Guilty secrets

In my post of 15 March 2014 — The online Scholar: A guide for PhD students (#187) — I suggested a list of points for a researcher to consider as they integrate online media into their writing processes. I even ran a few workshops on the theme for our University’s Institute for Academic Development (IAD).

But, we no longer need to remind new researchers to pay attention to email, online libraries, online survey tools, image libraries, video conferencing, virtual learning environments, smartphones, social media, and blogging in a research context. As I discovered, the audiences for treating online media as something special and unusual diminished rapidly after about 2012.

Now, navigating the uses and challenges of AI, in particular LLM-based text and data processing, takes over as a topic for which new scholarly practices are emerging. I was reminded of this as I responded to feedback on a recent article “Rethinking AI and Design in Light of the Hermeneutical Legacy of Adrian Snodgrass” that will appear this year in Architectural Theory Review. The publisher asked me to declare in my article the extent to which I used AI in the writing.

AI assistance to improve language and formatting, generate ideas, write code, and search for literature required no special acknowledgement. However, the publishers disallowed using AI to compose or generate text or images — without special explanation. I found myself adding a statement at the end of the article that the first draft of the abstract was generated with AI assistance.

Most researchers are aware of the consummate capabilities of AI to generate highly plausible and contextually relevant text. However, it is easy for discussions about plagiarism, copyright, academic integrity and (frankly) “what can I get away with?” to overwhelm a consideration of AI’s possibilities in contributing to quality research.

Guilt-free AI

In that vein, I have asked ChatGPT to adapt my post of March 2014 to accommodate newly emerging practices in the uses of online media in research writing — practices informed by AI.

I’ve modified ChatGPT’s update here. Most of the AI’s points under the headings were followed by a note encouraging readers to discuss protocols with supervisors, maintain the integrity of their own intellectual development, develop judgement, read carefully, avoid the generic, take care to develop depth and specificity, consider that any open-access work they generate may be scraped by AI, etc. The AI suggested, “discuss AI openly rather than treat it as a guilty secret.”

Revisiting online scholarship

What follows is an updated version of that earlier list, retaining the life-cycle structure of PhD research, but recasting “online scholarship” in the light of AI-assisted research and writing. I’ll indent this discussion as a quotation, though I modified it substantially.

1. I’m thinking of applying to enrol in a PhD

It is still useful to make visible your researching and writing abilities. AI tools can help refine a CV, improve the clarity of a writing sample, or test whether your research interests are intelligible to a wider audience. They can also help you identify gaps in your preparation.

2. I’m applying for a research degree and/or scholarship

Applications remain competitive, and selectors still look for evidence rather than aspiration alone. AI can assist here in practical ways: helping you organise your achievements, summarise prior work, improve the structure of a research statement, and tailor a proposal to the requirements of a funding body. It can also act as a useful critical reader: “What is unclear in this proposal?” “What assumptions am I making?” “What would a sceptical reviewer question?”

3. I’m preparing a research proposal

Most universities still provide guidance online about what a research proposal should contain: topic, research questions, context, methods, contribution, timetable, and bibliography. AI tools can be helpful at this stage, especially in exposing vague formulations. They can suggest alternative ways of phrasing a research question, identify possible methodological tensions, or help you distinguish between a topic, a problem, and an argument.

4. When I’m on programme and have firmed up my research topic

It may still be useful to put some account of your research online: a short profile, a project summary, a page on your university website, a professional network entry, or a carefully maintained blog. The purpose is not self-advertisement alone. It helps make your work discoverable, allows others to understand your interests, and can contribute to the formation of a scholarly network. AI impinges on this visibility. Public descriptions of your research may be scraped, summarised, indexed, misread, or recombined by automated systems.

5. During my research

Researchers now work among a dense ecology of online and AI-supported systems: library databases, citation managers, transcription services, survey platforms, image archives, mapping tools, collaborative documents, coding environments, translation tools, and generative AI interfaces. LLMs can assist with summarising material, comparing arguments, generating code, enlivening spreadsheets, cleaning data, drafting interview questions, translating passages, and exploring alternative formulations of a problem.

[I add that it’s worth being aware of the possibility that supervisors, examiners, critics and audiences may themselves read some of your work filtered through smart search engines, text-to-speech synthesis, and AI-generated summaries.]

6. When I’m developing articles for peer review and publication

Journal submission systems, open access arrangements, preprint repositories, and online peer review processes are now routine. As indicated above, publishers increasingly have explicit policies on AI use. These policies vary. Some allow AI for language editing, formatting, translation, coding assistance, or searching the literature. Some prohibit AI-generated text or images unless clearly declared. Most do not allow AI tools to be listed as authors, since an AI cannot share responsibility for the work.

7. After I’ve published an article or presented at a conference

A short online commentary or lay summary remains a useful way to extend the reach of academic work. AI can help generate versions for different audiences: a short abstract, a blog post, a script for a podcast, a set of social media posts, or a plain-language summary. It can also help turn a conference paper into a more accessible account without losing the argument. See 5 above about acknowledging where you have used AI.

8. While developing my network

Academic networking still involves conferences, workshops, seminars, reading groups, public lectures, edited collections, exhibitions, and collaborative projects. Online media continue to assist with publicity, event organisation, registration, recording, and dissemination. AI now adds further possibilities: drafting calls for papers, producing programme summaries, generating accessible descriptions, preparing captions, translating announcements, or identifying adjacent fields and potential contributors.

9. When I am teaching, demonstrating, or tutoring

Putting teaching notes online is now only one part of a larger question. Students will almost certainly use AI tools in some aspect of their learning: to summarise readings, explain difficult concepts, draft essays, write code, generate images, or rehearse arguments. The issue is not simply whether to allow or prohibit these uses, but how to help students to use such tools critically.

10. When I’m writing drafts and submitting to my supervisory team

In 2014 the question was whether to place drafts, notes, or summaries online. Now the question (for discussion with supervisors) also includes whether and how AI is involved in the drafting process. Has AI been used to improve grammar? To restructure a chapter? To produce a literature review? To generate paragraphs? To summarise sources? To translate material? To analyse data?

11. When I’m mentoring other students

Mentoring now includes sharing experience of AI as part of the changing infrastructure of scholarship. This might involve practical advice: which tools are useful for note-taking, transcription, coding, literature management, translation, or drafting. Do not assume that fluent prose is good scholarship.

12. As I’m finalising my thesis

If you have used AI tools in any substantial way, check your university’s current regulations. Some institutions require a declaration. Some distinguish between language correction and content generation. Some require that candidates retain records of prompts, outputs, or uses. Policies are still evolving.

13. After I’ve passed my PhD and submitted my thesis to the library

The decision about making a thesis publicly available online remains important. Some candidates delay release because they plan to publish a monograph or articles from the thesis. Others favour immediate open access. AI adds another consideration: publicly available theses may be indexed, summarised, mined, and incorporated into larger information systems. That is not necessarily a problem; indeed, discoverability is often desirable. But candidates should understand that open access now means machine readability as well as human readership.

14. When I’m looking for a job

An academic job search still requires an up-to-date CV, publication list, online profile, teaching record, research statement, and evidence of impact or public engagement. AI can help adapt these materials for different posts, shorten or clarify statements, prepare interview notes, and rehearse possible questions. It can also help compare job descriptions with your experience, identifying where you need to provide stronger evidence.

15. When I’m requesting references

It is still helpful to make the referee’s task easy. Provide a current CV, a short statement of the post or scholarship, the deadline, the criteria, and a reminder of relevant achievements. AI can help you prepare a concise briefing document or extract relevant points from a longer CV. It can also help draft the initial request email.

16. During my career after the PhD

The final point from 2014 still stands: keep up with the changing and challenging world of online scholarship. But the challenge has shifted. It is no longer enough to maintain a blog, an online profile, a repository presence, or a professional network. Scholars now need to appreciate how AI systems mediate searching, reading, writing, citation, translation, image production, data analysis, teaching, peer review, and public communication.

Having said all that, a researcher may indeed reject AI for writing on various grounds, including the way machine learning mines existing repositories of texts, and various putative misuses. The critique of AI is a research field in its own right.

Note

• Featured image is of a second-hand bookshop in Alnwick, Scotland, modified by ChatGPT to show the presence of students. Prompt: ” Please generate a variant of this picture of a famous second hand book store to look more academic, i.e. with students/researchers. Delete the shelf mark “GARDENING.” Here is the original: