1 of 14

Researchers Writing With(out) AI

MyFest25

Shyam Sharma, Professor of Writing & Rhetoric

State University of New York at Stony Brook

Greetings…

In this session, I’d like to facilitate an exploration of how AI is reshaping the functions of writing in academic research and scholarship. After a brief discussion of how AI-based writing is impacting research/scholarship in different disciplines, so we can all apply the discussion to our respective disciplinary challenges and opportunities, we’ll consider if and how we may need to redefine writing in the age of AI and how we could create our own frameworks and boundaries for using and not using AI in the writing process.

Description submitted–shorter–To be used in public-facing materials like the schedule or website.

In this interactive workshop, we will explore how academic writing works as a vehicle for thinking and knowledge-making in the process of research, exploring what is at stake when researchers from across the disciplines delegate the "writing" part of the research process to AI tools. With hands-on activities, we will assess how using AI "to write" might enhance or undermine academic rigor, originality, responsibility. We will create personalized rubrics and flowcharts to guide ethical uses of AI in field-specific ways for ourselves. Scholars of any discipline who want to approach writing with (or without) AI more thoughtfully are invited.

Longer: Include the purpose, activities, and what participants can expect. This will be included in the schedule and website.

In this interactive workshop, we will explore how academic writing functions not just as a means of communication but as a central tool for thinking, inquiry, and knowledge-making throughout the research process in any discipline. As generative AI tools become more capable and more common, researchers across disciplines often struggle to decide when and how to let AI tool write for them, assist their writing, or avoid AI tools when writing. While such tools may support certain aspects of the research process, they also pose risks: to originality, depth of thought, ethical responsibility, and the researcher’s intellectual agency. These risks are particularly pronounced in the humanities and social sciences, where writing more substantially embodies the research process. Toward the sciences end of the spectrum, AI poses risks because writing tends to be defined more narrowly, letting AI encroach thinking and meaning making.

Participants will do two activities. First, adapting a writing with and without AI rubric, they will identify what roles writing plays in their research and determine which of those roles are best preserved for human judgment and insight. The second activity will help them build a personalized decision-making flowchart to guide field-sensitive and ethically grounded choices about when, why, and how to use (or avoid) AI tools in their writing during research (broadly defined). Scholars from all disciplines who want to approach writing with (or without) AI more thoughtfully are warmly welcome.

2 of 14

Making Knowledge Through Writing

Humanities: we use discourse as a dominant tool of knowledge making – we dialog, synthesize, and interpret (theorize from) published texts
Sciences: we gather new/hard data from labs, patients, field, community – there’s a “control” mechanism that’s more transparent, systematic, etc
Social sciences: we combine both methods – with relative control

AI poses more risks to Humanities (and qualitative research) whenever researchers delegate synthesis to AI tools (and not just processing/analysis)

3 of 14

While researchers in the sciences may approach “writing” more as a vehicle for new ideas based on real-world data than as a source of ideas (if the researchers are honest), researchers in the humanities risk losing social value of the knowledge they create and contribute if they don’t use AI very cautiously.

4 of 14

Impacts of AI in Research via “Writing”

Still “writing” is a lot more than what scholars beyond the humanities think – that increases risks of AI use across the board
Delegating “writing” delegates intellectual work, professional growth, ethical/social responsibilities, etc
Even delegating “summarizing” of information to AI poses risks
So, the risks are distinct but serious across the board

What do we do?

Rethink, redefine, relearn “writing” in the age of AI
Create and adapt tools to exercise agency and maintain ethics

5 of 14

Activity 1: Rethinking “Writing”

Look at the “writing with(out) AI” handout and take notes.

What “writing” tasks could you do “better” without AI?

For more originality, agency, depth, etc, on your part as a scholar
For slowing down, thinking through/grappling, having a better grasp
For being ethically more responsible, sensitive, accountable, etc

What tasks involve minimal or no risk when you write with AI?

Where do you think you’d need to rewrite the meaning of “writing”?

Discussion

6 of 14

“Dumber” tasks you may get done with AI tools

generate Tuesday and Thursday dates in a table column from Jan to May
separate student names from this messy table I copied from an LMS
extract emails from this table and separate them with commas
point out paragraphs in this manuscript that I can cut or condense
summarize this article to help me pay attention to the theme of X as I read it
suggest possible titles for a journal article based on my abstract

Dr. Shyam Sharma State University of New York at Stony Brook

7 of 14

More cognitively significant tasks

interact to remind/brainstorm your own ideas – not learn from the AI tool
interact by asking how, why, etc, as you flesh out/refine my own ideas/perspectives
summarize an article to help you pay attention to theme X before you read it closely

draft a conference paper based on your notes and multiple rounds of interaction
analyze a set of articles, drawing your attention to gaps or patterns for you
analyze the data you uploaded, under 5 key themes for discussion

develop a intellectual position through extended interaction with an AI tool
write up a “lit review” section based on 20 articles you uploaded
write up an article’s “discussion” section based on study results provided

write up the social impact section for a grant proposal
substantively revise a journal article based on your notes
write profiles for effective outpatient dept. care of (say) Moroccan geriatric patients

Dr. Shyam Sharma State University of New York at Stony Brook

8 of 14

Prevalent myths about “writing”

Linearity myth: writing is a single, simple, linear process
Transience myth: you can “write” after you’ve learned how to once
Transparency myth: you can “write it up” once you have data/ideas
Hierarchy myth: you can outsource “lower-order” writing tasks
Content myth: content is separate from the “form” of writing
Style myth: good writing is mostly about “brilliant” writing styles
Genius myth: good writers are born “geniuses,” not educated

Dr. Shyam Sharma State University of New York at Stony Brook

As a professor of writing and rhetoric, I come across a lot of bad ideas about writing, and now those ideas are getting on steroid due to the influence of AI regarding what it means to write. Here are just a few:

Linearity myth has it that writing is a single, simple, linear process – in reality, writing very recursive, multi-layered, and very complex.
Transience myth tells us that we can “write” after we’ve learned how to write once and for all – in reality, we continue to learn increasingly advanced skills throughout our academic and professional lives.
Transparency myth means that we can “write it up” once we have data/ideas – in reality, writing begins with note-taking and facilitates all stages of the research and publication processes.
Hierarchy myth claims that we can outsource “lower-order” writing tasks – but serious meaning-making work happens even while editing.
Content myth is about content as separate from the “form” of writing – in reality, you can’t separate meaning from text any more than you cannot separate the dance from the dancer.
Style myth means good writing is mostly about “brilliant” writing styles – but writing is also a science, which is why I have a master’s and doctoral degrees in the theories, methods, pedagogies, and programs of academic and professional writing.
Genius myth tells people that good writers are born “geniuses,” not educated – in reality, there’s no such thing as good writers without good writing.

These are some bad ideas the public and many even in the university believed; by the way, my dissertation was a study of how engineering scholars view language and writing and how that view could shape the way they teach and mentor the next generation of engineers as knowledge makers. I learned that engineers tell you a lot of bad ideas about writing but they are in practice, in their skills, they are quite sophisticated about with writing.

One common bad idea is that writing is generally a single, simple, linear process: if you think about scholarly writing, for instance, first you read, then you annotate relevant current literature, then you gather and analyze your data, then you outline, then you draft…. etc. But as Charlex Bazerman, one of the most respected scholars in my discipline of Writing and Rhetoric, shows, writing is is an unpredictable, iterative, multi-layered, multi-process process. There are many, overlapping, or entangled layers or processes that come together in different combinations depending on the nature of the writing you’re doing. Some of those processes include the physical process of sitting down and getting ideas on paper or screen as words and sentences (that’s writing) and a variety of pre-drafting processes or tasks such as note taking during preliminary research or literature review (where an intellectual process starts taking over), recording during study design and implementation or collecting of data (which may involve a social process of engaging patients and colleagues), annotations of sources or data while finding patterns or analyzing data (which may involve the use of technological assistance), outlining the draft, and so on – and even after we have a rough draft or sketch, we still have many other processes to go through, such as revising, editing, formatting, and submitting, then peer feedback and another round of revision and so on. Bazerman suggests that one, “writing” is a much broader set of activities than putting pen to paper or fingers to keyboard, and, two, even the narrowly defined act of putting pen to paper is distributed across physical, intellectual, social, and emotional domains and tasks across an often lengthy and always multilayered set of processes.

Two, there is the myth of transience or the false idea that once you learn to write you have learned it for good. It missed the point that a third grader masters third-grade level writing, a college student needs to learn a whole new set of skills after being a great writer in high school, and a new assistant professor or company manager needs to keep learning new writing skills as they advance in their careers. This myth leads to claims like: I am a good writer, or I am a terrible writer. In my dissertation, I began by extensively reviewing misconceptions like this one about writing especially in the STEM fields. Building among other scholars on the historian of academic and professional writing, David Russell, I showed how these myths affect scientists.

Three, there is the myth of transparency, or the idea that writing (or at least “good writing”) objectively translates pre-existing knowledge. In reality, writing embodies thinking and the process facilitates the birth and growth of ideas, as well as facilitating a lot of other tasks, from the physical to the social to the intellectual in the process.

Four, there is the myth of hierarchy, even among writing teachers, or the idea that because things like editing are less intellectually engaging than drafting, they can be outsourced to someone else. In reality, editing can make a vast difference in what your text conveys.

Five, there is the myth about content, especially the idea that content is separate from the form of writing, an extension of the idea that ideas and language are separate. In reality, it is only when you’ve given content the form of writing that you’ve realized the idea in a communicable form.

Six, there is the myth of style, or the idea that effective writing is characterized by brilliant or appealing style–that you need to be a humanities person to write effectively.

Finally, and these are only the most common ones, there is the myth of genius, or the idea that there are born writers, people who have a special kind of talent or gift. In reality, writing is a teachable skill and it is learned and improved lifelong.

Now, briefly, why do these myths emerge and persist? Russell has theorized and shown that writing was not everyone’s daily business until recently (in fact, reading wasn’t either until the Renaissance in Western Europe and much later in other places). Once it became part of everyday life and every profession, it dramatically evolved from simpler tasks like describing and recording manual tasks and processes to much more complex tasks like gathering vast amounts of data, generating and presenting diverse perspectives, and disseminating advanced new knowledge. While people across the professions kept believing that writing is recording, the professions themselves demanding and rewarding far more complex types of writing. And people, including professionals who practice the complex tasks of writing, kept defining writing as mere recording of thoughts. The gaps between reality and discourse about writing are so persistent that often university leaders (like one particularly crazy dean about a decade ago) tell writing professors that “anyone who can write can teach writing” – which is like anyone who can read nursing books can treat patients.

- - - - -

By the way, I tried to use ChatGPT to help me remind the myths about writing because I couldn’t recall some of them off the top of my head, and I was too lazy (at 2am after a long teaching day a few weeks ago) to look up my dissertation or the book by Russell. Unfortunately, GPT included a few non-existing “myths” that I had never read in the literature (it most likely made them up and made them sound like they’re actually from current scholarship because they seem equally convincing). Anyway, it did help me to remember the style myth and the genius myth, and I’m acknowledging that now.

9 of 14

How AI is putting old myths about writing on steroids

Linearity: just use AI it will help you skip through the tough step
Transience: you don’t have to learn it anymore, as AI can “write”
Transparency: AI is “better” at writing it up than humans are
Hierarchy myth: it’s okay to “offload” lower-order tasks to AI
Content myth: AI knows it all, because it has infinite datasets
Style myth: AI can write in any style you prompt it to write in
Genius myth: anyone with AI and prompting skills is a genius writer

Dr. Shyam Sharma State University of New York at Stony Brook

With the advent of AI, the already insidious myths or misconceptions about writing have been dramatically aggravating and spreading like wildfire. I am one of the five editors of a book titled “Bad Ideas about Writing and AI,” and there are so many bad ideas that we received more than 120 chapter submissions and selected more than 50 unique bad ideas for inclusion. The book is coming out next year.

Here is how the writing myths of the previous slide are being put on steroid by AI.

Many people see writing as a linear task to shortcut with AI, but in reality, AI disrupts thinking, diminishes agency, and obscures engagement at every step. The old myth that writing is learned once—and now, not needed at all—has deepened, as some liken AI to calculators, ignoring the intellectual labor writing demands. Others believe AI can express their ideas better, mistaking grammatical polish for depth or clarity. As AI becomes the new grammar cop, surface-level correctness is mistaken for effective writing, sidelining logic, ethics, and responsible use of sources. The false split between content and form has also grown, with users trusting AI to “generate content,” despite its reliance on plausible-sounding patterns and limited data. Even the belief that style alone makes writing good has warped further, as AI mimics tone without real rhetorical awareness. Finally, while AI might boost confidence, overreliance turns it into a minefield of ethical, professional, and relational risks.

10 of 14

Dr. Shyam Sharma State University of New York at Stony Brook

11 of 14

Activity 2: Creating Your Own Flowchart

Take a look at my general flowchart. Thinking of the context of research in your field and for a particular project, check off questions you would ask, leaving out what you wouldn’t and adding what you’d need to add. Create your own flowchart (by hand).

Prepare to share your flowchart with a small group.

Breakout room discussion – followed by a brief discussion.

12 of 14

Maha Bali: My own past use of AI tools

Task/Tool	ChatGPT [paid]	Copilot	Poe (paid)	Gemini	Claude	Keenious	Typeset/ Scispace	Deep Research
Brainstorming	Yes	Yes	Yes	Yes	Yes			Yes
Finding articles	Unreliable	Semi reliable*		Semi reliable*		Yes. OK.		Yes
Summarizing/ paraphrasing	Yes	Yes	Yes		Yes		Yes	A bit
Writing abstract	(Via poe)		Yes
Data analysis	(Via poe)		Yes		Yes
Translation	Yes						Yes

* Semi reliable means it provides reasonable summaries and real sources, but the sources may not say what is in the summaries

I also encourage you to create your own toolset, focusing on what you want to learn to do as a researcher and scholar with and without AI – and I say with and without AI not only because AI tools are not yet reliable but also because you’ll need to do most of the scholarly things while walking, talking to someone, presenting, or debating – you can’t just run to your room and ask the correct AI tool to do it for you.

Presented at a doctoral student program in Morocco by Prof. Maha Bali, this is a personal AI-assessment rubric that shows a very pragmatic approach: test tools until you know what works best for what contexts and purposes – beyond simply relying on one or two tools to try to do everything.

- - - -

I also want to share a serious caveat that Maha shard along with this slide: In “Situating Search,” Bender and Shah argue that “removing or reducing interactions in an effort to retrieve presumably more relevant information can be detrimental to many fundamental aspects of search, including information verification, information literacy, and serendipity.”

Here’s Maha’s slide.

Open Praxis: best guidelines

13 of 14

Thank you!

14 of 14

	Genuine Scholarship…	AI-Generated Content…
Nature & Purpose	Is driven by curiosity, passion, and the purpose of advancing disciplinary understanding and/or addressing societal needs. It is a means to higher ends. Test it: Write up a research note for your next article. Then ask ChatGPT to do the same, providing it some context.	Is plausible word patterns limited by underlying data, app design, market forces; also limited by user knowledge, skills, ethics. Ask this: Does AI just wow me with “smart” ideas or reflects my field’s & society’s needs? Am I contributing new knowledge to my discipline or solution for society—or am I just producing something that sounds scholarly but fools my readers & society?
Evidence & Method	Reports findings from new data, using rigorous methods, and transparent documentation of process; research is designed upon scholarly context/ tradition; theory/perspective is applied or derived methodically. Test it: Read the lit review and methodology of an article you wrote. Then use Elicit to write that lit review and ask Claude to draft that methodology (with some prompts).	AI generated content has no access to real world data; it is not based on transparent method (AI is “black box” by definition). Ask this: What role did disciplinary conventions of data gathering and analysis play in my writing compared to those of the AI tools? Can I use AI generated content as “findings”? What databases does this AI tool have access to? How does AI define reality/ontology, knowledge/epistemology, and methodology? How does AI affect my rigor, transparency?
Novelty of thought & Interpretation	Involves original synthesis and advances new ideas, through cultural and value-based reasoning, and intuitive/human insight; process rewires brain. Test it: Feed some data from your research to Gemini, asking it to interpret and discuss it. Compare with your writing.	Mimics logical patterns; can short circuit user reasoning, nuance, agency; can’t do human-like “abductive” reasoning. Ask this: What conceptual nuance, contextual meaning, or cultural grounding of my writing does AI miss? Is my AI use making me mechanical, shallow, unoriginal, and less and less capable of “hard” thinking?
Intellectual Integrity & Social Responsibility	Demands citation, transparency; is accountable to peers and publics; shows contextual sensitivity, cultural awareness. Test it: Ask DeepSeek to write a citation-involved paragraph.	No credit to source; no ethical stance or accountability for error or harm; no regard for privacy, intellectual property; no concern for society, profession. Ask this: Are AI’s citations real, reliable, relevant? Who do I credit for the intellectual property? Sources? AI? Myself? Who is responsible for any adverse impacts? Can I fully reveal to my colleagues, my children, and other stakeholders how I used AI in this work?
Public Impact & Scholar Identity	Aims to advance knowledge, empower people, or improve society. Researcher/scholar builds a scholarly voice, identity, respect. Test it: Summarize a recent publication of yours for a community group. Then ask CoPilot to do it. Which version sounds like you?	Detached from community, consequence, or purpose. Produces content, not contribution–short circuits learning. Ask this: Does this content express my scholarly voice and identity in society? Does it help me grow as a scholar? How would publishing AI content affect my recognition in my discipline and society?

Dr. Shyam Sharma State University of New York at Stony Brook