How to Preserve Your Unique Author Voice After AI Writing and Editing
“Oh! I simply love reading her books. The storyline builds effortlessly and the characters truly come to life!”
“Professor Arnold has published a new textbook on handling social transitions in adolescents. I bet it’ll be as good as his last one.”
There’s something about an author’s writing style that attracts you to their work again and again. Over time, you recognize this style, making them your favorite writers. In simple terms, it is the author’s voice speaking directly to the readers! But unlike fiction, academic writing and editing usually follows a fixed style of presentation. The language is typically formal, with very little leeway to express uniqueness. Yet, you can weave storytelling into your research papers. We have previously explored this concept in detail; you may read those blogs here:
- Can Research Articles Follow a Story Arc? Freytag’s Pyramid for Effective Storytelling in Scientific Papers
- Importance of Storytelling in Research Writing-Scientific Storytelling Part 1
- 3 Best Techniques for Effective Storytelling – Scientific Storytelling Part 2
- The Secret to Narrating in IMRaD – Scientific Storytelling Part 3
In this article, we focus on humanizing AI writing! No, not a humanizer tool, but simply learning how, despite the use of AI writing assistants, your voice and unique writing style can be retained when drafting research manuscripts.
What is Author Voice in Writing?
What AI Does to Your Writing Style
- Recognize transition words that appear repetitive and replace them
- Retain nuance by adding field-specific terminology
- Vary sentence length in paragraphs to ensure ebbs and flows in writing
- Fix sentences that overclaim findings and use cautious hedging
Frequently Asked Questions on Loss of Author Voice in AI Writing
What is Author Voice in Writing?
Author voice is the style of writing adapted by individual authors over a period. It allows authors to showcase their signature through written content by using choice words, phrases, expressions, and variations in sentence length that can be defined as their own.
How do you recognize your writing style? Ask yourself these questions:
- What are my preferred words for transitioning between ideas?
- How often do I use question statements in my writing?
- Do I prefer writing in the active voice? Or do I use passive voice more?
- What is my preferred sentence length? Short and direct or long and layered?
- When it comes to field-specific jargon, do I use it directly without explanation or do I elaborate it for a general audience?
From your days at the university, you would have written assignments, essays, lab reports, and research papers. Some of these were perhaps written in the pre-AI era! Use them as references to identify your unique writing style.
What AI Does to Your Writing Style
With generative AI tools like ChatGPT, Perplexity, and Claude, authors can prompt the tools to generate content for them. Obviously, this raises the question of AI authenticity. But let’s say you do not use the AI-generated content as is in your manuscript and simply refer to it when preparing your research manuscript. Does this still impact your writing style? Yes, indeed!
Repeatedly reading the AI text alters how you express ideas because you tend to be overly familiar with AI vocabulary. Perhaps you start using words and phrases like “moreover,” “it is important to note that…,” and “furthermore” frequently in your writing without intending to. Or maybe the AI tool’s way of making strong claims without using hedging language has now wormed itself into your writing as well, causing you to minimize using qualifiers like “may,” “might,” “suggests,” “indicates,” “appears to,” “possible that,” etc. in the Results and Discussion sections that require you to report findings without overclaiming.
So, what is the solution? Should we abandon AI altogether?
Not exactly. AI tools, such as Paperpal, that are built for academic writing are indeed great for saving time and working efficiently. You can obtain an outline for structuring your paper, paraphrase your sentences, and even use the tool to brainstorm ideas as you develop your research manuscript. Essentially, you should ensure that AI authenticity is verified by human review.
How to Edit AI-sounding Text
This brings us to the best ways to paraphrase AI-generated text.
1. Recognize transition words that appear repetitive and replace them
Formal or academic writing commonly uses transitional phrases like “moreover,” “however,” “not only…but also,” “it is important to…” etc. These are not technically incorrect. Yet, their repeated excessive use makes the writing sound like AI prose.
Solution: Use a range of connectors. Some sentences can use “moreover,” while others can use “in addition,” “furthermore,” “beyond this,” or “additionally.” In certain cases, you may avoid using connectors and simply state the information directly. Avoid using “it is important to note that…” at the beginning of sentences. It is very ChatGPT-like! Instead, use “notably” or simply remove the phrase completely.
Maintain variations in your word choices so that AI paraphrasing becomes easy.
2. Retain nuance by adding field-specific terminology
The biggest concern about using AI writing tools is the loss of nuance and context, which often remains unique to a study topic depending on the research field. For example, if you prompt ChatGPT to interpret your findings to help draft your Discussion section and paste the results of your experiment within the prompt, there could be two major risks:
a) Your unpublished research findings have been shared with a generative AI tool. Could it be used to train the model? Or worse, when you eventually publish your paper, will the findings be flagged as plagiarized because an AI tool has perhaps used them as reference to output information to its users?
b) Interpretation of findings requires critical thinking, which is a human skill and not a task for machines. The tool could convolute the implications of your findings and align the interpretation with its own training datasets rather than your intended meaning.
Therefore, prevent this loss of nuance by retaining technical jargon where relevant and using field-specific terminology.
Example
AI sentence: SPSN is a concept close to RSN. However, SPSN possesses distinctiveness in that it encompasses professional and practical aspects, whereas RSN does not.
Human-edited sentence: Although SPSN is conceptually related to the RSN, it distinguishes itself by encompassing professional and practical aspects, such as nurses’ growth orientation, active involvement in teams, and interpersonal consideration, whereas the RSN emphasizes adaptability and psychological persistence in challenging situations.
3. Vary sentence length in paragraphs to ensure ebbs and flows in writing
Have you seen the latest social media posts that sound robotic?
“It is not X, but Y that is the problem. There is a genuine concern regarding Y, and the solution is simpler than you think! All you have to do is follow A, B, and C. Do this consistently to alter your life in the best way possible!”
It sounds cliched. But ignore the content and pay attention to the length of sentences. It is typically one positive idea and a negative idea spliced with a comma, followed by a surprising element, three quick fixes (why is it always three?!), problem reiteration, and finally offering encouragement to take action.
Now, the same issue occurs when AI is instructed to generate text that “sounds academic.” The lack of variation in sentence length and repeated phrasing makes your manuscript draft monotonous, especially when readers encounter the same pattern in each paragraph. Therefore, rephrase such AI-generated sentences to maintain variations in sentence length. Here’s an example:
AI text: Improving athletic performance is a major area where assessing physical risk plays an important role. Traditional evaluation methods based on past analysis and general guidelines often fail to capture physical stress during sports activities. Therefore, it is important to improve assessment methods to better reflect the changing nature of competitive sports.
Human-edited text: Biomechanical risk assessment plays a pivotal role in enhancing athletic performance. Conventional evaluation methods based on retrospective analyses and ergonomic guidelines often fail to capture real-time biomechanical stress during dynamic activities, strenuous workouts, and athletic movements. Therefore, assessment techniques must evolve to reflect the changing nature of competitive sporting events.
4. Fix sentences that overclaim findings and use cautious hedging language
AI tools tend to sound confident when they respond to user prompts. Even if hedging is used, the wording is quite generic: “this information may vary,” “results can differ,” “it is important to note,” “in some cases,” etc. But with authors? Hedging language should be based on an assessment of the evidence.
For example, you do not need to use modal verbs and probability adverbs with well-established facts and theories. Only when you are genuinely uncertain about data or if expert opinions differ on a topic or when you have limited access to data, use hedges.
Frequently Asked Questions on Loss of Author Voice in AI Writing
1. What are three things to always check in AI-written text?
Any AI-generated content must be thoroughly vetted for syntax, diction, and authenticity of citations.
- Syntax: Check for sentence length, active and passive voice usage, subordinate clauses, and transitional wording
- Diction: Specific vocabulary and word choices that make the tone robotic and monotonous
- Citation authenticity: Always verify whether the citations quoted as references by gen AI tools are indeed accurate
2. How to avoid overusing hedging language in academic writing?
It is alright to use hedging language in academic writing. However, you must be careful not to misuse or overuse it, especially in the Methods and Results sections.
- State the facts when you are certain of your data and the corresponding findings. Second-guessing your results could make the reviewers doubt the validity of your reported findings.
- Avoid using hedges for well-known scientific facts. Hedging language is needed only for uncertain or limited data, unclear evidence, or varying expert opinions.
- Do not repeatedly use hedging as a habit, because it can weaken your overall argument. Unnecessarily layering hedging with “may,” “might,” “could,” “appear to,” etc., even in well-supported claims, impacts the research’s messaging.
3. Which AI tool best retains author voice?
Generative AI tools and chatbots like ChatGPT are not suited for academic writing. If retaining author voice is a concern, use AI writing assistants that only fix your writing rather than generating their own content.
Tools such as Paperpal, SciSpace, and Jenni AI are designed to support academic writing. This reduces the chances of the author’s writing style being compromised. Use the suggestions and outlines from Paperpal to build your manuscript. Brainstorm ideas, paraphrase text, check for plagiarism, and chat with PDF to enhance your writing rather than replace critical thinking.
4. If AI tool fixes my grammar mistakes, will that lead to loss of author voice?
Not necessarily. Author voice is not determined by grammar. So, if an AI tool corrects grammatical errors in your writing, it is only improving your content and not replacing your ideas. Your writing style is determined by syntax, diction, active vs. passive usage, and variation in sentence length. Focus on these aspects and let the tool take care of English grammar.
