Annotations

(1/12/2026, 10:33:49 AM)

Go to annotation “In our view, Plato’s ultimate concern in this passage is with responsibility: once written down, it becomes much harder to identify who is responsible to and for the meaning down, it becomes much harder to identify who is responsible to and for the meaning of a claim, and – equally important for Plato – who is responsible for the effects that of a claim, and – equally important for Plato – who is responsible for the effects that such a claim might have on those unable to properly contextualise and interpret its such a claim might have on those unable to properly contextualis” (Flenady and Sparrow, 2025, p. 3)

Go to annotation “(Bearman, Ryan, and Ajjawi 2022) has expressed fears that GenAI will, for example, deepen rather than relieve the ‘epidemic’ of academic overwork and precarity example, deepen rather than relieve the ‘epidemic’ of academic overwork and precarity (Watermeyer et al. 2024), and ultimately destroy the reading and writing skills on which (Watermeyer et al. 2024), and ultimately destroy the reading and writing skills on which the humanities depend (Freiberg the humanities depend (Freiberg 2024). One common response to scepticism” (Flenady and Sparrow, 2025, p. 3) I’m medium rare among these three camps as of now

Go to annotation “We cannot reasonably expect students to take responsibility for their so-called ‘collaborator’ and reasonably expect students to take responsibility for their so-called ‘collaborator’ and ‘tutor’ because GenAI is fundamentally unable to assume responsibility for itself or ‘tutor’ because GenAI is fundamentally unable to assume responsibility for itself or others. Drawing on recent accounts of GenAI outputs as irresponsible ‘bullshit,’ we others. Drawing on recent accounts of GenAI outputs as irresponsible ‘bullshit,’ we argue that commentators should in turn ‘cut the bullshit’ and not refer to GenAI in argue that commentators should in turn ‘cut the bullshit’ and not refer to GenAI in higher education as a collaborator and tutor – words that falsely imply that GenAI is a higher education as a collaborator and tutor – words that falsely imply that GenAI is a responsible responsible agent.” (Flenady and Sparrow, 2025, p. 3) Critical point here -AI Cannot take responsibility And therefore can never be a real collaborator

Go to annotation “Where most people content. Plato’s response was to restrict his ‘real’ doctrine to an oral conversation shared between the select few: philosophers present to and thus directly responsible for the meaning of their words, and those ‘initiates’ (citizens, always male) privileged with the time and opportunity to learn to take those words in the right sense (Hadot 2002). The modern academy is of course very different to Plato’s. Even though elite universities demonstrably privilege those from certain race and class backgrounds (Bhopal and Myers 2023), they are at least in principle egalitarian and meritocratic. Equally obviously, the modern academy depends on writing and written archives. Nonetheless, tertiary insti- tutions remain acutely sensitive to Plato’s second worry: universities work to redress the ‘defencelessness’ of words by ensuring that written claims are properly connected to those responsible for them via rigorous practices of attribution: we know who the authors are, we know what institutions they belong to and represent, and we know how to contact them. Whether or not we in fact get a response from a given author when we attempt to engage them in written argument, we know who is ultimately responsible for defending a work, and to whom our response to that work ought to be directed. At the same time, we work to reduce the ‘hobnobbing’ of written words by ensuring those who engage with academic texts have either been trained, or are in training, to possess disci- pline-specific skills that allow them to contextualise, interpret, and assess the text in ques- tion. Together these two features of the modern academy – named authorship and attribution combined with disciplinary-specific training in the assessment of claims – have worked to neutralise Plato’s fear words. However, that we work as conspicuously as we do to neutralise Plato’s worry suggests that his concerns remain relevant. In the age of the internet, of course, the impact of ‘lies’ ‘fake news’ and ‘bullshit’ on public opinion is a serious concern.2 Here again Plato’s second worry seems legitimate. In the case of internet sources, we do not know who or where the authors reside, who their words are intended for, or even – increasingly – if they are human. Thus, it has become even more important to inculcate university students with the capacity to dis- tinguish between those works that can be sourced to a responsible author located within a discipline and an institution and those dangerous words for which authors do not take responsibility. In our view, those who respond to critics of AI in education by pointing to Plato’s fears regarding writing unwittingly draw attention to the ethical heart of the matter. The intro- duction of GenAI teaching systems is irresponsible in the sense that it introduces into the academy words without an author – words for which no one is responsible (Bingham 2024).3 This, we suggest, has the perverse outcome of demanding students take respon- sibility for discourse that they are not yet in a position to properly assess. Moreover, to the extent that GenAI teaching systems substantially replace students’ interaction with real human teachers in the future – very much a possibility (Selwyn 2019) – it will become increasingly difficult for students to develop the skills to assess GenAI outputs, just as students become increasingly divorced from the human relationships that give their own academic responsibilities meaning. Irresponsible Bullshit Harry Frankfurt famously defined bullshit as ‘lack of concern with truth, or an indiffer- ence to how things really are’ (Frankfurt 2002, 340; Frankfurt ence to how things really are’ (Frankfurt 2002, 340; Frankfurt 200” (Flenady and Sparrow, 2025, p. 4)

Go to annotation “Moreover, to the extent that GenAI teaching systems substantially replace students’ interaction with the extent that GenAI teaching systems substantially replace students’ interaction with real human teachers in the future – very much a possibility (Selwyn 2019) – it will real human teachers in the future – very much a possibility (Selwyn 2019) – it will become increasingly difficult for students to develop the skills to assess GenAI become increasingly difficult for students to develop the skills to assess GenAI outputs, just as students become increasingly divorced from the human relationships outputs, just as students become increasingly divorced from the human relationships that give their own academic responsibilities that give their own academic responsibilities meaning.” (Flenady and Sparrow, 2025, p. 4)

Go to annotation “That is, GenAI is designed to produce a particular reaction in its users – to convince its users of the usefulness of the response – rather than to say users – to convince its users of the usefulness of the response – rather than to say some- thing true of the world. They are, then, fundamentally thing true of the world. They are, then, fundamentally bullshitters.4 Importantly, to say AI outputs are bullshit is not to claim that” (Flenady and Sparrow, 2025, p. 5)

Go to annotation “Importantly, to say AI outputs are bullshit is not to claim that they are always untrue, nor to say that they can play no role in the justification of belief. A bullshitter might in nor to say that they can play no role in the justification of belief. A bullshitter might in certain cases achieve their rhetorical goals via the utterance of true claims. Similarly, AI certain cases achieve their rhetorical goals via the utterance of true claims. Similarly, AI might – perhaps even for the most part – produce outputs that reliably correspond with might – perhaps even for the most part – produce outputs that reliably correspond with states of affairs in the states of affairs in the world. That the outputs of” (Flenady and Sparrow, 2025, p. 5)

Go to annotation “As such, ‘hallucination’ is a misleading term: it gives the impression that GenAI wants to report on the world, but suffers some term: it gives the impression that GenAI wants to report on the world, but suffers some breakdown in that reporting (Hicks, Humphries, and Slater breakdown in that reporting (Hicks, Humphries, and Slater 2024). Secondly, this way of assigning responsibility has the” (Flenady and Sparrow, 2025, p. 6)

Go to annotation “Secondly, this way of assigning responsibility has the perverse outcome of making individual students entirely responsible for the truth of the claims in the work they individual students entirely responsible for the truth of the claims in the work they submit, in place of genuine collaborator and teacher-student relationships that are submit, in place of genuine collaborator and teacher-student relationships that are characterised by mutual responsibility to the characterised by mutual responsibility to the truth. Calling GenAI a ‘collaborator’ disregards” (Flenady and Sparrow, 2025, p. 6)

Go to annotation “Calling GenAI a ‘collaborator’ disregards accepted understandings of that term: to col- laborate is to work together with another agent who is responsible for their laborate is to work together with another agent who is responsible for their contributions, who is capable of holding themselves and others to collective standards governing a” (Flenady and Sparrow, 2025, p. 6)

Go to annotation “The limitation of this approach, however, appears once we consider the nature of pedagogical of this approach, however, appears once we consider the nature of pedagogical respon- sibility as holding mutually between teachers and sibility as holding mutually between teachers and students. As students, one important way we come to a sense of” (Flenady and Sparrow, 2025, p. 7) Reminds me of Gulya’s idea that relational teaching Is the future

Go to annotation “More concretely, if students come to feel they are writing with and for a machine rather than with and for human beings – if their education with and for a machine rather than with and for human beings – if their education emphasises prompting and revising GenAI outputs, with submissions in turn to be emphasises prompting and revising GenAI outputs, with submissions in turn to be marked by GenAI – it is plausible to suggest that they will not feel responsible to marked by GenAI – it is plausible to suggest that they will not feel responsible to repre- sentatives of disciplinary standards, and thus not to disciplinary standards at all. sentatives of disciplinary standards, and thus not to disciplinary standards at all. Asking students to critically interrogate GenAI outputs in line with disciplinary standards thus students to critically interrogate GenAI outputs in line with disciplinary standards thus faces a significant conceptual problem: we may come to ask students to exercise faces a significant conceptual problem: we may come to ask students to exercise respon- sibility to a standard outside of the human contexts in which those standards have sibility to a standard outside of the human contexts in which those standards have mean- ingful ingful grip. This” (Flenady and Sparrow, 2025, p. 7)

Go to annotation “However, the evidence available thus far is ambivalent, with GenAI in some cases promoting disengagement (Lo, Hew, and Jong ambivalent, with GenAI in some cases promoting disengagement (Lo, Hew, and Jong 2024). Indeed, there is already evidence that students are more likely to rely uncritically 2024). Indeed, there is already evidence that students are more likely to rely uncritically on ChatGPT – effectively using it to cheat – in cases where they feel ‘degree apathy,’ that on ChatGPT – effectively using it to cheat – in cases where they feel ‘degree apathy,’ that is, where they do not feel invested or engaged in their education (Playfoot, Quigley, and is, where they do not feel invested or engaged in their education (Playfoot, Quigley, and Thomas 2024). It is plausible then, on current evidence, to predict a vicious circle: Thomas 2024). It is plausible then, on current evidence, to predict a vicious circle: sub- stantial” (Flenady and Sparrow, 2025, p. 8) The argument that students need to evaluate AI outputs critically falls apart When we realize the skills needed to do that can only be taught by humans. Need to bring this up with Katy.