ‘powerful knowledge’.

3 A Critique of ‘Powerful Knowledge’

In the 1990s and 2000s in Higher Education Studies, there has been a lively debate around the notion of ‘powerful knowledge’. The concept was initiated by sociologists of education, (Muller, 2000; Young, 2007), to counter what they saw as the deleterious consequences of constructivist views of knowledge for education. Below we develop a partial critique of their conceptualisation of powerful knowledge based McGilchrist’s (2019, 2021) theory of human cognition and Stenger’s (1985, 1997) studies of the practice of science. The point of this critique is to show that even in one of the most influential ideas of the last few decades in higher education studies, the recognition of the significance of the embodiment of the brain for people learning, enacting and generating knowledge, is absent. Below we outline the key tenets of the argument for access to powerful knowledge through education as advocated by Muller and Young. We then engage in a partial critique of this position based on the theories of cognition and science discussed above.

Muller and Young wish to counter interpretivist, constructivist or conventionalist (in science) views of knowledge, which they believe lead to epistemological relativism (that all forms of knowledge are equal and should be celebrated). They argue that a consequence of this view of knowledge for education is the privileging of student-centred, context-dependent learning. Instead, Muller and Young advocate the differentiation of forms of knowledge, in which, because of its explanatory power, knowledge based on a realist epistemology is valued most highly for curriculum selection and teaching purposes (Muller & Young, 2019; Young & Muller, 2013). They believe that this heritage of specialised, ‘powerful knowledge’ is preserved in the modern disciplines—especially in the natural and applied sciences—which, unlike the humanities, have the property of building knowledge cumulatively on the basis of the empirical confirmation of theory. Furthermore, they argue that if (higher) education is to realise epistemic and social justice, it must do so by democratising access to this ‘powerful knowledge’.

Following Bernstein (2000), who built on Durkheim’s distinction between sacred and profane knowledge, Young and Muller describe both the internal and external properties of knowledge that make it specialised (or sacred, that is powerful). Internal properties include the coherence, systemisation and abstraction of concepts and the relations between them, as well as the validation of theory by autonomous evidence procured through rigorous methods. Unlike the humanities, it is these properties that enable scientific laws to hold for a wide range of phenomena and thus to build knowledge cumulatively through a process of subsumption. Young and Muller also describe the external properties of powerful knowledge which include the epistemic criteria and institutional and social norms used by scientific communities to generate, judge, validate and revise knowledge. They argue that it is these properties that enable powerful knowledge to achieve a high degree of ‘social objectivity’ that can transcend the contexts of its production as well as the interests of its producers, thus potentially enduring across space and time (Young & Muller, 2013, p. 237).

Regarding the internal properties of scientific knowledge, as noted above, Stengers (1997) is critical of the move from the concrete to the abstract that allows the presumed objectivity of scientific discourse. Contra Young and Muller, she argues that this move by scientific rationality imposes a reductive model on natural reality in order to achieve ‘universality’ through high levels of abstraction. Based on her studies of actual scientific practice, Stengers (1997) would challenge Muller and Young’s typification of scientific knowledge, because they focus on knowledge as reified object, abstracted from its processes and contexts which she wishes to re-insert. In fact, Stengers argues that science has attempted to ‘purify’ itself by stripping away its social and natural contexts of practice in order to conform to an ideal conceptual schema that generates models and facts as artefacts (Stengers, 1997 cited in (Olkowski, 2010) p.115).

Regarding the external properties of scientific knowledge, Stengers (1997) recognises the importance of this aspirational ideal for scientific communities committed to what has been termed the ‘epistemic interest’. However, Stengers (1997) also challenges scientists to be more open and honest about the political and social ‘contamination’ of their practices and the implications thereof. Given her understanding of experiments as singular, contingent events, she suggests that the role of scientific communities should be to encourage and support scientists to be imaginative and take risks in experimental practice. Furthermore, she challenges scientific communities to give up on the attempt to conceal their relationship to society and to power and to rather be open about the limits of experimental practices (Stengers, 2000 cited in (Olkowski, 2010) p.124).

Stengers’ argument has implications for powerful knowledge. In addition to this, from a different angle, Young and Muller have got into trouble with their definition of powerful knowledge. They are frequently misunderstood by critical social theorists who point out that regardless of attempts to democratise access, all production and exercise of knowledge entails the production and exercise of power, despite attempts by expert communities to restrain the latter (Zipin et al., 2015). Social reproduction theorists also point out that, as sites of power and privilege, there is a tendency for (higher) education institutions to reproduce rather transform the power relations of society (Apple, 1976; Bowles & Gintis, 1976).

In their 2019 paper, Muller and Young recognise the problem, admitting that ‘no knowledge, including specialised knowledge …. can remain innocent of power relations’ (2019, p. 208). In this paper they try to rescue their position by using Spinoza’s distinction between potentia (power as transformative) and potentas (power as domination), to make a clean distinction between ‘powerful knowledge’ and ‘knowledge of the powerful’. Despite their efforts, and whilst this analytical distinction might be useful to think with, it is clear that in practice, as illustrated by Stengers (1997), this distinction does not hold. This is because society—its institutions and power relations—pre-exists as a necessary condition for agents (for example scientists) to act within specific roles—making it impossible to keep potentia uncontaminated by potentas.

However, importantly, Stengers does support the realist view of knowledge that Muller and Young want to protect; on the grounds that the data (selected from nature) can and does talk back to scientists to either confirm or falsify their hypotheses. In this regard, Bhaskar’s (1978) critical realist metatheory is pertinent and supports the case that Muller and Young want to make for a realist view of knowledge. But crucially, and contra Muller and Young, Bhaskar does not base his argument on epistemology—which, with the epistemic relativists, he agrees is always fallible and partial. Instead, in line with Stengers’ reasoning, Bhaskar focuses on ontology, locating causal mechanisms in what is ontologically ‘real’. He posits that the (explanatory) power of knowledge is based on its ability to access and grasp the nature, mechanisms and stratified structure of the world (depth ontology). This includes the stratified nature of human being in the material, embodied, social and ideational worlds (Bhaskar’s four planar social being). Bhaskar’s position is thus congruent with McGilchrist’s (2021) who agrees that there is a reality (ontology) independent of our cognition (and epistemology) and that our knowledge of it is always mediated by our minds and bodies. That is to say that the ways in which we have learnt to experience, observe and understand the world shapes the ways we interact with and in it. We ourselves are part of the cosmos and so we can only participate in it in ways that are constrained by the causal mechanisms of the physical, social and cultural worlds.

It is important to note that all four sources that we draw on for our argument, McGilchrist, Stengers, Bhaskar and Muller and Young, would agree that the world exhibits an autonomous ontological reality—especially the natural world, which as noted above, autonomously validates or falsifies the hypotheses of scientists. What is crucial here is that the scientist cannot force nature to respond the way the scientist wishes: there is, in effect, a dialogue between humans and nature, not a dictatorship (Olkowski, 2010, p.116). It is this autonomous feedback from the ontologically real world that permits scientists to exercise epistemic judgement and criteria to judge some forms of knowledge as better than others. This allows for epistemic pluralism but is not epistemically relativistic. Nonetheless, the particularity of the knower’s (the scientist’s) experience will influence the choice of criteria used in this judgement.

4 The ‘Knower’ Who Produces the Knowledge Matters

We have established that the artificial separation of knowledge from its knowers is untenable. Here we discuss the problematic implications of doing so—first for knowledge creation and secondly for knowledge acquisition. Firstly, regarding knowledge creation; there is always a particular embodied person situated in culture and in time, educated in a particular way, performing an experiment. Their historical and socio-cultural context including their educational background will all influence the way in which the experiment is carried out and the way in which the data gets interpreted. Furthermore, the personal metaphors that the individual scientist uses to imagine their work is deeply impacted by their own web of interests and life experiences (Blackie, 2021, 2022). We must retain the notion that it is impossible to separate ourselves from the whole that we are exploring (McGilchrist, 2021). Stengers and Prigogine likewise argue that ‘all scientific activity is time-orientated and the scientist must come to see herself as part of the universe she describes’ (Stengers and Prigoine (1985, p. 301) cited in Olkowski, 2010, p. 117). Knowledge creation is therefore always particular and that which is accepted as ‘established knowledge’ is always situated in a particular time and space. Secondly, the separation of knowledge from the knower is problematic with regard to knowledge acquisition. Information stored in libraries—textbooks, monographs, edited volumes, journals and so on—may provide clear robust descriptions of aspects of the world. But until that information has been processed and integrated into the embodied mind of a singular person, it has little value. Taken to its hypothetical extreme—what is the value of a library and all it contains once humans cease to exist? We argue therefore that the ‘power’ in ‘powerful knowledge’ comes primarily through the enactment of that knowledge in the world by the person who has acquired it. The ‘enactment’ of knowledge can be transformative for both the knower and the world, for example in the interplay of different ideas; intervening in the material world; building physical structures in the real world; and the development of new conceptual and social structures. The point is that it is only once the knowledge is understood and embodied that its power is realised and this always entails more than knowledge in any ‘pure’ or abstract sense. This is crucial for epistemic insight. The fact that Newton’s Laws can be written in a way that appears to have no cultural reference does not negate the fact that Newton lived in 19th Century England. Although Newton’s laws remain valid within certain parameters two centuries after their formulation, a student learning Newton’s Laws in the 21st Century will gain from knowing this history and the predominate understanding of nature within which they were developed. This allows the student to see clearly their limitations in light of developments of quantum physics. Furthermore, the students will be learning and experimenting from within a particular location. Newton’s Laws will be put to a particular use framed by the student’s lifeworld. This has ethical as well as political implications. For example, whilst one student may attempt to create a safer wheelchair ramp to improve the lives of paraplegics, a second student may turn their attention to the development of a projectile delivery system to cause maximum disruption of a Martin Luther King Day parade.

The separation of knower from knowledge means there is no recognition of the ways in which the knower has been shaped by their lifeworld (Blackie, 2021). As a consequence of the modernist conceit to prioritise science and reason over imagination and intuition, knowledge is presumed to be socially neutral. Ethical questions are seen as irrelevant, but this position cannot be defended. Ethical behaviour requires a certain level of reflexivity (Eagleton, 2004), which in turn depends on right hemisphere functions—imagination and intuition working dialogically with detachment (a form of reason). Eagleton (2004) argues that we should understand fact and value to work dialogically and not separately—as has often been assumed when knowledge is separated from its knower and the context of its production. Values inform the ways in the knower chooses to employ knowledge.

We have argued that the move to deny the personal in science in pursuit of some objective, universal truth is problematic. This extends to the social and the normative although we have not elaborated on these aspects in as much detail here. Whether acknowledged or not, power and the ethical implications thereof lurk in the shadows, exerting forces on and through actors and institutions in the system. Rather than attempting to remove issues of power and ethics from the practice and use of science, we argue that scientists should rather be aware of their scientific, ethical and political responsibility and affirm the selective and partial character of their knowledge (Stengers, 2000 in Olkowski, 2010 p.124).

Thus we suggest that the emergence of AI tools is one of the greatest opportunities afforded higher education in a long time. The disembodiment of the human mind and the reductive metaphor of the ‘human-as-machine’ are unfortunate by-products of the European Enlightenment and the Industrial Revolution respectively. The call here is not for a luddite-type response—to ban the use of AI tools. Rather AI tools have brought into sharp focus the need to rethink what we are trying to achieve in higher education. We have advocated that our model of cognition should re-focus on the acknowledging and developing the imagination and intuition of the human knower. As increasingly AI tools take over the analytical functions and demands of society, this should free us up to complement the capacities of AI with the distinctively human capacities for holism, relationality and creativity. Furthermore, it will be increasingly important that we teach students how to think reflexively about the knowledge generated and therefore to use it in ethical and responsible ways.

5 Some Principles for a Meaningful Response to the Challenge of AI Tools

We think it is premature to attempt to list here a set of practical institutional strategies and teaching methods to both complement and deal with the affordances posed by AI to higher education, particularly to specific disciplines and professions. Instead we summarise below some of the premises and principles suggested by our discussion above, which we believe should inform future institutional and disciplinary practices in an era of generative AI:

  1. 1.

    We need to consciously make a shift in metaphor from ‘human as machine’ to ‘human as eco-system’ in administrative, evaluative and pedagogic practices. AI tools could then be located as one part of the complexity that that forms the eco-system. Pedagogically this means bringing the person—the embodied knower—into view. Blackie (2022) terms this ‘knower awareness’. There are many ways to do this, one is to make visible the knower through the use of personal metaphors. Often in teaching, the instructor will use an example from their lived experience to make a concept more tangible. Illustrating molecular rotation through the example of an ice skater spinning more quickly when she pulls her arms into her body is often used (Adendorff & Blackie, 2020). Students in warm Southern contexts may offer other more meaningful life experiences such as crafting spinning tops. Allowing discussion around the use of a range of metaphors is a powerful way to illustrate how we can draw our own life experiences and imaginations to enrich learning. Such an approach can also be used to gain knowledge about knowledge as required by epistemic insight if students are prompted to reflect on what they have learnt and on how the metaphors used by their peers afford different nuances of understanding.

  2. 2.

    We need to expand our understanding of cognition, to restore recognition of the value and powers of intuition and imagination. This move should not in any way prejudice the importance of reason and science. We should hold all four faculties in tension if we are to claim epistemic reliability and validity—what some, including McGilchrist (2021), would call ‘truth’. Furthermore, with Stengers (1997, 2000), we must recognise that both reason and science are bound and limited by time and context. Ideas do not usually drop out of nowhere. Students should understand that even the most brilliant of scientists are products of their embodiment in a particular environment (Blackie & Adendorff, 2022). For example, Einstein’s focus on time was likely influenced by his working as patent clerk over a period when there were good number of patents being filed to solve the problem of accurate local time precipitated by an increasing mobility of people by the rail network. Would he have been musing about time in the same way had he been in that office 10 years earlier or 10 years later? We should promote reflexivity and ethical development through modelling and teaching epistemic insight. Epistemic insight is essentially knowledge about knowledge (Billingsley & Ramos Arias, 2017). This includes recognising that what we know is limited by the current understanding of the community working in that field (Stengers, 1997, 2000). Secondly, it entails acknowledging that there is always a person using, enacting and creating knowledge within a community of practice (McGilchrist, 2021; Stengers, 1997, 2000). Furthermore, if we are educating scientists, we must presume that they will put that knowledge into effect somewhere. Stengers (1997, 2000) is unequivocal in her position that power is always operational in science. Therefore appropriate development of ethical considerations is not a ‘nice to have’ but should be an essential aspect of all science education. This broadened understanding of the elements of knowledge creation should inform epistemic insight as it is introduced at a secondary school level.

  3. 3.

    We should engage in cross disciplinary dialogue. In higher education each discipline tends to operate within its own silo. There may well be good reasons for retaining clarity of boundaries and structure in an undergraduate curriculum where students are moving from being consumers of codified information (secondary school) to producers of knowledge (postgraduate degrees). Nonetheless, we should find ways to encourage cross-disciplinary thinking through the kinds of problems we pose to undergraduate students. Engaging with colleagues and ways of thinking in a different discipline tends to make one more conscious and reflexive about assumptions made in one’s own discipline, thus developing knowledge about knowledge. Furthermore, increasingly postgraduate students are required to engage with a multiplicity of knowledges and research communities in order to address our most pressing and complex problems. In such environments one is required to take cognisance of the assumptions and simplifications of the world inherent in one’s home discipline. This requires some level of ‘knower awareness’. The promotion of epistemic insight at a secondary school level could substantially assist in preparing students for this cross disciplinary engagement.

6 Conclusion

Much of the conversation around higher education appears to be implicitly reliant on some version of ‘the human as machine’ model. In an age of generative AI—if that is our guiding metaphor, higher education has no future. AI tools are faster, more reliable and more efficient at functioning as machines than human beings. This reductionist view of knowledge and knowing will make the human being obsolete in an era of AI tools. Instead, we have advocated a broader view of education in general and higher education in particular. Herein, we have shown why the move to abstract, universal truth claims and the concurrent erasure of the particularities of the human investigator no longer holds, even in its own terms. This erasure leads to naïve empiricism or the epistemic fallacy (Bhaskar, 1978)—what we experience and know about the world is conflated with the world itself. We easily forget that our concepts are contingent—singularities situated in culture and time and therefore always partial and fallible. Therefore we have argued that the invitation for higher education at this time is to make visible the deep value of the imaginative, intuitive and occasionally brilliant connections that can be made by any one person or team, grappling to understand a singularity at a particular time. We must foster these sensibilities in the next generation. We are deeply curious about the multiplicity of strategies and approaches that will arise in higher education as we do so in the next decade. We are committed to communicating with those involved in holistic approaches to engagement with knowledge such as the community who foster epistemic insight.