How to use Generative AI for reading and note taking

Introduction

Some Library subscribed ebooks readers and academic databases will provide GenAI features such as summarisation, interactive chat, audio and other formats conversion, or translation in different languages.

Gemini Notebook (formerly Google NotebookLM) is one of the University approved tools together with Gemini and can support your reading and note taking skills.

A number of free or freemium AI and GenAI powered tools and apps are also available that can help with your comprehension and interpretation of texts. These tools are not supported by the University of Sheffield and therefore have not been assessed and approved from an information security point of view. If you choose to use any of these tools, check the terms and conditions before signing up for anything.

Use of Gemini Notebook and other GenAI tools may not be appropriate in every context and at each stage of your learning process.

For a full learning experience and comprehensive understanding it is essential you read, engage with, and respond to the text yourself first. 

Use Gemini Notebook or other GenAI features in ebook readers and academic databases as a way to help you understand, rather than a replacement for reading and to help you organise your thinking, generate questions, and keep track of key arguments. 

Familiarising yourself with a text first, before introducing generative AI tools, will allow you to get the most out of them. 

Employing it well should develop your information literacy and your critical thinking as well as your comprehension of texts.

Visit the Reading and understanding information hub 

Gemini Notebook

In Gemini Notebook you can upload information sources and use it to summarise them, to clarify complex topics, identify recurring themes, and compare different sections of a book or an article.

You can chat with and ask questions about the sources you have uploaded. You can also turn them into different formats such as study guides, mindmaps or audio.

Opportunities 

• Gemini Notebook is designed to work directly only with the texts you upload and will usually ground its responses on that material, reducing the likelihood of hallucinations and fabrications.
• It can provide notes and summaries of texts in a range of different formats. For example, audio, including in the form of a podcast, flash cards, mind maps, or quizzes.

Challenges

• Gemini Notebook can still hallucinate. It works best when asked specific questions about the content of the text you have uploaded. This means you have to read the text in advance yourself. The further you move away from these specific interrogative questions, the more likely hallucinations and fabrications may occur.
• It works best for sources that follow the conventions of academic writing, e.g. texts with clear introductions that set out arguments and are made up of paragraphs that develop a single main point. The more experimental or unconventional the writing, the less likely it will generate helpful responses. 

How does Gemini Notebook differ from Gemini?

Getting started with Gemini Notebook 

This step-by-step advice provides a walk through of how to use Gemini Notebook to enhance your approach to reading.

1. Start with a minimum of a skim read of the text

   • What to Do: Before uploading a text, undertake at the very least a 15-minute skim read through the article or chapter. Look at the abstract, headings, subheadings, introductions, conclusions, and topic sentences. You may also read the text more carefully and/or write a short paragraph or set of bullet points capturing what you have taken away from the text and what further questions it has raised. 
   • Goal: You are trying to form an initial sense of the author’s purpose, scope, and key arguments without the aid of the GenAI tool. Note down terms, issues, or events that seem important as well as questions or points of confusion. This stage is essential to get the most out of Gemini Notebook and to avoid overreliance on its AI outputs.
   • Risk: missing out this step will leave you unfamiliar with the overall structure and tone of the text, making it harder to verify ideas or report on the text later on.
      

2. Upload the text to Gemini Notebook

   • What to do: After uploading the text, Gemini Notebook will generate a summary and suggest questions. 
   • Goal: Analyse the content of the text.
   • Risk: While convenient, the automatically-generated summary and questions can oversimplify, omit crucial issues, and above all discourage you from engaging actively with the text.  Consider ignoring them altogether at this stage. Relying on them is likely to hinder, not help, your comprehension.
      

3. Pose new questions to help comprehension

   • What to do: Go back to your own initial read through and use that as the basis for formulating new questions. 
   • Goal: Analyse it in depth by posing your own questions. Your own informed questions are much more likely to help your understanding than the ones generated by Gemini Notebook. Aim to develop your understanding of the text through dialogue, so ask follow up questions, and keep going until you feel you are developing a confident understanding of the text. If you are confused by something, query it.
   • Risk: AI tools such as Gemini Notebook work best when prompted with thoughtful, interrogative questions about the text itself. Vague prompts produce vague answers.

4. Verify, re-read the original closely, and take careful notes

   • What to do: One of Gemini Notebook’s most useful functions is that it directs you to the text related to the questions you’ve asked.  Most responses to your prompts will include numbers in grey circles. Click on these to locate relevant sections in the original source. You should now carefully read those passages and the surrounding text in the document  you have uploaded.  Note down any key points in your own words, record key quotations with page numbers and clear quotation marks, and jot down ideas for discussion and reflection.
   • Goal: this ensures that you check the accuracy of Gemini Notebook’s output, undertake a closer reading of the original text and produce your own reflections.
   • Risk: Lifting notes directly from Gemini Notebook’s responses is certainly quicker but comes with considerable risks. You are more likely to understand and remember information when putting arguments into your own words, and clear quoting and paraphrasing reduces the risk of academic misconduct. 

5. Multiple texts

   • What to do: One feature of Gemini Notebook is the ability to upload two or more texts at once.  Upload the relevant articles or chapters together (both texts must be selected) then ask Gemini Notebook to draw connections or contrasts. 

   • Goal: This allows you to see how different authors frame similar issues, use evidence, or position themselves within a debate.

   • Risk: Use Gemini Notebook’s comparisons only as a starting point. Multiple text comparison is particularly difficult as GenAI tools lack the knowledge and understanding of their wider theoretical context and research background. 

A note of caution on multi-document summarisation

Multi-document summarisation is particularly difficult for GenAI tools. 

Gemini Notebook, as well as other GenAI tools which claim to be able to summarise and synthesise multiple sources, have a key issue: they have no knowledge of the wider theoretical context. The context is about the circumstances and reasoning that form the setting for the development of ideas and theories, and in terms of which they can only be fully understood.

This is key in being able to produce correct interpretations, analysis, evaluation, synthesis, and comparisons.

Furthermore, the theoretical background and how authors positioned themselves within it is not always explicitly set out in every single piece of writing they produce. Their theoretical framework of reference can change over time or different authors can use different terminology. 

Remember, GenAI tools lack real world understanding. They have no way of comprehending whether what they have generated is true or not as they generate text by predicting the next most likely word. At the same time their summarisation will appear polished and eloquent. GenAI tools excel in producing very convincing and articulate outputs, but it doesn't mean they are correct or accurate.

With thanks to Dr  Andrew Heath, School of History, Philosophy and Digital Humanities at The University of Sheffield who developed guidance on using Gemini Notebook, much of which has been used to inform the development of this page.

Next steps 

• How to use Generative AI for assessment
• How to read for understanding | StudySkills@Sheffield
• How to use Generative AI critically | StudySkills@Sheffield

Further Resources 

• Principles for Generative AI in learning and teaching