Experimenting with Google’s NotebookLM to create a podcast from a source
Today, I decided to experiment with Google’s NotebookLM, which is an AI tool geared towards note-taking and research based on select sources that a user uploads to a notebook. In a nutshell, it’s an AI collaborator that analyzes documents that users upload so that the user can then interrogate the document and even create a podcast that summarizes the source(s). The possibilities for such a podcast creation seem numerous if we consider Bloom’s Taxonomy. (see below).
This graphic, released under a Creative Commons attribution license, provides a quick overview of Bloom's taxonomy of the kinds of cognitive processes often asked of students in educational settings. The graphic reflects the 2001 revision of the original Bloom's taxonomy of educational objectives. For more on Bloom's taxonomy, see the Vanderbilt University Center for Teaching guide on the subject: cft.vanderbilt.edu/guides-sub-pages/blooms-taxonomy/.
To test it out, I uploaded a recent paper that I had written which I have available as a pre-print. The paper is a systematic narrative review of actions for decolonizing higher education in the context of the UK. I use this paper for a few reasons:
I wrote the paper, so I’m very familiar with the content (I have subject matter expertise)
Since I authored the source that I used, I can easily check NotebookLM’s podcast/audio against the content of the paper
I wanted to see what NotebookLM would produce from a piece of work I know well
The paper itself is about 5500 - 6000 words long and the podcast that NotebookLM created was around 20 minutes in length. You can read the paper here and listen to the podcast at the link below.
Initial reflections
Overall, the audio summary is pretty spot-on in providing the overarching ideas of what the sole source is all about. The key arguments appear to be conveyed concisely. That said, these are more generalizations without reference to specific examples. In addition, some notable issues should alert users to treat such tools cautiously.
Content
The podcast interlocutors state that the research I’ve used is all from 2024 when in fact, the research itself (i.e. the systematic narrative review) was completed in early 2024. The individual papers that I reviewed were published between 2020 and 2024. So, this is a significant error in my view as the implications for researchers/students and those interested in the topic would be surprised to hear that all this research suddenly popped up in 2024. The result is that this denies a history of the research, detracting from an established and growing tradition for such research.
Although the overarching ideas are conveyed, there is no reference to specific sources that I used in my paper. So, whereas an academic and/or research-focused podcast might cite and bring in sources, the podcast that NotebookLM creates is more of a summary discussion of the source while omitting the sources themselves. This, I think, is insufficient for academic levels of scholarship and research.
Since the focus of the podcast is on the ideas of the paper, this can be valuable for getting the gist of a topic and perhaps surface-level details; some concepts may be expanded upon. However, since no citations/authors from the source are mentioned, the listener will need to do a lot more research and investigation to understand who said what, and when. The listener will get a bird’s-eye view while missing out on specific authors.
Delivery
The voices of two avatar-interlocutors are obviously from North America from the sounds of their voices. This is, however, problematic for a few reasons.
Podcasting is seen to convey a ‘sonic whiteness’ (see Brekke, 2020 and Smith, 2022 to better understand this concept) which colors the avatar-interlocutors in a certain way (white) through a combination of their accents, inflection and word choice. The delivery of the NotebookLM podcast appears to mimic, in some ways, NPR’s own style, where interlocutors speak in clear diction, representing a 'General American English’ accent that one might hear on national news outlets. This accent tends to overshadow regional accents, much in the way that the Queen’s English and, subsequently, BBC English were the norm across Great Britain.
The style is, in many ways, conversational. While this is beneficial in breaking down complex topics, the use of filler words can sometimes appear too much and detract from the message. I didn’t count the number of times ‘right' is used as a way to convey ‘cool,’ ‘okay,’ ‘I understand,’ and others, but I started to notice it half-way through.
Possible applications
Listening to a podcast about an area of interest
How many times have you wanted to go outside and perhaps listen to a podcast about something related to what you’re studying/researching/generally interested in?
In this case, The podcasting feature of NotebookLM could has potential to take a few sources to give the listener a good, general overview of the topic.
Reviewing and revision
Are you preparing for an upcoming exam? A viva perhaps?
I can see possible use cases for undergraduates, postgraduates and doctoral students who might want to get to grips with basic concepts and/or revise for exams (or perhaps vivas).
A student could upload their source(s) and then create a podcast to listen to afterward.
This will present problems for universities that don’t allow uploading of content to other tools such as ChatGPT and others. This may also present issues for students who could benefit from using such tools like NotebookLM for the purpose of listening to content.
Learning for understanding, and others
Interrogating sources
I haven’t tested the querying features of NotebookLM, but for the podcasting feature, the possibilities seem numerous for a few levels of Bloom’s Taxonomy especially related to remembering and understanding. This could expand to include developing a learner’s abilities to apply and analyze knowledge through listening to others discuss the concepts.
References:
Brekke, A. (2020). 11. The Sound of Yellow Rain: Resisting Podcasting’s Sonic Whiteness. In J. Hendricks (Ed.), Radio's Second Century: Past, Present, and Future Perspectives (pp. 173-190). Ithaca, NY: Rutgers University Press. https://doi.org/10.36019/9780813598505-013
Google (n.d.) NotebookLM. https://notebooklm.google.com/
Hosseini, D. D. (2024, May 28). Actions for decolonizing higher education in the UK: a systematic narrative review of empirical studies. [pre-print] Doctoral assignment. University of Strathclyde. https://doi.org/10.31219/osf.io/baqws
Smith, A. D. (2022). The Black Podcaster-Scholar: A Critical Reflection of Using Podcasting as Methodology as a Black Doctoral Student. Social Media + Society, 8(3). https://doi.org/10.1177/20563051221117576
