As with any work invested in thinking about place, we are conscious that place has formed the work at hand. This paper has been developed across different places over the course of three years, on Noongar and Yamatji Country, as well as being developed at a conference on Gadigal Country.
Caitlin (8th November 2023)
I write this introduction on Yamatji country, Nabawa, a five-hour drive north from Boorloo/Perth. Nabawa is on the border between the semi-arid and Mediterranean climate zones; when we arrived on Friday it was 35°c, we had to turn the AC on, and the drip of the condensation punctuates these sentences as even as a metronome. It's harvest season. The grain trucks rumble past, there is a haze that makes the sky -- the sunset -- a brilliant orange.
In Nabawa, I'm emailing Catherine about the paper we are delivering together. The sound of emails being sent and arriving is the same ping here as in anyplace, anyspace. We are sharing this paper, the same way we share the coastline running along/through the Noongar Countries in which we live, the suburbs of the universities in which we teach creative writing, and the way we share a love of Randolph Stow, the poet and novelist I'm in Nabawa because of. I'm here because Nabawa is in the Chapman Valley, southeast of where Stow turned Northampton stations into fictional locations and northeast of where he was evacuated as a six-year-old in World War Two to live with family on the sandy plains of the Greenough River. I'm here because the work I do as a poet requires physical engagement with place, or at least I say it does on grant applications and acquittals, on the increasingly numerous statements of research significance my non-traditional research outputs require as justification in the academy (statements which very well could be AI-generated for how little they capture the actual experiential dimension of a poem bubbling out from close-grained observation of a singular location across a swathe of time).
In his work on Stow's Geraldton novels, Tony Hughes-d'Aeth aims for such a 'close-grained treatment of the material dynamics of colonisation' (1). These dynamics, he argues 'in particular, the mode of agrarian land usage ... tend to be mystified by the metaphysics of alienation of a certain school of Australian criticism, and occluded in the more typological forms of postcolonial critique' (1). I am in Nabawa because I too am drawn to the material rather than the metaphysical. One way I think about this, is that I like to ask/answer questions with my body as well as my mind, or rather I do this with both, an irreducible multiplicity. The question I am emailing Catherine about, sitting here in Yamatji Country, listening to the trucks cannon along roads built long after Stow left, is what are the pedagogical intersections, complications, contradictions of writing in/of/from/with place in the spaces of Generative AI?
Catherine (7th April 2025)
The question Caitlin offered by email in 2023 is still humming somewhere in the digital archive of my email. Still on Whadjuk Noongar boodjar in Boorloo/Perth, not too far from my office on campus where it first appeared for me, I can summon it to appear again on my screen, and contemplate it anew. Some aspects of the conversation have developed, others continue to circle around the same conclusions. GenAI has both continued and changed in the same way -- it constantly makes itself anew. Its presence on Whadjuk Noongar boodjar, where I am now, where I was in 2023, continues to add complexity to my teaching on and of this place.
My responses to the advent of this technology have been complicated by time, and by the things that have happened in the time since: the evolution of GenAI, the advent of university policy approaching it, the evolution of my personal world, even the changing bodily space from which I write. The arrival of a new child, who sits with me as I type this. The ping of email from Caitlin here, in my spare room, is the same as it was in my office, when she first emailed me from Yamatji Country, the same as anyplace, anyspace. But in my spare room, surrounded by baby paraphernalia and with a sleeping child in a bouncer at my feet, it sounds somehow different. Simultaneously, it connects me to a version of myself that exists in spaces other than this. I mute it, to avoid waking the baby. But her question is still open, and we have returned to it several times over now: how does GenAI alter and impact teaching creative writing, and particularly the writing in/of/from/with place? It is not a question with a definitive answer, we have realised. Rather, it is one that stays with us, one that we have learned to carry day-to-day, between the various spaces we inhabit in our academic lives.
We begin this paper with a full-body question, because what GenAI is designed to do is to narrow, to answer, where we wish in this dialectic to expand, explore. We also begin with an acknowledgement of Country because one immediate complication of GenAI (particularly compounded by online classroom spaces) is where do we consider the act of writing to be occurring when it is being produced by GenAI? (We note the instinct to use the word produced rather than written in the previous sentence -- a concept we will return to.) Prepositional thinking is implicit in this questioning, as a function of relationality and thus positionality, a counter to the anyplace (/nonplace) and inherent lack of embodiment symptomatic of GenAI. Acknowledging Country, acknowledging our positionality and settler colonial heritage on Country, represents likewise an acknowledgement of the reality that Jason Lewis points to in the opening of his introduction to the position paper Indigenous Protocol and Artificial Intelligence: 'scholarly traditions that homogenize diverse Indigenous cultural practices have resulted in ontological and epistemological violence, and a flattening of the rich texture and variability of Indigenous thought.' (4) From the outset, it is important to recognise that the body- and placelessness of GenAI constantly risks such homogenisation in flattening diverse cultural knowledge and ontologies into data. There is, Lewis also notes, a long history of technological advances being used against and to the detriment of Indigenous peoples (6). And as Hèmi Whaanga makes clear: 'With homogenization comes loss.' (35)
Bringing this context to the creative classroom is a first gesture to the nature of the dialectic that we wish to employ here. Forces of homogenisation can be met with and challenged by unbounded thinking, and thinking about place and situatedness across different scales is useful grist. This might be one reason why the very first protocol listed in the Indigenous Protocol and Artificial Intelligence position paper is 'Locality': 'Indigenous knowledge is often rooted in specific territories. It is also useful in considering issues of global importance.' (Lewis et al. 21) Throughout this paper, we will be asking what future might a place-based approach to creative writing pedagogy offer in a time of GenAI? How might such approaches speak back to the flattening and generalising imperative of (current and continuously mutating) GenAI programs? This is driven by both theoretical interest in the possibilities of GenAI -- what texts and places might our students, might we produce now -- and a practical one, as Catherine put it at one early meeting: 'there is a suspicion all I am doing is managing [Gen]AI -- checking for use in assignments -- rather than productively questioning this reality in teaching' (pers. comm.).
We will be expounding on three primary frames exploring the intersection of GenAI and place in the creative writing classrooms. We will begin by asking, in what way does the particularity of writing place put pressure on GenAI? Beyond, the next two frames of approach reflect our pedagogical motivation to search for ways to foster cultural safety in and out of our classrooms: in the second section, we will explore how the nexus of GenAI and place in the Creative Writing classroom requires more than simple 'management' to support our diverse student bodies; and in the third, we will consider value structures at this nexus-point -- both in the context of degrees and as a concern across the literary sector. The fourth frame to our thinking, implicit throughout the paper and expanded on across the work as a whole, is an attempt to respond to the geopolitical dimensions of GenAI as a phenomenon, particularly in reference to Indigenous knowledges and (place-) writing on Country.
Place and GenAI
In her 'On Freedom', in the section on climate change after acknowledging that she has no 'special love of "the local"', Maggie Nelson quotes Bruno Latour on the false binary between local and global: 'The planet is much too narrow and limited for the globe of globalization; at the same time, it is too big, infinitely too large, too active, too complex, to remain within the narrow and limited borders of any locality' (188). The negotiations required to exist between and within local and global are fundamental to approaching life -- and therefore writing -- in a world of climate catastrophe. What GenAI does in a creative writing (or any) classroom, is make us constantly aware of how the pedagogical spaces we occupy are not fixed to any one place.
But this is of course, not something which has come about purely with the advent of GenAI. Returning to Hughes d'Aeth's article of Stow, he frames Stow's local novels 'firmly within a global extension of agricultural production into new zones of exploitation during the late nineteenth and twentieth centuries' (1). Stow's at the time thinly-veiled portrayal of his own settler familial ancestry is both very local and particular (for instance the individual stations are easily recognisable in their real-life counterparts) but also inextricably intertwined with global material forces. This is one way we teach students to read creative writing texts and to craft their own, as belonging to a particular person in a particular place but shaped by a series of forces beyond that, forces which are not always easily, or desirably, knowable. The difference Gayatri Chakravorty Spivak positions, way back in 2015 well before GenAI, between the global of 'Planetary' and 'planetarity', where:
The globe is on our computers. No one lives there. The 'global' notion allows us to think that we can aim to control globality. The planet is in the species of alterity, belonging to another >system; and yet we inhabit it, on loan. It is not really amenable to a neat contrast with the globe. I cannot say 'the planet, on the other hand.' When I invoke the planet, I think of the effort >required to figure the (im)possibility of this underived intuition. (291)
Caitlin (10th November 2023)
I am writing this while opposite a large grain silo, a truck marked oversize, a drought diminished flock of corellas on the wires, but I am also writing this while my son streams 'My Friends Tigger and Pooh' from Disney+ and I scan from the window to any of the nineteen (yes, I counted) tabs on my desktop. I do not know what question you could ask GenAI to get it to bring together all of these diverse sources of input, the question is simply my being here and noticing what is around me.
Writing on 'AI and the Creative Process', James Hutson turns to Margaret Bodewood's 1978 'Tripartite Theory of Creativity' of 'exploration, transformation, and imagination', to argue that while GenAI can work well with exploration and transformation as AI 'has computational strength to follow specific algorithms to create something and can also use random exploration to produce unexpected results', it cannot replicate 'intuition, insight and personal expression of the human condition'. (GenAI can help you write of place, but it cannot write from place, and it especially cannot negotiate the postmodern tension of occupying multiple understandings of place and space simultaneously, let alone hold space for the unknown.
This is another way of reframing what a common complaint about GenAI-generated writing is: the generic. The writing produced by programs such as ChatGPT is characterised by a flattened tone, neutrality, a sense it could have been written by anyone in anyplace. Even the most recent successes (as of April 2025), such as OpenAI producing a short story which met with acclaim from writers including Jeanette Winterson, who called it 'beautiful and moving', are without a rootedness in place. Instead, what is celebrated about OpenAI's short story is its metafictional conceit of being written by AI. As part of this the AI narrator openly muses on how setting works in fiction, writing,
This is the part where, if I were a proper storyteller, I would set a scene. Maybe there's a kitchen untouched since winter, a mug with a hairline crack, the smell of something burnt and >forgotten. I don't have a kitchen, or a sense of smell. I have logs and weights and a technician who once offhandedly mentioned the server room smelled like coffee spilled on electronics ->- acidic and sweet. ('A Machine-Shaped Hand')
This story succeeds because it builds character from exactly the flattened tone and neutrality which has thus far been key to creative writing produced by GenAI.
A core tenet of writing place is particularity, specificity key to understanding ecologies and the diversity of lives supported within. This is not a singular specificity, but an entangled one, reflective of the way we are all tangled up together. The question then becomes how, in the creative writing classroom, might thinking about place and GenAI help us recast our understanding of what the specific means in writing contemporary places? Thinking of place in terms of the geopolitical contexts of GenAI can offer ways to diversify critique around GenAI for our students. Firstly, there is the consideration that when it comes to the creation and subsequent operations of GenAI, place still plays a role in shaping who may use what GenAI and for what purposes. For instance, at the university where one of us works, certain uses of ChatGPT are permitted where DeepSeek is not to be used as it is 'considered an unacceptable risk on Australian Government systems' (pers. comm., 3 March 2025). This follows the Australian Government Protective Security Policy Framework (PSPF) 'Direction 001-2025', which directs entities to 'manage the risks arising from DeepSeek's extensive collection of data and exposure of that data to extrajudicial directions from a foreign government that conflict with Australian law' (1). So, while the information and writing provided by GenAI is hypothetically or seemingly neutral, place (where the searcher is vs. where the servers are) dictates use conditions. The seeming neutrality also obscures biases in outputs and inputs, as 'both the data that is used to train AI systems... and the codes written by human engineers are sources of bias, as engineers have their inherent bias... that is likely to pass onto algorithms if left unchecked' (Ho et al. 1). The ethical implication of this as regards GenAI as classroom collaborator will be discussed in more detail below, yet it is worth noting how place shapes these biases as the coders themselves are not outside the influences of their own geopolitical locations.
It is necessary to locate these nationalistic politics within an overarching ethos of planetarity, which is the holistic framework into which the individual points in this article nestle. (Wherever there is place, there is also the planet.) At tension with a nationalistic ethical concern, for instance, is the broader concern around the environmental costs of GenAI. While the search processes of ChatGPT have been likened to 'powering up a nuclear reactor to use a calculator' in terms of their inefficiencies of resource uses, DeepSeek has been touted for searches which use energy scaled 'up-and-down depending on the complexity of the prompts' (Howson). Because of the emergent nature of the technology, there is a paucity of information on the isolated costs of the use of GenAI in educational contexts. What exists, however, indicates a dramatically increased environmental cost for classroom and university practices which incorporate GenAI, despite universities in Australia being rated against the Sustainable Development Goals (SDGS) which prioritise environmental responsibility. Assessing the energy costs of GenAI in a design class, Lupetti et al. found incorporating GenAI 'can easily double the energy costs associated with students' use of computers'. If we are invested in place-based pedagogy as a way of shaping ethical and responsive classroom practices, then GenAI offers ways to think broadly about the local-global nexus, as well as the environment.
Moving from the philosophical to the practical, however, one of the most common ways GenAI is framed as enabling creativity is in opening up resources to the creator, whether these be knowledge, processing capacity or production. GenAI in this lens is often framed as collaborator. Studies have shown that it can be favourably used in the initial stages of developing a story idea, but that stories developed with large reliance on GenAI show less divergence -- or particularity -- from one another (Doshi and Hauser 1). In our teaching across universities and units, we both use a mapping exercise that illustrates these different potentialities for GenAI to be used to deepen engagement with place. Following Vanessa Berry's Mirror Sydney, Caitlin asks students to think about a place they know well, preferably one they attach personal memories too. She then asks them to sketch it, then in any way which works for them to write their memories on the map. Then students are asked to search for knowledge beyond their own, to find any resource or resources on that place, then layer this impersonal -- or generic -- knowledge alongside their own. GenAI in this instance could perform multiple functions following student directions, it could help sketch, or it could be directed to finding information on that place. Here, in this exercise, the generic of GenAI emerges as a potential pedagogical resource to think through how places have competing meanings, as well as a problematisation of the meta and paratextual, how authorship is increasingly shared, a space to think about what positionalities they are occupying across the shifting scales of place/space.
Classroom as Place
Catherine (22nd November 2023)
In my version of the creative mapping exercise, I draw from paired notions of the 'deep map' and the 'thick map' -- versions of mapping which seek to encompass and embed a breadth of subjective response to place alongside the layers of context that Caitlin describes. The assignment version of this was originally designed as assessment for the unit 'Writing/the Environment', coordinated in its first iteration by Tony Hughes-d'Aeth and myself. It is one I have used personally now in multiple units as a classroom exercise, and it asks students to return repeatedly to their chosen place, to observe it at different times of day and seasons. The exercise is intended to encourage reflection not only on subjectivity within place, but situatedness as well -- to ask students to consider how they are embedded and complicit in the complex system of that micro-environment.
As Robyn Longhurst suggests, 'Bodies exist in places; at the same time, they are places.' (337) This is something our Western Australian bodies are conscious of each time we travel east and finds ourselves caught hours out-of-context in a new time zone. Contemplating body in this way is suggestive of interoception, the ability of the body to feel within itself, to be perceptive of what is happening within itself, (most commonly riffed on in popular contexts as 'mindfulness'). Judith Bishop discussed Antonio Damasio's work on interoception in an Australian University Heads of English (AUHE) forum on AI (22 November 2023) as definitive of human subjectivity, and the way human writers are perceptive in a manner distinct from and as yet unobtained by GenAI. Drawing a line from interoception to perception opens the possibility of likewise connecting between lived experience of place, physical manifestations of situatedness, and the conceptual theorising of positionality.
Situatedness, in the creative writing classroom, is a productive means to cognise positionality, core tenet to ethics within place-writing, and likewise to pedagogical ethics. The self-reflexive recognition of positionality from a teacher as well as students lends itself as a leveller to the kind of decentred power which sustains an inclusive classroom. Contemplation of situatedness and power in the classroom is taken up in research focused on digital creative tools as well, as demonstrated in the work of Jack Tsao and Collier Nogues. While an interest in positionality would seem antithetical to pedagogical approaches centring around digital engagement, much of this research turns on the manner in which digital apparatus and tools can alter or shift specificities of the classroom as a place. Tsao and Nogues, analysing their co-curricular program of teaching creative writing in Hong Kong, highlight the use of GenAI as a potentially 'liberating' site of collaboration, an 'AI-empowered "learner-as-leader" model, emphasising and furthering research around increased student agency and redistribution of power in the classroom.' (3) In contrast to our creative mapping exercises, which push beyond digital engagement to physical being-in-place, Tsao and Nogues invoke GenAI and the digital space as a tool which can alter students' experience of the classroom as a place. The imperative to the experimentation with GenAI in their project is in the pursuit of intellectual equality and emancipation, taking up the definitions of these terms offered by French philosopher Jacques Rancière. Their project responds to an anxiety that traditional teaching practices limit possibilities for creativity in the writing classroom in 'presupposing a dependency on learned teachers' (2) -- a mode of situatedness which diminishes a consciousness of positionality, maintains a centred power and reduces inclusivity.
There is a distinction to be made in this conversation therefore between the pedagogy of place-writing generally, and the classroom as a place: not always physical, but always with an implicit onus to managing student experience and safety. Already, in creative writing, this requires care. Tsao and Nogues highlight the possibilities GenAI affords when adopted in the classroom via 'machine-in-the-loop' processes of writing, seeing it as promoting student agency in learning (2--3). There are positive implications, they argue, in the shift it can occasion in classroom power dynamics, given that all 'writers can engage with GenAI on terms of equality, alleviating concerns about teachers' potential and contempt.' [1] (3) Their program saw students experiment with AI tools to create/generate -- the words are used interchangeably at different points (4) -- graphic short stories or poems, producing drafts, final submissions and reflective writings as well as engaging in peer feedback and surveys after the fact. In discussing their pedagogical methodology, Tsao and Nogues lean on the transformative capacities of GenAI, coupled with its inherent person- and placelessness, as implying a non-judgemental and non-critical collaborator which might all the same inherently provide feedback in the form of co-creation for students to work with in gaining confidence and experience in writing: 'GenAI served as an unfamiliar site to affirm their equality in this experimental endeavour, fostering the exploration of a creative product without predefined answers.' (4) The use of GenAI in this role allowed for the decentralisation of the teacher presence, in that 'we explicitly communicated that we did not have much more expertise than the students and therefore were not able to correct or provide answers.' (4)
But, even setting aside scepticism that GenAI can be approached as a non-judgemental and non-critical co-collaborator, a significant component of managing the safety of the classroom is managing cultural safety -- something that is potentially risked in Tsao and Nogues' approach. Jerlyn Ho et al. make clear that AI in its programming and in its sampling from pre-existing datasets is biased towards subjective and judgemental positions which exist as 'ghosts in the machine' (1), and are all more insidious for the seeming objectivity of AI analysis. In Tsao and Nogues' work, underpinned by Rancière's drive towards emancipation from linear models of knowledge transmission and their implicit hierarchies of knowledge (Rancière 15), the imperative to decentre both positionality and authorship risks removing incidentally subjective responsibility to cultural sensitivity in the classroom. Ho et al. examine the biases embedded within the neural networks of GenAI, 'present in training data, model architecture, and user interactions' (4). They argue such biases begin at conception, with 'Large Language Models' (the base architecture supporting GenAI programs such as ChatGPT) 'subject to bias or inaccuracies present in its dataset. Many stereotypes, including gender and racial ones, are often sown at this stage' (4). Engaging with GenAI as a co-collaborator risks inherently bringing such biases into the development of work, and subsequently into the classroom. The position paper Indigenous Protocol and Artificial Intelligence supports this idea -- number five (of seven key protocol) is to 'Recognise the Cultural Nature of all Computational Technology' (Lewis et al. 22).
There are similar reasons to question Tsao and Nogues' approach to their classroom experiment as well. While they acknowledge some of the dangers of AI globally, they put forward highly positive outcomes, and ultimately 'present the possibility of GenAI in promoting a more distributed and diverse conception of creativity accessible to all.' (10) But aspects of their pedagogical design can arguably be seen as biased in similar ways to the design critique of Ho et al. For a start, their sample group of student participants, as Tsao and Nogues themselves recognise, 'primarily consisted of students with higher educational backgrounds and potential socioeconomic advantages, who may hold views on art and creativity that do not fully reflect the wider population.' (4) They acknowledge this as a warning against generalising the results within broader teaching. But this is potentially suggestive alongside of a lack of cultural and socioeconomic diversity in the group, with the relative social and economic privilege of their 'primarily local' (4) students perhaps minimising imperatives for the trial to address (or approach) questions of cultural safety in the learning experience. Cultural safety is multifaceted and intersectional, something GenAI flattens -- erasing all forms of diversity within its sample as it homogenises output. This perception of privilege might have contributed to the lack of reflection in the article around the risk to the cultural safety of students as they engaged with GenAI as a review and collaborator/co-author in their work; a risk born of the overwhelming whiteness of the data subset that GenAI models tend to draw from, including those used in the project [2]. This same design in an Australian context, for instance, would leave Indigenous students incredibly vulnerable to implicit forms of colonial violence via that same whiteness and the homogenising force of GenAI algorithms as a whole. The need for the Indigenous Protocols in Lewis et al.'s work to include 'Relationality and Reciprocity', 'Responsibility, Relevance and Accountability', and 'Develop Governance Guidelines from Indigenous Protocols' as numbers two, three and four suggests the respective general lack of these attributes in contemporary technology (Lewis et al. 21).
Extending on this risk is the ethical question of working with GenAI programs which have been proven to have employed stolen IP in training their algorithms. Again, this is a phenomenon which leave Indigenous peoples specifically vulnerable, involving the theft of sovereign knowledge and communal cultural and intellectual property. This dilemma was not approached in the study, which opens instead with Tsao and Nogues' definition of GenAI as utilising 'extensive unstructured and publicly accessible datasets' (1). It is, however, potentially a factor which has influenced the findings, in the sense that the creative work produced by GenAI in the trial was shown to adhere broadly to precepts of craft, defined by creative conventions which students were then encouraged to critique and build from. While Tsao and Nogues argue that this 'reinforced the consciousness, emotional intensity, and a drive for action in the creation process that was evidence of intellectual emancipation' (8), this is also predicated on a responsiveness to convention, working from a base position derived from homogenised data. Matthew Salesses aligns craft and sociocultural expectations, describing the risks inherent to teaching craft: 'Craft does not separate the author from the real world ... Craft is never neutral. Craft is the cure or injury that can be done in our shared world when it isn't acknowledged that there are different ways the world is felt.' (22)
Place-writing, with its capacity for contemplating situatedness and positionality, can enhance the latent capacity in creative writing for (as Salesses puts it) 'acknowledging the different ways the world is felt'. Dominique Hecq emphasises the manner in which 'the practice of writing is an experiential activity which mobilises both unconscious and conscious processes' (7), producing an embodied knowledge -- it 'entails seeing, knowing and being in the world: as such creative writing is a perspective, an epistemology and an ontolog' (4). Directed towards a contemplation of being-in-place across these strata, this capacity in creative practice can constitute a powerful means for both the consciousness and articulation of standpoint. Recognising this capacity, in an interview on settler poetics recorded for my poetry unit, Nadia Rhook speaks similarly of the 'double vision' of poetry, through which a settler subject might try to see the living operation (and privilege) of colonialism in their life, as well as simultaneously working to cognise Indigenous sovereignty.
Catherine (12th April 2025)
In 2020, Westerly awarded the Patricia Hacket prize to Koreng Wudjari Noongar poet Cass Lynch, for her work 'Five Haiku'. The poems, originally published in Westerly 64.1 (2019), were released free-to-access after the prize was announced, and I immediately began to use them as a teaching text, compelled in part by the complex interactions they suggest between language, translation and the forces of colonisation. Using bold font to prioritise Noongar language, each haiku is comprised of the Noongar original, a poetic translation and a direct translation, both in English. These two versions surround the Noongar text. The first reads:
The weeping river / gathers salt from the earth / to embitter the sea
Waliny bilya / baal wedjan djalam boodj-ool / warn wardan nyorn ngibart-abiny
Crying river / it gather salt ground-from / make ocean sad poison-become
(Lynch 21)
Craft, in Lynch's poems, lies in the deliberate placement of the translations, acting each against the other, and in their completeness to speak to an embodied Noongar identity. The poem is not any one version of the haiku, but all three versions together. They turn on the experience of a being-in-place made particularly powerful in my classroom by the students' presence on the Country described within.
Is GenAI craft? Can it produce craft? Salesses argues that to use craft is to take up and engage with a pre-determined audience, to sit within a shared bias for the structuring of creative work in a set way (23). GenAI does this, but unknowingly so in its production of text. In stark contrast, writing into the continued colonisation of so-called Australia, Lynch -- a member of the Wirlomin Noongar Language and Stories project, which focuses on language and culture revitalisation ('About') -- is simultaneously empowered in reclaiming Noongar language, and wielding it within a work crafted to mourn the damage that colonial violence has inflicted. The work anticipates the predominance of settler-readers (leading with the poetic translation), but simultaneously insists on offering space to Noongar readers in acknowledging the primacy and importance of language to place, and the myriad meanings possible within the Noongar text (via the transliteration). It meets, in this, Salesses's definition for craft of producing an embodied knowledge which positions its reader accordingly. The work acts to encourage the settler-reader to engage in the same critique of language primacy and translation within the colonial context. In comparison, in its unknowingness, there appears to be an argument to understand GenAI as the opposite of craft, as detached from such systems of sociocultural expectation. It replicates rather than accepts or understands the rules of craft. In this sense, it could potentially be seen as a tool akin to Rhook's 'double vision', to push students to cognise their own expectations as to how a text should be composed, in making accepted norms of composition amusingly and/or alarmingly blatant. This is in part the argument taken up by Tsao and Nogues. But there is a more insidious risk within this that GenAI, even when employed productively and in complement to student writing, cannot make visible its audience, inherently pre-determined in the way Salesses describes. Rather, it obfuscates its engagement with bias for the end user. In considering GenAI as compared to a poem like Lynch's, the idea that it could be viewed as detached is immediately questionable. Where GenAI's reader is assumed via the same flattening logic as its text-production, writing for the lowest common denominator within its sample, Lynch's work demonstrates the careful contemplation of the reader that we encourage students to consider. In the place-writing exercise we set our students, imagining who they are writing for is another kind of imagining place.
Each GenAI program inherits craft and bias as functions determining what constitutes prose and/or poetic structure and form, in the coding for generative response and in the (white, Western) sample material it draws from. Reiterating the position of Ho et al., James Hutson makes clear, 'Generative AI models undergo training on existing human-made art, thereby forging a link with human creativity and the cultural context from which it arises.' This link is multivalent -- GenAI draws from the products of human creativity (and their cultural context) to produce work which mimics and thus extends that context, in the same effect as an echo chamber. Ho et al., in discussing predictive AI (underpinned by the same neural network technology as GenAI), describe this effect as a 'feedback loop where the algorithm's predictions are continually validated by the influx of data generated' (2). It risks therefore flattening and compromising diversity of thinking in the classroom in the same way -- it is incapable of writing 'from' or 'for' any specific position. It can only write 'of' its sample, with the knowledge it produces constrained by mimesis. In the same way, as a friend (Mike Griffiths) posts on Facebook, the thinking that students are begin trained to do is shifted accordingly: 'we aren't preparing students to drive poesis, only to accelerate invidious mimesis.' (Facebook post, 10 April 2025) The mimesis that results, moreover, is based in the replication of pre-existing bias.
The non-place of GenAI is thus not whole nor utopic but constrained by inherited systemic sociocultural bias.
Caitlin (7th March 2025)
A humorous -- if slightly concerning -- example of this occurred in my own place-based creative-critical practice as I was responding to Catherine's notes on another paper. In this paper I had used the term 'so-called Australia', and Catherine had suggested a note to explain the critical lens around the term for an international audience. When I placed the term into Google to see what would happen, the below was returned at the top of the page under 'AI Overview':
the expression 'so-called' Australia is often used to question the legitimacy of Australia's claim to continent-hood, as it's a relatively small landmass compared to other continents like >Africa or South America. (Google 7th March)
I took a screenshot and sent it to Catherine. We laughed via emoji but it was unsettling on two levels: the obvious being that while we immediately recognised the flattening of geological continent with geopolitical nation state as a function of bias in the code, the same might not be true for others, including our students; the second and more insidious was how I was unable to disable the AI Overview function of Google, while I can choose (at this stage) to engage with ChatGPT many other AI integrations are not as easy to separate oneself from and do not seek active consent.
It is clear then that at the same time as flattening the diversity of the classroom, GenAI flattens embodiment within that space, and the outcomes made possible accordingly, both for teachers in the moment, and potentially for students vocationally. Students engaging illegally with GenAI are effectively stepping outside the ethical responsibility of cognising positionality within the classroom, but without ever leaving the sociocultural context which makes an awareness of positionality so vital. This is, perhaps, of particular relevance to the online classroom, wherein the need to protect inclusivity is heightened by the lack of containment to the space, and the manner in which it brings people together whilst inhabiting different places. Rachel Adams argues for the need to decolonise GenAI, including in developing 'an alternative imaginary for transcending the normative binaries that AI fortifies' (261). Introducing the imperatives behind the Indigenous Protocol and Artificial Intelligence position paper, Lewis warns:
we are concerned that the Western rationalist epistemologies out of which AI is being developed are too limited in their range of imagination, frameworks, and language ... If we insist on >thinking about these systems only through a Western technoutilitarian lens, we will not fully grasp what they are and could be. At best, we risk burdening them with the prejudices and >biases that we ourselves still retain. At worst, we risk creating relationships with them that are akin to that of a master and slave. (6)
In contemplating the classroom as a place, this risk extends into the pedagogical structure of the relationship between student and teacher. If any aspect of the learning environment is structured around or involves AI -- which it almost invariably will, when we note the predominance of student interaction with AI -- relationships with those programs become part of the student-teacher dynamic. In arguing for Indigenous Protocols and the active involvement of these in AI development, Lewis suggests ontologies of kinship in particular 'can point us towards potential approaches to developing rich, robust and expansive approaches to our relationships with AI systems' (8). Protocols developed on this basis 'reinforce the notion that, while the developers might assume they are building a product or tool, they are actually building a relationship to which they should attend.' (8) By the same thinking, developing an awareness of positionality as a teacher, and encouraging students to do likewise, is a means for cognising relationality that might be applied to thinking about GenAI. Building on Lewis's distinction between product and relationship in development, it might also be taken as a feature of practice distinguishing text creation from text production. Where production is extractive and placeless, creation is relational and situated.
Catherine (8th November 2023)
Caitlin emails me from Yamatji Country, and her description of place prompts me to recentre my own being on Whadjuk Noongar boodjar, the derbarl yerrigan close by. At my computer, I am both here and online -- a simultaneous subjective state which in the function of preposition approaches something of impossibility, alterity. The online classroom is sympathetic to the state of alterity, and engaging with GenAI in this context can serve as a prompt to contemplate what subjectivity is and means when rendered through the digital. There is value to be gained here. Apart from anything, there is real reason in the face of climate crisis to develop our ability to engage with states of alterity. But it cannot be at the cost of an ethical engagement with the world -- 'real' and/or online.
Structures of Value
Understanding and responding to the distinction between production and creation is implicit in extending any critique of GenAI in creative writing. This is a nexus point in our dialectic in the sense that this distinction implicates all other concerns. Where creation implies some aspect of originality, production is suggestive of a process of making or forming which is dependent on the supply of raw materials. Without completely eliminating the possibility of GenAI as creative -- which would entail a broader philosophical conversation around the definition of creativity -- it is necessary to note that GenAI is undeniably invested in production. This is implicit in the technology itself. The majority of Large Language Models (the base model for GenAI) function as 'transformers' -- this is, for instance, the 'T' in ChatGPT, a 'generative pre-trained transformer'. Transformers, developed by researchers from Google in 2017 (Vaswani et al. 2), convert text into 'tokens' which can then be contextualised via an 'attention model'. The revelation of the 'transformer' was to base text output exclusively on multi-head attention (including self-attention, meaning the output text becomes considered as input as it is generated), rather than employing a recurrence model to attach significance to key tokens (Vaswani et al. 2). As the Google research team boast, this 'allows for significantly more parallelization and can reach a new state of the art in translation quality after being trained for as little as twelve hours' (Vaswani et al. 2). The tokens taken from the prompt are assessed according to word embedding tables, a form of distributional semantics, allowing the mechanism to effectively define and respond to key tokens in reference to the data of its training (Vaswani et al. 5). In approaching this development as a creative writer, we can see that the output of such models is constrained, however, not only by this data, but also by the semantic function by which it pays attention.
Caitlin (23rd April 2025)
Sidebar: for the longest time I said that Michael Bay's Transformers 2 was my favourite movie, precisely because it required absolutely no attention. Lacking any plot, and balanced on the twin-axels of explosions and intergalactic exiled robotic lifeforms, it was just colours unspooling on a very large screen as the human and robot heroes travelled from location to location. It was a kaleidoscope of a film; in that it had all the same elements of all the other Hasbro films (for instance the GI Joe movies) but shuffled (rather than transformed) into a new order. As David Edelstein writing in New York Magazine says, 'Much of the movie is just computer-generated hash, weightless even with nonstop BOOMS and METAL GROANS and THUDS' (6 July, 2009), which I appreciated.
Paying attention is a key tenet of place-writing -- fields such as deep mapping turn around the premise of careful and prolonged attention to a specific site. With attention, in human creation, generally comes respect -- it is the act of valuing the complexity of a place as a whole. This too can be transformative, and when it is read, that experience becomes collective. But GPT attention is singular even when 'multi-headed' -- it flattens the collectivity of engagement between an author and readers into a one-directional system of assessment, producing a weighted set of valued components. What is transformed is thus the input and data combined, rather than the subjective experience. There are broader questions of ethics at play in this context as well -- as noted above in relation to the work of Tsao and Nogues -- including that of piracy and plagiarism in the programming of GenAI, let alone as an issue in student applications. Production is predicated on the use of extant IP, not always legally employed. The acceleration of technology development has exacerbated this. Matthias Bastian reported in 2023 that GPT-4 had one trillion parameters (the number being 'an indicator of its ability to recognize complex patterns and relationships in data and develop new emergent capabilities' (Bastian)), increased from GPT-1's 117 million. But in this, he as relying on anonymous sources, as this information was no longer disclosed by OpenAI, in the same way that the sources of training data have not been disclosed since 2020 (Bastian ; OpenAI). And this year, the exposé released by The Atlantic evidenced that Meta engaged in wide-scale IP theft via literary piracy site LibGen to train its flagship AI model, Llama3 (Reisner).
Catherine (28th^ March 2025)
The news of the Meta theft has broken, and my Facebook feed is filled with friends decrying it. I use the search tool and find my own work too has been stolen, message Caitlin via a Meta platform and hear hers has been too. The emotional responses within our cohort seem to range from shock to anger, to cynicism, to a vague humour that one's work has garnered this attention. I feel all this, and also a deep sadness that the ten years of work that went into writing my novel might now been reduced to the encounter of milliseconds from a non-human reader.
Our consciousness of GenAI's theft and piracy is not new, as much as it might suit corporate interests for this to be assumed. The 'Shaxpir' bot, an early and privately owned GenAI program specific to creative writing production, was found to have stolen IP and enabled plagiarism back in 2023. Holden Sheppard is a West Australian author based on Whadjuk Noongar Country, and one of many authors globally whose work was stolen and uploaded to the 'Shaxpir' bot. He argues that 'Because all AI is trained using texts that were -- and still are being -- fed into it without the consent of the artist, any writer using G[en]AI for their own work is participating in the infringement of other artists' copyright, to which they did not consent, do not know about, cannot opt out of and are not compensated for.' The risk for Sheppard, thus, is literal -- there are genuine implications here for income and career.
There is a vocational imperative to producing graduates capable of contemplating and critiquing AI, which is partly a question of market: the ability to argue for and protect the originality of human creativity. But at the same time, the ability to productively and sensitively employ AI is a skill-set many employers are now seeking, including in the public sector. This has been recognised by Tertiary Education Quality and Standards Agency (TEQSA), whose guiding principles for assessment and AI, begin with the principle that 'Assessment and learning experiences equip students to participate ethically and actively in a society where AI is ubiquitous' (2). The ability to critique AI can be seen as a measure of degree-value in broader terms, though, too. In a context wherein there is genuine anxiety amongst the student body and the creative industries that AI will steal jobs, graduates able to explain and convey to an employer the need for human response to AI function, and the necessity in ethical terms to critique AI whilst still productively leveraging what it can offer, are powerful examples of the base value of the critical thinking that an Arts degree can confer. More specifically, the value of creative writing as a discipline of study is made visible in the possibilities it offers to approach such critical discourses in both these ways.
As an example, in their April 2023 'Editorial Position Paper' for Education, Technology and Society, Gwo-Jen Hwang and Nian-Shing Chen provide a series of recommendations for how to best instruct students to use GenAI in writing tasks. Key here are to 'be specific', 'provide context' including where 'your question is situated or based', and to use 'role play', where you 'tell G[en]AI what role you want it to play, and ask it to act as a specific role'. Applying this thinking in the context of place-writing exercises akin to our creative mapping, we can see the possibilities this might offer students to engage in the ways Caitlin suggested before, expanding their own maps consciously via AI. But at the same time, the context of plagiarism suggests a need for some caution here, in terms of the implications for the writers unseen behind the program.
Moving iteratively, these recommendations highlight the limitations of GenAI -- its inability to be specific or situated, its lack of consciousness as to its own implicit sense of positionality and audience. Critiquing this approach highlights conversely the value of these capacities in the writers we produce. Doing so in the classroom might equip students with the ability to articulate for themselves the value of their own creative skills. It also potentially places the discipline of creative writing -- along with many other creative practice disciplines relatively new to the academy -- outside changes to university wide learning and teaching philosophies particularly in relation to assessment. For instance, the recent adoption of language at multiple universities across Australia around 'lock-down' browsers or exams, 'secure' and 'insecure/open' assessments. What is particularly interesting here from within the framework of the place-based creative writing classroom, is the emphasis on location -- or place -- in securing assessments. The University of Sydney's 'Two-Lane' approach to AI (lane one: secure vs. lane two: open), specifically states that:
Any assessment which is not supervised in a secure face-to-face environment is 'open' or 'unsecured' and hence a de facto (unscaffolded) 'towards lane 2' assessment. This includes online >quizzes, take-home assignments, and orals or presentations held using web-conferencing software. (Bridgeman and Liu)
In order to compensate for the placelessness of GenAI, we see universities altering the frameworks of the long-held push towards hybrid learning, to reinscribe the value of in-person learning, if only for assessment purposes.
In the University of Sydney's 'Indigenous Strategy 2021--2024' -- 'One Sydney, Many People' -- the vision statement reads:
Through our shared responsibilities to the Aboriginal Lands upon which the University stands, we create a genuine sense of belonging among all students and staff. As one of Australia's >most eminent universities, we demonstrate visible leadership by fulfilling our social contract with Australia's First Peoples. (7)
Obviously, universities serve many purposes and have many strategies to do so (as evident by the many strategy documents); but, thinking from within our creative writing pedagogies of place which forefront positionality, one cannot help but notice a tension between the rapid adoption of AI -- and its learning, teaching and assessment strategies -- and the values and visions of many universities' commitments to First Nations Peoples. In the Sydney strategy, responsibility to Aboriginal Lands is centralised, further in the document 'Environment -- A sense of Place' is one of four 'Strategic Focus Areas' (10). At a basic level, we have already discussed how the adoption of GenAI in the classroom is at odds with this responsibility in terms of the current environmental costs of GenAI, while the flattening of the specificities of place into a generic 'setting' is reductive and not responsive to Indigenous understandings of the specific Countries in which we live. In the introduction to Indigenous Protocol and Artificial Intelligence, the authors open by saying 'Our aim, however, is not to provide a unified voice. Indigenous ways of knowing are rooted in distinct, sovereign territories across the planet. These extremely diverse landscapes and histories have influenced different communities and their discrete cultural protocols over time' (Lewiset al.4). In our place-responsive creative writing classrooms, we ask our students to consider positionality as a way of asking them to think about where their own individual voice comes from and to whom or where it is responsible. This is not something which can be easily or thoroughly accomplished with GenAI.
Caitlin (22nd April 2025)
It so happens that in teaching from place, I also reached for the same set of poems by Cass Lynch. I teach the poems in week one of the 'Writing Poetry' unit I coordinate, as part of the introductory business of getting students acquainted with one another and also grounding them in the locale of the university which is on unceded Whadjuk Noongar boodjar. I ask students to think of a place they know well, often, but not always, these will be places also of Whadjuk Noongar boodjar. The students are asked to identify seasonal markers -- kigo -- for those places. Then before drafting their own haikus, there is discussion of Lynch's sequence framed by a short, informal artist's statement offered by Lynch over email (pers. comm. 13 Jan. 2023):
When I started I was putting them on Instagram, and I have a lot of Noongar mob following me, so when I wrote the first haiku ... I wrote it in Noongar, and then I provided what I call an '>English poetic translation' which captures the vibe of the Noongar words, but then I added the literal translation so my Noongar followers wouldn't be locked out of their language, I made >it transparent which words meant what ... the haikus capture the translation process in real time, the haiku form and the three lines of translation have escaped the binary of translation, >they make translation a prism or a lens flare rather than a reflection.
The poems originated on Lynch's Instagram as a way of writing into and responding to Country, and are -- as her artist note makes clear -- intended for specific audiences. As Catherine notes above, a responsible reading of these poems is beyond ChatGPT, in a similar way that being responsible to place requires paying attention, listening and being in place, all of which are not easily engendered by GenAI.
Conclusion(s)
The questions we have raised here are contingent on and complicated by the continuing evolution of GenAI, which we have witnessed over the space of this research and attempted to signal in the shifting timestamps of this essay. This is a constantly emerging pedagogical space, and as teachers we will always be behind the eight-ball in dealing with the follow-on effects it might have for creative writing students. This implicates (and will continue to implicate) students' writing within the classroom, but also more conceptual questions of what it means to teach place-writing, what it means to write in place, and to receive writings from place. It implicates alongside the vocational outcomes and aspirations of students, both in affecting the writers they are becoming, and affecting the industries they will move into. As we have attempted to recognise throughout, the risks faced in this are multiplied for culturally diverse students, and particularly for First Nations students, as AI exacerbates the colonial imperatives at place in the geopolitical space and in the classroom alike -- a double-scale colonial economy which continues importing tools generally devoid of the ability to recognise place. Working with Indigenous protocol in developing AI might in the long-term help counteract this, in the (unlikely) event corporate AI developers take that step. In the short-term, the imperative remains to teach in ways that these forces might be actively cognised and critiqued.
Catherine (7th April 2025)
Working on this article while on parental leave, my baby sits at my feet in her bouncer, occupied by a mythical creature and two invasive species. It was AI that, last night, uninvited, offered me parenting advice on weaning infants when I absent-mindedly searched online for baby recipes. Its bland helpfulness was deceptively benign, until I came to wonder about the predictions it made of her growth, the assumptions it made of her age, the amount it knew already of her life. It is an invasive species within my family home -- in the six months of her living, it has become even more ubiquitous and obvious in search engines and social media. In the same period, it has become clear that my own work, my own IP, has been stolen and used for its training.
I call Caitlin again driving home from the daycare drop-off, my eldest covered by the care of others, and the space of a day before me. I plan to work on our article as my baby naps, but I'm hampered by a deep ambivalence about our subject matter, and the anxiety that a conclusion is not possible. Even as I type this, the shared document in which we draft our work offers predictive text anticipating the direction each sentence will take. I am not sure if I am relieved or discomforted by the regularity with which it predicts this incorrectly.
I will return from my parental leave to a teaching environment which is different to the one I left -- quite literally, with the advent of new assessment policy and AI guidelines. The landscape around creative writing as a discipline is eternally evolving, but more rapidly now than I am accustomed to.
Caitlin (15th April 2025)
I am now writing this from the future (or at least that's how it feels), my son in the gym creche for two hours while I try and survive school holidays. We have been travelling the breadth of different Noongar countries -- Wirlomin, Whadjuk and soon Yued. Passing between them, I think I can see some markers of difference, slight changes in the colour of the sand, the density of the trees, but my eyes are still a tourist's. What stays constant, for the most part, is the connection outwards via the internet, although at Tony's Bend campground in Dwellingup there's no reception, and at Grey Shack Settlement the imported repeater has stopped providing any type of reception and the slow walk up a sandhill is required to navigate any telecommunication. Catherine and I have been writing this via a shared Google doc, and messaging each other via a messenger service we both want to leave because the company which owns it has harvested our work, our labour, without compensation. Very occasionally we talk on the phone, normally on drives from dropping our children places. Catherine often wants to discuss what we will conclude the paper with, there is a separate shared document with dot-points of ideas: theft, ongoing colonisation, the capitalist business of universities versus the creative practice of our classrooms. It's the anniversary of my father's death from an illness he couldn't move past. But this is not how GenAI functions, there is no sudden precipitous change, rather it moves slowly, the way climate change moves incrementally against our small lives. Though it also moves in the way of the increased extreme weather events the shared Country is seeing, such as the drought deaths of the trees along the street verges of my suburb by the Derbal Yerrigan. I tell Catherine I have no conclusion, only the same sense of competing scales I try to get my students to find ways of writing, as well as the same sense of responsibility to this shared place onto which the University maps itself.