Drafting from scratch
While even the most advanced LLMs are not good enough yet to generate quality academic writing from scratch, as a form of ideation to create sub-par drafts to help overcome writer’s block it’s certainly very useful. It may be more valuable for those with English as a second language, but even then that would be more useful to help with re-writing, rather than writing from scratch. Academic writing is somewhat unique not only in its form but the fact it requires rigorous citations from literature and empirical sources as well as complex abstract reasoning. More importantly, academic writing reflects substantial thought and work as part of the entire research process. And while AI cannot be considered an author because it’s not a human agent and cannot be held accountable, using AI generated text may be seen as passing off others’ words (even though AI isn’t human which makes this idea philosophically tricky, but many publishers view this as a form of misconduct even if it doesn't perfectly fit the definition; certainly in the public sphere the mere use of words like 'delve', 'intricate' and 'meticulous' in a research paper invites ridicule on social media because these are infamous overused Chat GPT words which are uncommon in academic writing) as your own. The reality as of 2024 is that any academic would and even should be disappointed by the quality of academic writing that even the most advanced LLMs generate from an initial prompt.
Where it can be much more helpful with 'from scratch 'drafting is for short, less formal writing that is associated with academic research. For example, drafting a paper or conference proposal abstract, a personal profile summary for a webpage, a generic tick box form statement, a blog, survey or interview questions and so on. With suitably detailed and explicit prompting and, where applicable, grounded data to help inform the draft, it’s more than capable of producing something which only requires minimal editing, because it’s fundamentally a short written form of communication.
Here's an example the entire text of a PhD thesis from 2016 (consent provided by the author, using Claude 3.5 Sonnet because Chat GPT can't process an entire PhD thesis) and asking it to draft a conference abstract based on the thesis content:
Prompt:
Attached is my PhD thesis. Can you please use this to draft a conference abstract for an event on researching online communities (i.e. the primary focus of the conference abstract is on the methodology and less so on the substantive subject matter of the thesis)?
Response:
Title: Analysing Resistance Discourses in Online Conspiracy Theory Forums:
This paper discusses a methodological approach used to analyse resistance discourses within online conspiracy theory discussion forums, drawing on a PhD thesis examining how members of these forums think about and discuss strategies for resisting the perceived "New World Order" conspiracy.
The research leveraged the unique affordances of online forums to access rich, naturally-occurring discussions as they unfolded in real-time. This allowed for analysis of the collective, bottom-up processes of meaning-making and strategy development, avoiding the limitations of top-down or post-hoc approaches common in social movement research.
Methodologically, the study employed discourse analysis of selected discussion threads, guided by theoretical concepts from Alberto Melucci's work on social movements. This enabled unpacking of how forum members collectively constructed and negotiated ideas about power, morality, goals, and tactics through their online interactions.
Key methodological insights include:
1. The value of online forums for accessing "uncontaminated" communicative construction processes
2. Importance of anchoring analysis closely to forum texts as the unit of analysis rather than speculative inferences about members' offline beliefs or actions
The paper argues that online forums provide an excellent opportunity for examining real-time socially constructed meaning-making processes through conversational discourse among emergent ideational communities, while also highlighting methodological and ethical considerations for this type of research. It contributes to discussions on innovative approaches for studying online spaces as sites of discursive resistance.
Re-writing
While there is some overlap between the idea of drafting and re-writing, particularly with the example from the previous section since an abstract is a kind of re-write, this particular section focuses on actual re-writing of your own drafts (note: be sure to avoid using tools that use anything you upload for LLM training, as it would effectively be leaking your research, MS Copilot should always be the default). This is an area where it’s generally much better, because it doesn’t need to generate information as such, it’s a pure language transformation task where it's given the source data. This could be re-writing something in a different prose style (and even language), or perhaps converting dense text into an FAQ or bullet points. Here’s an example using the same PhD thesis as above within the prompt:
Prompt:
That's great, thank you. Can you now please provide content for slides with succinct bullet points relating to the thesis' methodology that can be used in a powerpoint presentation for this conference?
Result:
Slide 1: Research Objective
Analyse resistance discourses in online conspiracy theory forums
Focus: How members discuss strategies to resist the "New World Order"
Key concepts: Agency, power, morality in resistance discourse
Slide 2: Data Source
Two popular conspiracy theory forums: Above Top Secret and David Icke Forum
Selection of discussion threads focused on resisting the New World Order
Natural, real-time discussions unaffected by researcher presence
Slide 3: Theoretical Framework
Adapted from Alberto Melucci's work on social movements
Key concepts: 'Action system' and 'ideology'
Focus on collective meaning-making processes
Slide 4: Methodological Approach
Qualitative discourse analysis
Close reading and interpretation of forum texts
Analysis of how ideas are constructed, negotiated, and contested
Slide 5: Analytical Process
Initial open coding using ATLAS.ti
Identification of recurring themes and patterns
Application of Melucci's concepts to organise and interpret data
Slide 6: Ethical Considerations
Use of publicly accessible forum data
Anonymisation of forum members
Reflection on researcher's role and potential biases
Slide 7: Methodological Advantages
Access to 'uncontaminated' communicative construction processes
Ability to observe meaning-making in real-time
Rich data source for bottom-up analysis of collective ideas
Slide 8: Methodological Challenges
Defining boundaries between conspiracy theory and social movement discourses
Avoiding speculative inferences about members' offline beliefs or actions
Balancing close textual analysis with broader theoretical insights
Slide 9: Key Insights
Online forums as valuable sites for studying emergent ideational communities
Importance of context in shaping resistance discourses
Fluidity and contestation in collective meaning-making processes
Slide 10: Future Research Directions
Combining online and offline ethnographic approaches
Longitudinal studies of how online discourses evolve over time
Comparative analysis across different types of online communities
As with any generative AI outputs, it will never be perfect first time but using it for drafting or re-writing can help get over writer's block and ease the process of making a start.