See discussions, stats, and author profiles for this publication at: [Link]

net/publication/376519857

The other side of AI: Writing, thinking and creativity in an age of Artificial
Intelligence

Article · December 2023

CITATIONS READS

0 466

1 author:

Steven Thurlow
University of Melbourne
5 PUBLICATIONS 20 CITATIONS

SEE PROFILE

All content following this page was uploaded by Steven Thurlow on 15 December 2023.

The user has requested enhancement of the downloaded file.

Journal of Academic Language & Learning ASSOCIATION FOR
ACADEMIC
LANGUAGE AND
Vol. 17, No. 1, 2023, T92-T102. ISSN 1835-5196 LEARNING

The other side of AI: Writing, thinking and

creativity in an age of Artificial Intelligence
Steven Thurlow
Academic Skills, The University of Melbourne, Australia
Email: sthurlow@[Link]
(Received 24 September, 2023. Published online 9 December, 2023.)

Artificial Intelligence (AI) programs such as ChatGPT have illuminated the

previously quintessential human capacity for creating writing. This paper in-
vestigates the complex intersections between writing, thinking and creativity
through addressing two key dichotomies that characterise any act of writing:
namely, by first framing writing as either recording (already known) infor-
mation or as an act of creation, and secondly, writing as the expression of
convergent and/or divergent thinking. It speculates how each of these con-
trasting forces may intersect with AI in the academic writing space. It con-
cludes by considering implications of these intersections for the practices of
Academic Language and Learning Advisers and academic writing teachers.
Key words: Academic writing; Academic writing pedagogy; Creativity;
Thinking styles; ChatGPT; Artificial Intelligence

1. Setting the scene: The Graduate Student Orientation Fair

I am on the Academic Skills booth at this high-spirited event. An unpredictable tide of students
sweep past my table, many high on free chocolate and lollies. Among them, tight-knit groups of
international students, often clustered by country of origin, linger expectantly by my table. Rais-
ing my voice above the din, I respond to their enquiries with what I hope is a friendly yet knowl-
edgeable air. A tall, male student stops by, looking like he is up for a chat.
So, hat is this thing, this Academic Skills?
I respond with my usual spiel about helping students with their academic and
study skills writing, reading, time management and the like. I ask him what
he is studying and find out he is starting a doctorate in IT systems design.
Do ou think ou ill need help ith an thing like reading and riting? I
ask, a tad hopefully.
A quizzical look overtakes his face.
But no e ha e the AI and it helps so much ith the riting. What can ou
offer more than that?
My face drops a little. Why would he think such a thing?
Unfazed by my reaction, he continues. AI is fantastic; so con enient. No I
need not worry so much about the writing, putting together the words. I can
concentrate on the ideas, the real work of the PhD. The program can fix up
m language, make it better. Makes me feel rela .

T92 © 2023 S. Thurlow

T93 S. Thurlow

He fails to pick up my distress at his words and I offer no defence of my work; this casual negation
of what I do, day in, day out. I have heard many arguments around AI in the past year and wit-
nessed much hand-wringing in the process. But this commencing student s words have the effect
of rattling me more than I can ever let on. It seems my job is on the line here.
His words continue to ring in my ears as I leave the event. Graduate research students, such as
my questioner, clearly appreciate the intrinsic benefits of AI; how it can fix their writing and
therefore, free them up to focus on the real part of their studies generating the innovative ideas
and novel notions that drive their higher degrees. This sets me reflecting on my job and indeed
the whole apparatus of teaching academic writing I have been enmeshed with for the past twenty-
five years. How can we approach learning and teaching academic writing and the thinking that
both rest on in a world rapidly transformed by generative AI? And what is the place of creativity
human and non-human in this new world of large, disruptive and constantly evolving large
language models?

But first things first. Writing has always been enmeshed with technology. Since symbols, letters
and words were first set down in the sand or on stone, writing has moved with the times. Paper,
quills, pens, typewriters and keyboards are all artefacts that have directly emerged from develop-
ments in the technology of writing and the process of capturing words for posterity. In late 2022,
the world of writing was shaken up yet again with the flashy relaunch of widely accessible, gen-
erative Artificial Intelligence (AI) programs. Offerings such as ChatGPT and other generative
language models have shown us how written communication might be approached and accom-
plished differently.
This article takes a step back from the recent development of large language models (LLMs) to
consider the perspectives of both a practising academic writing teacher (myself) and wider impli-
cations for the practice of Academic Language and Learning (ALL) professionals. I begin by
investigating two dichotomies that I believe exist at writing s core:
1. Writing as recording (already known) information versus writing as an act of creation.
2. Writing as the expression of convergent versus divergent thinking.
As part of my analysis of each of these pairings, I contemplate how they might intersect with AI
in the academic writing space.
But who am I to speak about this large and rather daunting topic? I settled upon these two dichot-
omies through undertaking my own doctoral research into creativity in graduate research writing
contexts in my own workplace. During this investigation, I witnessed the struggles of my student
participants Arts doctoral writers all of whom wished to include elements of creativity and
creative expression in their academic work. Many constraints, however, blocked these desires
to write creatively. Uncovering this conundrum coincided with the emergence of LLMs such as
ChatGPT, and it seemed highly appropriate to apply my research findings to these new tools
which have re-focused attention on the place of human creativity in writing.

2. Writing as recording (already known) information or writing as an act of

creation
One useful way to frame the development of writing over time is to position it within the socio-
historical contexts in which it has traditionally occurred. While writing originally evolved to rec-
ord trade and official histories, it developed into a more creative practice over time, facilitated by
the development of simple alphabets. These alphabets enabled people to capture their own
thoughts, wishes and desires in a way that could either be private or public. Thus, although writing
was invented to record or document information, it rapidly evolved into a method to aid thinking
by freeing the imagination and creative thought and therefore, bring new knowledge into being.
T94 Writing, thinking and creativity in an age of Artificial Intelligence

This fundamental dichotomy between writing as mere documentation and writing as personal
creativity is burnt into the soul of writing; it is part of its DNA .
This conflictual duality is reflected in the role of writing in academic contexts. Since the emer-
gence of the first universities in medieval Europe, higher education institutions have frequently
been associated with the preservation of existing information. The convention and conformity
that still characterise many disciplinary discourses spring from this period when scholars worked
with a limited range of sacred and canonical texts that students were obliged to copy verbatim as
they were read aloud by senior scholars. Indeed, it is from the Latin term lectura meaning to
read or read aloud that we are left with the word lecture (French & Kennedy, 2017). Alt-
hough the tradition of academic writing as verbatim recording declined after the invention of the
printing press, vestiges of it still remain in the contemporary university. Indeed, many academics
might agree with Dunleavy (2003) when he frames academic writing as an act of commitment,
a decision to firm up what we think, to prevent constant reprocessing and reconfiguring (p. 26).
In this process of preserving information, learning about what is expected and acceptable in aca-
demic writing and then adhering to an existing template when producing a written piece still oc-
cupies a major part of any student s experience of writing for the Academy.
Existing alongside this pressure for conformity in writing, however, the presence of imaginative
creativity can also be felt in academic writing contexts. At a fundamental level, it is there in the
act of committing words to the page or screen. For, before the writer writes, there is nothing; a
vast emptiness. Then, the physical act of writing occurs and becomes visible to a reader. Thus,
even though written script can perform so many different functions, the conjuring trick of pro-
ducing a piece of writing is inherently creative, if not downright magical. In fact, Yagelski (2011)
observes that during the act of composition, so engrossing does the physical act of writing become
that it is almost synonymous with [one s] consciousness [and] in this sense, is a way of being; it
is an ontological act (pp. 103-4). In other words, writing brings ideas to textual life.
As well as ratifying or extending what is already known, writing can also function as a way of
knowing or bringing new knowledge into the world (Richardson, 1998; Locke, 2015). In fact,
Ong (2015) believes it has the capacity to change the way we think, asserting that written literacy
has transformed human consciousness through a fundamental restructuring of our thought pro-
cesses. Confirming this wider and intrinsically creative dimension of writing, Hayot (2014) de-
clares:
Writing is not the memorialization of ideas. Writing distils, crafts and pres-
sure-tests ideas it creates ideas You cannot know what your ideas are, or
mean, or do until you set them down in sentences. (p. 1)
In sum, this framing of writing as simultaneously a process of thinking and as a method of dis-
covery is in almost direct contrast to the idea of writing as mopping up after knowledge has
been created; an exceedingly common and in some disciplines, almost ubiquitous, view of its role
in higher education.

3. AI intersections and affordances

Enter AI. As we attempt to reconcile the intensely human act of writing with what emerging
LLMs can provide, many questions emerge. For instance, how will these new tools impact on the
thinking skills required for any piece of academic writing? And, by potentially removing the em-
phasis on the creative and/or original thinking as lying at the base of writing, are we returning to
an emphasis on writing as recording as opposed to writing as creating knowledge? Ultimately,
comparing these programs to human writers, how effective are they in the business of knowledge
creation and dissemination so valued by the Academy? Recent research would suggest that they
may be equal to, if not better than, many student writers in the performance of several core aca-
demic tasks such as reflective writing (Li et al., 2023), although other scholars view its creative
abilities more cautiously (for example, Cropley, 2023). I cannot pretend to offer any definitive or
T95 S. Thurlow

indeed, comprehensive answers to these important and complex questions, rather, tapping into
the spirit of the zeitgeist, I will ask ChatGPT to respond.1
My first question gets straight to the point: How well can ChatGPT perform in the generation or
creation of ideas required for a piece of academic writing? As per usual in the world of AI, the
ChatBot is not backwards in coming forward to praise its capabilities in this area, albeit with a
standard let out clause:
Chat GPT can be a valuable tool for generating ideas required for academic
writing, but its effectiveness may vary depending on the specific topic and
context. Here are some ways in which ChatGPT can assist in generating ideas
for academic writing. (OpenAI, 2023)
We then hurtle into the predictable, multi-bullet pointed list format seemingly treasured by all AI
ChatBots. It spits out eight points in the blink of an eye (OpenAI, 2023):
1. Brainstorming ideas
2. Outlining and structure
3. Research directions
4. Thesis statement development
5. Content expansion
6. Citation suggestions
7. Counterarguments and analysis
8. Conclusion and implications
Analysing these in relation to our previous discussion of writing as thinking, I note the signifi-
cance of point #4 on its list ( Thesis Statement Development ). On this point, ChatGPT confi-
dently asserts that it can, help you formulate a clear and concise thesis statement that summa-
ri es he main arg men of o r paper and s gges a s o refine and s reng hen o r he
sis (OpenAI, 2023). Sounds wonderful, but I probe for more information. As expected, I get a
degree of backtracking from its initial bold claim:
In point number 4, I meant that ChatGPT can assist in formulating a prelimi-
nary thesis statement for your academic paper. It can provide suggestions
for how to structure and phrase your thesis, helping you articulate the main
argument or point you intend to make in your paper. However, refining and
finalizing the thesis statement to accurately represent your unique perspec-
tive and argument would typically require human judgment and expertise.
(OpenAI, 2023)
In other words, it seems that we are not going to be given a definitive thesis statement, only help
articulating and refining one that we might have to come up with ourselves. This is significant as
any ALL Adviser would recognise that this task the creation of a controlling argument that
drives an entire paper forward is one of the most important tasks for student writers and one
of their most fundamental challenges. However, I am left wondering if Chat GPT always keeps
to these strict, self-imposed boundaries around thesis statements and suspect that many student
writers would expect it to come up with a ready-made argument for their papers.
Interestingly, I note that three points from ChatGPT s list of eight ways it can assist with idea
generation (Numbers 3, 5 & 6) offer the promise of insights and examples on a writing topic based
on literature in the field. However, ChatGPT proves to be a little coy about exactly what material

1
Note: Throughout this paper, responses from ChatGPT are identified by the use of a sans serif font which
are either placed within quotation marks or block indented.
T96 Writing, thinking and creativity in an age of Artificial Intelligence

on which it bases its ideas. When pressed on what constitutes this source material, it suggests that
a critical mind may be necessary to verify the relevance, accuracy and appropriateness of the
evidence and examples provided by the model as they may need to be tailored to your specific
topic and context (OpenAI, 2023). In a nod to the superiority of human capabilities in this area,
the bot advises that ultimately, human judgment and critical evaluation are crucial in determin-
ing the quality and appropriateness of the evidence and support suggested by ChatGPT
(OpenAI, 2023). This discussion highlights a major criticism that has often been levelled at AI
programs, namely, their reliance on spurious, overly general or indeed, fake references.
From an academic perspective, Number 6 is worrying. Upon further prompting, ChatGPT advises
that its basic analysis might not be deeply rooted in a specific academic discipline or belief
system (OpenAI, 2023) and that, again, critical assessment and refinement by real-life humans
may be needed to align the information it provides to a specific academic context. This admission
draws attention to the preference shown by AI programs to focus on general information rather
than on the specialised and disciplinary-specific knowledge more valued by universities.

4. Writing as the expression of convergent and/or divergent thinking

Thinking about thinking is of paramount importance in academic writing contexts, with one useful
approach being to consider how the act of writing can incorporate both convergent and divergent
thought processes. Razumnikova (2013) views convergent thinking as involving the search for a
single correct answer, whereas divergent thinking concentrates on producing a large number of
appropriate and adequate alternative responses and often is associated with creativity which in-
volves the generation of varied, original, or unusual ideas in response to an open-ended task (p.
551). In other words, convergent thinking reins in creativity through the application of rational
thought, implying a form of thinking that follows an expected path or route; not rocking the
boat . In contrast, divergence takes thinking into risky and imaginative unknown territory and
may be more concerned with connecting diverse elements rather than reaching consensus. While
it is commonly believed that human creativity is associated with divergent thinking, it can also
extend from a base of convergent thinking. For example, a creative product resulting from diver-
gent thinking may have been first conceived of as a more standardised or mundane version of its
innovative self.
With its focus on reaching one correct or logical answer, convergent thinking would appear to be
more attuned to those fields of study that may value a single answer to a difficult problem. Con-
versely, divergent thinking, with its encouragement of critical thinking processes such as assess-
ment of information, problem-solving and finding innovative solutions, collocates with the higher
order tasks required by areas which prioritise independent research (Strongman, 2013). Zemits
(2017) notes its special importance to Humanities students who need to step outside the writing
of experts and to reflect on what might be missing in a line of thinking (p. 176). This openness
to other viewpoints demands divergent thinking.
But how does all this apply to academic writing? Pennington (2014) sees these two thinking types
either irrational/chaotic (i.e. thinking divergently) or rational/ordered (i.e. convergent thinking)
as both necessary at different stages of any writing project, believing that writers with a creative
mindset need to be able to flexibly switch between them to effectively accomplish different
stages of writing work. For example, initial brainstorming for ideas on which to write may involve
chaotic thinking, whereas planning writing calls for the operation of an orderly mind; a difference
Greetham (2008) describes as being between the writer as artist and the writer as editor or crafts-
man [sic]. This division necessarily involves a shift in focus from the creative activity of con-
verting your ideas into language to a more self-conscious focus on the way you ve used words,
phrases and structures (Greetham, p. 280). Many writing specialists would concur that effective
academic writing requires both the creation of ideas ( the artist ) and conversion of those ideas
into acceptable form and/or expression ( the editor ). While on a personal level we might be
T97 S. Thurlow

drawn to one or the other, a seamless merging of both is a prerequisite for any piece of writing to
truly fly.

5. AI intersections and affordances

The development of AI programs has the potential to illuminate the place of convergent and di-
vergent thinking styles in academic writing. When I ask ChatGPT how these patterns of thinking
might relate to academic writing, it begins by giving me a handy (but unsourced) definition and
breakdown of convergent thinking:
A process that involves narrowing down options to find the best solution or
answer to a specific problem or question. It's characterized by a linear and
focused approach. In the context of academic writing, convergent thinking is
important for: thesis development; organising and structuring [content]; crit-
ical analysis and; problem-solving. (OpenAI, 2023)
That is all well and good, but AI programs such as Chat GPT often tap into another, more insidious
form of convergence, for ChatGPT reproduces information often North American in origin and
resulting from a mash up of academic and non-academic sources packaging it all up using a
determinedly convergent writer s voice. In this way, AI bypasses the messy irrationality of in-
progress writing with the presentation of an easily digestible (but frequently trite) product to the
consumer in a matter of seconds. Adding to the allure of this glossy package, the voice of authority
through which this information is delivered makes it difficult to resist.
Regarding divergent thinking, ChatGPT recognises the messiness inherent in this alternate think-
ing style when it states:
Divergent thinking is a process of generating a wide range of ideas and pos-
sibilities without necessarily focusing on finding a single "correct" answer. It
encourages creativity, flexibility, and exploring different perspectives. In ac-
ademic writing, divergent thinking is relevant for: brainstorming and idea
generation; developing original insights; innovation; addressing complex is-
sues and; creativity in expression. (OpenAI, 2023)
Teaching writing deep inside the Academy, I find the final item on the list ( creativity in expres-
sion ) particularly interesting and truth be told somewhat liberating. When I ask ChatGPT to
explain it further, it comes up with a refreshing list of techniques including the use of colourful
introductions and memorable metaphors. From my own research, however, I note that these forms
of creative expression are often strongly resisted by academic readers (Thurlow, 2021), and any
writer seeking to introduce the balm of creativity into their academic work needs to tread care-
fully.
On balance, however, ChatGPT concedes that both styles of thinking are necessary in academic
writing, sensibly asserting:
Both convergent and divergent thinking are valuable in different stages of
the academic writing process. Convergent thinking provides the structure
and focus needed to present a coherent argument, while divergent thinking
adds creativity and depth to your ideas. Effective academic writing often in-
volves a balance between these two thinking patterns, adapting their use
based on the specific requirements of your topic, discipline, and writing
goals. (OpenAI, 2023)
Interestingly, ChatGPT appears to recognise the vast difference between these two thinking styles
and, when prompted, categorises itself into the convergent side of the equation:
T98 Writing, thinking and creativity in an age of Artificial Intelligence

The answers I provide are primarily the result of convergent thinking, rather
than divergent thinking. This is because my responses are based on patterns
and information present in the training data I've been exposed to. I analyze
the context of the question, identify relevant patterns, and generate re-
sponses that align with the patterns and knowledge I've learned.
Divergent thinking involves generating a wide range of creative and unique
ideas, which is a more human-like cognitive process. While I can sometimes
offer creative examples or alternative viewpoints, my responses are more
focused on providing accurate and relevant information based on existing
pa erns in he da a The e en of crea i i and inno a ion in he re
sponses I provide might be limited compared to what a human writer with
the capacity for divergent thinking could offer. (OpenAI, 2023)
This last point is significant. It highlights the inability (or unwillingness?) of AI to go beyond the
data, to travel on unmarked paths that might well lead to the novel or original knowledge that is
more valued by the Academy. It stays on the well-trodden path of what is already out there
tirelessly repackaging the efforts of others in a dogged but endless quest to satisfy its users. An-
thropomorphising it, I can see AI as a plodding but determined student writer, not yet a flexible,
highly professional writer, exceedingly keen to please its mistress or master and learn from its
mistakes, but with a somewhat over-confident attitude.

6. Implications for ALL practice

The conversation between a student and an ALL Adviser (myself) which opened this piece show-
cased two very different reactions to the potential of AI and highlighted emerging tensions at
universities as they respond to this new technology. As you will recall, the graduate student ac-
centuated the positive particularly the way in which AI might enable him to focus on his research
without the pain of writing up his doctorate. From my experience with student writers, this atti-
tude might be one that is reasonably common, characterised by an enthusiastic embrace of AI
technology with a lack of knowledge about (or a negation of) the enduring link between thinking
and writing. Other students, however, are more cautious, reflecting their instructors concerns and
institutional alarm regarding academic dishonesty and the lack of transparency about source ma-
terial, among other issues. My own reactions of cautious distrust may also be typical of writing
practitioners working at universities, coupled with fretful feelings as we contemplate losing our
roles as trusted guides to the occluded world of academic writing by generative AI tools.
But, as ALL advisers and teachers, how might we respond to our AI-curious students either
those displaying the firm opinions of my friend at the Orientation Student Fair or more hesitant,
or indeed, silent others? How might we, as the professionals in part responsible for the writing
development of post-entry students in Australian higher education, view the emergence of the
brave new world of AI and the shiny promise it brings to student writers? These questions deserve
our time as we grapple with the seismic changes to our roles created by the on-rush of generative
AI.
To begin this discussion, let us recap the two dichotomies nestling at the heart of this piece. The
first dichotomy investigated how writing can record the writers already known thoughts, or ex-
tend them into creative territory. Intimately connected to this first dichotomy, the second dichot-
omy examined the place of convergent and divergent thought patterns in producing effective ac-
ademic writing. During this foggy dawn of the era of AI-generated and supported writing, ALL
Advisers need to balance several, sometimes conflicting, responsibilities regarding these inter-
twined areas.
In sum, at the time of writing (August 2023), in the pre-drafting stages of writing, AI (exemplified
by ChatGPT) seems able to provide potentially useful input in relation to:
T99 S. Thurlow

1. Generating ideas for a written piece on a reasonably clear-cut topic

2. Outlining these ideas in a (usually) logical sequence
3. Finding references to support ideas (although, as stated previously, some of these may
prove to be spurious, overly general or indeed, fake)
4. Providing a paraphrase or summary of difficult reading material to aid understanding
5. Organising notes from reading including identifying major themes on which to base later
writing.
Rowland (2023) provides more useful detail on the use of generative AI in these preliminary
phases. I highlight them as apart from #3 they tend to involve the core thinking work where
a writer s creativity would be expected to emerge. This is important for our first dichotomy and
would appear to place the student writer who accepts AI s suggestions at the recording end of the
continuum of how thinking and writing intersect. However, the question of whether this kick-
start in providing, sequencing and supporting ideas may aid later expressions of creativity and/or
originality in subsequent drafts or, rather, to continue to lock it down, remains open for debate
and indeed, forms part of an evolving and much larger discussion on AI and creativity.
To some extent, the second dichotomy extends the first, with the pull and push between conver-
gence and divergence in academic writing contexts lying at its core. This piece has explored how
a written product may result from both irrational, divergent forces and more rational, convergent
approaches. With its programmed preference for a convergent approach, AI programs such as
ChatGPT steer the user to safe information on a topic already published material that is acces-
sible to its web crawlers. As such, it does not know what it does not know and cannot suggest any
novel or innovative idea that is not already out there, which may pose limitations on its use by
those exploring new research-based worlds, such as doctoral candidates.
Both dichotomies are intimately connected to our work as ALL Advisers, for ultimately, our work
as third-space professionals in the neo-liberal university (Whitchurch, 2008) positions us as both
convergent enforcers of the disciplinary and institutional systems that employ us and as dissenters
encouraging divergence and other types of freeing up in thinking and writing. This can be a
difficult balancing act to get right.
Let us explore our role as enforcers first. This function sees us as upholders of the often obscure
rules of academic writing to those learning the craft. We routinely find ourselves in this space
of surveillance as we review student writing, frequently in the frantic days (or even hours) before
a high-stakes assignment is due. Working with students in this fraught environment, we help to
ensure that their work meets the specific requirements demanded by their lecturers and disci-
plines. Although I do not frame this service as necessarily constraining thinking and the outcomes
of thinking (such as written creativity and/or an original product), often the timing of this eleventh
hour checking work serves to firmly place it and us at the recording or locking down end
of how academic writing occurs.
The other potential role of the ALL Adviser is one of liberator, helping to open writers up to
different ideas, approaches and techniques in their writing and in some scenarios freeing them
in their academic work. This dimension of our role coalesces in the one-on-one writing tutorial,
particularly when a student brings in an early draft of their work. In these encounters, students
often present the with a task rubric which requires them to corral their bright but unwieldy ideas
into an ill-fitting straitjacket template or genre. In direct contrast to these demands for uniformity
are criteria for assessment which commonly prioritise the writer s own ideas/content at the ex-
pense of the marks allocated to written expression and other mechanical criteria such as refer-
encing or formatting. To me, this suggests that writers may be more creative or freer with their
ideas if they are able to fit them within the parameters of the task at hand. Therefore, when advis-
ing these writers, I prioritise strategies to give them the self-confidence to discover and emphasise
their own ideas and often leave detailed discussion of how to write it up for a follow-up session.
T100 Writing, thinking and creativity in an age of Artificial Intelligence

The key to success in this type of writing tutorial is often the discovery and amplification of
writer s voice. While writer s voice remains a contested term in the literature of writing pedagogy
(see, for example, Tardy, 2012), the task of persuading the reader to believe a line of reasoning
and then guide them to a convincing conclusion is almost impossible without one. Due to its
significance, I spend much time in individual tutorials helping students find and grow their own
voice while maintaining the courage to use it throughout their work. I persevere with this work
because I see that tapping into this inherently divergent voice may enable the writer to produce a
novel, creative and/or original piece of work that satisfies both their creative urge and helps them
receive a pleasing grade.
Writer s voice has the potential to become a hot topic again as AI infiltrates the spaces where
writing is born and shaped. At the time of writing (August 2023), programs such as ChatGPT can
present innumerable ideas on a mind-boggling range of subjects. Having trawled the web and
amalgamated all that is known about a topic, these ideas are then regurgitated through an author-
itative bot voice stating how the user should approach their task and urging them to follow the
formulaic and determinedly rhetorical moves it lays out. The way it outlines this information may
indeed be one way to approach a topic, but there are surely others. I suspect many students, how-
ever, will find the lure of information presented in such a solid and seemingly legitimate order by
this convergent AI voice extremely hard to resist.
Focusing on the writer s voice/s adopted by AI programs such as ChatGPT soon conjures up other
issues. A major reservation for me in relation to the use of AI tools to support academic writing
is that the strength of claim advanced by this self-assured voice can be rather strong and, at times,
inappropriate for certain disciplines. As ALL Advisers, we know that many academic tasks need
to be written in a formal, objective writer s voice, but this is not a blanket rule. The role and value
of both the subjective and objective writer s voice shifts dramatically across disciplines and dis-
ciplinary discourses. Crucially, an overly objective voice may be off-putting to readers in some
disciplines due to its contrived and inscrutable nature. Indeed, from a reader s perspective, the
most memorable and often the most highly cited works may exhibit a warm, guiding and even
subjective voice, not the fro en tundra of formal objectivity. However, achieving a suitably ten-
tative, disciplinary-appropriate yet accessible voice is something that, currently at least, seems to
largely escape ChatGPT. Without some active prompting to tone down its over-confident stance,
students who need to write using the tentative approaches preferred by many Humanities and
Social Sciences-based disciplinary traditions may find this assertive stance disconcerting. As
such, this an area where the nuanced and developmentally focused advice given by human ALL
Advisers to writers on when to withhold and when to reveal their voice appears to have a decided
advantage over AI.
The definitive and take no prisoners style in which AI often delivers information presents other
challenges to our ALL practice. Taking the example of ChatGPT once again, I can confirm its
determination to avoid saying, no, I don t know the answer to your question , at all costs.
Whether this alarming inability to admit ignorance is the result of programming flaws or reflects
a programmed wish to maintain its knowledgeable face and satisfy users at all costs, many AI
programs have become well-known for creating fake information based on non-existent sources.
With this stance, AI can be an embarrassingly poor model for aspiring academic writers to follow.
As ALL Advisers, we can only hope this aspect of AI is thoroughly ironed out in the coming
months/years, but in the meantime, we need to be on high alert to these issues.

7. Conclusion: Learning writing by breaking the rules of writing

Each time a writer begins their work, there are two writers inside of them. One is the writer who
has something to say; the creator who seeks to make the page sing with their novel ideas. The
other is the constrained writer, held back by their topic, genre, self-confidence and a host of other
intrinsic and extrinsic restraints. ALL Advisers see these two writers every time a student comes
T101 S. Thurlow

to us seeking help with their work. And while we may not see which writer ultimately prevails in
their submitted product, we are all witnesses to the struggle that ensues as these duelling personas
battle it out either on the main stage of the page/screen or, more often, in the wings.
AI will undoubtedly change the outcome of the adversarial scenario depicted above but in what
ways is still unclear. As we have seen, AI programs such as ChatGPT offer us an inexhaustible
supply of ideas for writing and willingness to produce models of work for no or minimal cost to
users. In fact, one of AI s core strengths is its ability to generate multiple ideas for a piece of in-
progress writing, outlining these ideas in a seemingly logical order. These tempting offerings
sourced from every nook and cranny of the web present the student writer with an impressive
variety of ways to approach their topic and the promise of capturing a stellar mark for their work.
But while things may look rosy on the surface, delving a little deeper sees significant cracks creep
into this always ready, always eager to provide an answer hero narrative of AI. The uncertain
provenance and frequently formulaic nature of the ideas that it conjures forth are often com-
pounded by a lack of clarity regarding the reasons and rationale behind its choices in composing
its synthesised texts. Clearly, this approach is incompatible with what is expected in the Academy
where any (human) academic writer needs to be transparent regarding the what, where (from) and
why (chosen) on which their work is based. Thus, while AI can be a useful tool for the generation
of ideas, there are clear limits to its ability to mimic and/or extend human creativity.
Finally, as ALL advisers and writing specialists, we know that the most effective and memorable
writers are skilled at bending or even breaking the rules of writing. As Sharples (1998) astutely
observes, like most rules of everyday living, they are most useful when learned and then selec-
tively ignored (p. 5). This paradox characterises any act of writing and is keenly felt in academic
writing contexts where student writers are occasionally rewarded with high grades for bending
these rules in a disciplinary-acceptable manner. Yet this very human capacity for both coming up
with novel ideas for writing and exploring them using an engaging and inspiring writer s voice is
not something that AI programs such as ChatGPT appear to be expert at doing not yet anyway.
As ALL Advisers, I believe that we have all learnt to exist albeit a little uncomfortably at times
with the paradoxes presented by writing, thinking and creativity that have been surveyed in this
paper. AI, as the new kid in town , has yet to learn the nuances that we have accrued in our
everyday practice. But, as the last year has aptly demonstrated, it is a quick learner.

Acknowledgements
I thank Stephen Campitelli and Dr Justin Park for their feedback on earlier versions of this paper
and the JALL reviewers for their suggestions on the final draft.
The author used ChatGPT in preparing this paper with the aim of providing an authentic example
of use from a commonly used LLM. He has reviewed all content and wordings created by the
generative AI tool used, edited this content as needed, and takes full responsibility for the content
of the subsequent publication.

References
Cropley, D. (2023). Is artificial intelligence more creative than humans?: ChatGPT and the di-
vergent association task. Learning Letters, 2(8). [Link]
Dunleavy, P. (2003). Authoring a PhD: How to plan, draft, write and finish a doctoral thesis or
dissertation. Palgrave Macmillan.
French, S., & Kennedy, G. (2017). Reassessing the value of university lectures. Teaching in
Higher Education. 22(6), 639-654. [Link]
Greetham, B. (2008). How to write better essays (2nd ed.). Palgrave Macmillan.
 T102 Writing, thinking and creativity in an age of Artificial Intelligence

Hayot, E. (2014). The elements of academic style: Writing for the humanities. Columbia Univer-
sity Press.
Li, Y., Sha, L., Yan, L., Lin, J., Rakovi , M., Galbraith, K., Lyons, K., Ga evi , D., & Chen, G.
(2023). Can large language models write reflectively. Computers and Education: Artifi-
cial Intelligence, 4,100140. [Link]
Locke, T. (2015). Developing writing teachers: Practical ways for teacher-writers to transform
their writing practice. Routledge.
Ong, W. (2015). Writing is a technology that restructures thought. In P. Elbow (Ed.), Landmark
essays on speech and writing (pp. 83-96). Routledge.
OpenAI. (2023). ChatGPT (August 30 Version) [Large language model].
[Link]
Pennington, M. (2014). Towards a creative writing pedagogy. In H. Millan & M. Pennington
(Eds.), Creativity and writing pedagogy (pp. 17-24). Equinox.
Razumnikova, O. (2013). Divergent versus convergent thinking. In Encyclopaedia of creativity,
invention, innovation and entrepreneurship (pp. 546-552). Springer.
Richardson, L. (1998). Writing: A method of inquiry. In N. Denzin & Y. Lincoln (Eds.), Hand-
book of qualitative research (pp. 345-371). Sage.
Rowland, D. (2023). Two frameworks to guide discussions around levels of acceptable use of
generative AI in student academic research and writing. Journal of Academic Language
and Learning, 17(1), T31-T69. Retrieved from [Link]
[Link]/jall/article/view/915
Sharples, M. (1998). How we write: Writing as creative design. Routledge.
Strongman, L. (2013). Academic writing. Cambridge Scholars Publishing.
Tardy, . (2012). Current conceptions of voice. In K. Hyland & C. S. Guinda (Eds.). Stance and
voice in academic discourse (pp. 34-48). Palgrave Macmillan.
Thurlow, S. (2021). Creativity is for poets and pop singers isn t it? Academic perspectives on
creativity in doctoral writing. Arts and Humanities in Higher Education, 20(2), 187-
206. [Link]
Whitchurch, C. (2008). Shifting identities and blurring boundaries: the emergence of third
space professionals in U. K. higher education. Higher Education Quarterly,
62(4), 377-396. [Link]
Yagelski, R. (2011). Writing as a way of being: Writing instruction, nonduality, and the crisis of
sustainability. Hampton Press.
Zemits, B. I. (2017). Representing knowledge: Assessment of creativity in humanities. Arts and
Humanities in Higher Education. 16(2), 173-187.
[Link]

View publication stats