Writing a literature review often seems like an arduous task, even for seasoned researchers. The prospect of sifting through a mountain of papers and dedicating countless hours can be daunting. But what if there were strategies to streamline and enhance this process?
Welcome to a world where, for better or for worse, you can write a literature review that’s comprehensive, accurate, and well-structured without having to read a single academic paper. Not so long ago, the idea was preposterous!
“Just hold on a minute Bron!! Are you suggesting we should write literature reviews without reading any academic papers?!”
While you technically can, I beg you not to! Instead, I encourage you to leverage the power of AI to augment the process of writing a literature review so you never have to start with a blank page again.
The technology that makes this possible is the AI-powered product called Petal.
Petal stands for ‘Professors et al.’ and the team at Petal was founded by a team with PhD backgrounds which is a bonus when it comes to understanding the needs of researchers.
In this blog post, I’ll introduce you to Petal and share a discount code in case you want to try it out yourself. I’ll also highlight some of its key features that can assist you (but not replace you!) in the literature review process.
If you’re already familiar with the term ‘literature review’, please skip ahead. For those who are new to research, as I once was – scratching my head when someone told me I needed to write a literature review and had no idea what one was – I’ve included a summary below.
What is a Literature Review?
A literature review is essentially what it claims to be… a review of what’s been previously published on a given topic, as documented in (generally academic) literature.
Literature reviews are a necessary cornerstone of academic writing, offering a window into the existing body of knowledge on a given topic. They form an essential part of any research project, serving not just to acquaint the researcher with the current state of knowledge but also aiding in identifying gaps and opportunities for new inquiries and for formulating research questions.
In my academic field, a literature review included in a PhD thesis typically constitutes a chapter of approximately 10,000 words. However, the length may vary depending on disciplinary and university norms and requirements.
Literature reviews can serve many purposes, including mapping the current state of understanding, supporting/debating arguments, demonstrating the relevance and significance of the chosen research topic, formulating research questions, and more.
Conducting a literature review typically involves identifying academic papers related to the topic of interest, reading them, analysing them, and then synthesising the information contained therein to identify key themes, gaps in knowledge, trends in research, methodological approaches, and implications for the ‘real world’ and advancing theory.
This process has traditionally been time-consuming, not to mention labour-intensive, requiring researchers to manually read and analyse a memory-exhausting number of academic journal articles.
This is why I was rather excited (ok, I was downright ecstatic!) when I discovered an AI tool that can help with a considerable amount of heavy lifting involved in the literature review process.
Introduction to Petal’s Multi-Doc Analysis
Petal does many things, but its multi-document analysis feature is what I want to share with you in this blog post.
Petal’s multi-document analysis feature lets you upload a collection of academic papers i.e., the papers you want to write your literature review about. That collection of papers is called a ‘knowledge base’. This knowledge base becomes the training data that’s used to train a chatbot so you can ‘chat’ with your literature.
Once you’ve provided the chatbot with the papers, you’re able to interact with it to get it to answer questions and provide information based on the content of the uploaded papers.
It does this by leveraging Petal’s generative AI capabilities. Generative AI refers to a subset of artificial intelligence that focuses on creating new content.
Petal’s generative AI capabilities effectively enable it to ‘read’ and ‘understand’ the content in the knowledge base, and then, by asking it questions (i.e., prompting it), you can identify and generate summaries relating to key themes and concepts contained in those papers.
In what follows, I’m going to run you through an example of using Petal to analyse a knowledge base and generate a first draft of a literature review.
How to Write a Literature Review Using AI
1. Gather literature
To start the process, gather the academic papers you want to consider/include in your literature review – .pdf versions of each paper are ideal. The process of locating these papers can be facilitated by tools such as Petal and other artificial intelligence literature search tools such as Elicit.
2. Set up a Petal account
Set up your Petal account at www.Petal.org.
The free plan lets you work with Petal to analyse a single paper at a time. If you want to use Petal to analyze multiple papers at once you’ll need to spend a few dollars for the Advanced plan.
You’ll find the signup button on their home page. Click ‘Signup’, and enter your preferred email address. You’ll get a verification code to complete the signup process.
To get a 15% discount on the subscription, click this link and use the discount code broneager at checkout. The Petal team has generously offered me a small commission when you join using the link and code, at no additional cost to you. Doing so helps me offset the rising costs of running this website and testing AI tools to share with you (which are self-funded). I promise only ever to recommend tools I genuinely love, and Petal is one of them!
When you’re signed in, you want to click on the Document Analysis Platform option.
3. Set up a knowledge base
To get your knowledge base set up, upload/import all the .pdfs (the full-text versions of the academic papers) you want to consider in your review.
To do this, click +Upload/Import
There are multiple ways to upload papers into Petal.
If you’re uploading documents from your computer, when you click that option, a new window will open, allowing you to drop and drag all the files (comprising all the .pdfs of the full-text academic papers for my literature review). You could alternatively click ‘Choose file’ or ‘Choose folder’ and navigate to where the .pdfs are stored on your computer.
After indicating what files you want in your knowledge base, click the ‘Begin upload files’ button to begin the uploading process.
Depending on the number and size of your files, and the speed of your internet, it could take a minute or two for the upload process to complete. Trust me, it’s worth the wait!
When you uploaded the .pdfs into Petal, in the background you effectively trained an AI chatbot, which is now aware of the content contained in the documents.
4. Start chatting with your documents
Now that it ‘knows’ what’s in your documents, you can begin having a conversation with the chatbot to interrogate the contents of those documents.
To start interacting (i.e., ‘chatting’) with the documents you uploaded (rather than just one document at a time), you’ll need to select the ‘Multi-doc Chat’ option from the left-hand side menu.
If it’s your first time using Multi-doc Chat, you’ll likely see some instructional popups designed to orient you to Petal’s multi-doc chat interface. I suggest going through the tutorial to get familiar with the user interface.
When your orientation is done, it’s time to start chatting with your literature.
The quality of the information you can extract across all your .pdfs is somewhat determined by the prompts (i.e., instructional commands) you give to the chatbot. If you’re not familiar with writing prompts, I suggest checking out the following article I co-authored with a colleague: Prompting Higher Education Towards AI-Augmented Teaching and Learning Practice.
As an example of a prompt I might use to begin a chat session, here’s one from a recent project I worked on with a team of talented co-authors (publication forthcoming) on the topic of seasonal workers.
Act in the role of an experienced academic researcher, with expertise in Accounting and Accountability, and the seasonal work sector (a seasonal worker is defined as ‘an individual engaged in temporary employment that occurs during specific times of the year, typically in seasonal industries such as agriculture or tourism’). Acknowledge and await further instruction.
You’ll note that I added the ‘Acknowledge and await further instruction’ command at the end of my prompt. This is because the chat window only currently (at the time of writing) allows for up to 400 characters to be entered. By adding this instruction at the end of my prompt, I was able to artificially extend the length of the chat by letting the chatbot know I wanted to tell it more and that it should wait for a follow-up command.
Once you’ve entered your prompt, click the arrow next to the chatbot to run the prompt.
For your next prompt, I suggest providing the chatbot with a goal.
For example, if you want it to help you conduct a general overview/review of the information in the .pdfs, you could prompt it to provide you with a summary of the key themes contained across your literature library (i.e., conduct a thematic analysis of the literature).
Once it provides you with the key themes, you could use follow-up prompts to ask the chatbot to provide you with a summary of each theme, working systematically through each of the themes until you have generated a summary for each. This approach is a very quick way to generate a decent amount of text for your literature review! But remember, it should always be treated as just a first draft.
Because the chatbot will generally summarise information across all the .pdfs, you might want to include some instruction in your prompt to request the chatbot to indicate which .pdfs it is gaining information/insights from when answering your prompts/questions. This becomes useful when needing to include citations in your review to evidence claims.
In the seasonal worker example, I wanted the chatbot to use the information it had access to (i.e., all the .pdfs I previously uploaded) to answer one of our research questions. Thus, my follow-up prompt was:
Provide a detailed, evidence-based response, and always cite the .pdfs you obtained your insights from when forming that response, to answer the following question: “[INSERT MY FIRST RESEARCH QUESTION]”
If the chatbot doesn’t quite give you what you want, you can click the delete button and begin again with a new prompt. You could alternatively create a new chat session to begin again.
If the chatbot generates a section of text but then stops… you can prompt it to continue writing by writing ‘Continue’ in the chat box.
In the screenshot below, you can see that the response provided by the chatbot included links to the .pdfs from which the chatbot obtained its insights. Hovering your mouse over the link will provide more detail, or you can click on the link to load up the full-text version of the .pdf. Alternatively, you might be provided with a hyperlink in the text – click it to open the .pdf the information was obtained from.
Because, in the seasonal work example, I was using the chatbot to interrogate the collection of .pdfs to gain insights into specific research questions, I ran the previous prompt multiple times, while changing the phrasing to reflect the research question I wanted the chatbot to answer.
Provide a detailed, evidence-based response, and always cite the .pdfs you obtained your insights from when forming that response, to answer the following question: “[INSERT MY SECOND RESEARCH QUESTION]”
I continue like this until all my questions are answered.
By cutting and pasting the Petal responses into a Word (or equivalent) document, I find I remove writer’s block (i.e., blank page anxiety) and can begin editing/rewriting/adding/subtracting text to accelerate the writing process.
Using Petal lets you rapidly generate a first draft from which your human input can flourish.
How much does Petal cost?
Petal pricing used to be a little confusing. But, luckily, the Petal team listened to customer feedback and redesigned their pricing structure. Thanks, Petal team!!
You can choose to pay for an annual subscription (which is cheaper overall), or by the month. The monthly pricing is shown below, noting that signing up with a .edu email address affords you a small additional discount.
At the time of writing this blog post, the Advanced Plan (this is the plan you need to purchase if you want to use the Multi-doc chat feature) is $9.99 USD per month if signing up with a .edu email. If you want an additional 15% don’t forget to sign up using this link and apply the discount code broneager at checkout.
The credit system works by deducting credits every time you run a prompt. At the time of writing, credits worked as follows:
- Every query you make for a single-document chat will use 1 Message Credit (this fee structure applies when chatting with a single .pdf document, not the multi-document chat function)
- For multi-document chat, each query uses 8 Message Credits.
Petal’s AI functionality adds a valuable addition to your skill set when conducting multi-document analysis, writing thematic summaries, and generally writing up literature reviews.
It can gather insights and information from a collection of journal articles, allowing researchers to extract relevant details and incorporate them into their review of the current state of research.
However, it is important to note that while Petal can provide convenience and efficiency in gathering information, relying solely on AI tools may limit the exploration of future research directions or potential implications. Therefore, researchers must enhance their writing to also incorporate their insights and knowledge i.e., never rely solely on an AI to do your job!
At the end of the day, there is no substitute for knowing the information contained in the academic papers you’re including in your review.
Petal does soooo much more than I’ve covered in this blog post.
In my workshops and training sessions I deliver online for universities, we delve deeper to explore Petal’s other features including its reference management system, AI Table, AI Create, and learn ways to phrase prompts to get the best out of the tool.
In sum, if you’re anything like me and suffer from the condition that involves reading, and reading, and reading, and having the information enter your brain and rapidly fly out the other side while only retaining the overall gist accompanied by a failure to recall what paper that information came from when it comes time to write up your review… Petal is changing the game!
- To learn more about literature reviews, I recommend checking out the Thesis Whisperer’s fantastic list of #LiteratureReview blog posts.
- Dr Andy Stapleton shares an overview of the free Petal version on YouTube: The AI That’s Changing Academia? Must-See for Researchers!
If you’ve enjoyed this blog post, consider signing up for my newsletter ‘T3’, where I share tips, tricks, and tools on AI + Technology for the higher education sector.