In 2018, I received a grant that supported the adoption of an open educational resource textbook for my Introduction to Advertising large lecture course. Since that time, more than 1,000 students have not had to rent or purchase a book for the class. While actual savings can be difficult to nail down, I’m confident in saying that a minimum of $50,000 has been saved through the single adoption.
Unfortunately, the textbook is very dated. In fact, it’s always been dated. That’s one of the reasons that available as an OER. The book was originally published in 2009 by FlatWorldKnowledge, who was a pioneering company in the realm of open educational resources and digital textbooks. In 2012, the company transitioned to a hybrid model that blended open content with more traditional, paid options. While some content remained freely accessible, they introduced premium versions of their textbooks, which included additional study materials, interactive features, and supplementary resources. Perhaps most importantly they removed the open licenses from the textbooks.
Saylor Foundation, smartly, backed up all the original versions of the textbooks and is now the primary host of the works. As was documented by David Wiley at the time:
Saylor has done a Herculean job, backing up and providing free and permanent access to Word and PDF formats of every Flat World Knowledge textbook – with ePub versions coming in Q1 2013. They’re also inviting anyone who has remixed FWK books to contribute links to their remixes for Saylor’s new Bookshelf.
Each year that goes by has me contemplating if I should continue with the textbook or not knowing that it is not getting any younger. In class, I make sure that we hit all of the ways advertising has evolved since that time (of which there are no shortage!): mobile, social, influencers, programmatic advertising and other data-driven targeting strategies, branded entertainment, etc. More broadly, the textbook plays a secondary/tertiary role in the course. I don’t use any publisher slides or materials, so its value is really as an auxiliary that deepens their understanding of the course content and allows them to spend additional time with key concepts.
But my course evaluations continue to hit the same notes. First, they are extremely appreciative that they are not having to pay for the textbook, but that it’s also terribly dated. As an example, here is a paragraph explaining how Facebook is expanding to include advertising on its platform:
Facebook is ramping up its involvement with the advertising community as well. The company signed a deal in 2008 with Microsoft to let it provide Web search services and associated advertisements directly on the site—at least on the American portion of the social network. Microsoft already sells and manages display advertisements on Facebook, but the additional search function could allow the software giant to catch up to Google (which provides search on MySpace) and Yahoo! (which does the same for Bebo) in the search business.
I’ve stuck with the textbook for simple reasons. Free is forever the most popular choice, and though I continue to evaluate newer textbooks, I haven’t found anything that merits the cost of access. Plus, students aren’t exactly revolting.
The best option would be to revise the textbook since the license allows that, but revisions are an incredible amount of work. So enter my idea to see if ChatGPT could be an effective rewriter for the textbook. I decided to test out a chapter of the book just to see how well it would do. I started with this set of prompts:
You are a professor of advertising. You teach a class called Introduction to Advertising to 100 students every semester. You have used a dated textbook for the past five years. While it’s dated, students appreciated that it is free and accessible online. Do you understand your role?
You are going to use the open educational resource textbook as a base but rewrite it to make it more up to date with changes to the industry. Every time that you rewrite a section include some notes to what changes you made. Do you understand your duty?
I then started feeding it chunks of text and adding the edits to a fork of the GitHub repository of the original website of the text. I really like that Saylor was forward thinking when they put the websites on Github. This makes it incredibly easy to host a separate version as well as transparently show the edits. An example, here’s some before and after code:
You can see that there are some paragraphs that have been completely rewritten and others and that very lightly touched. The fully rewritten paragraph included a reference to “brand ambassadors” who can be seen as “walking billboard” in order to describe how advertising is continuing to evolve. Side note: Interestingly, this was something that was being watched closely during that time period. The early 00s had a brief fad called forehead advertising (also occasionally referred to as skinvertising…. also was cheekily dubbed foreheADs by John Carver of the London agency Cunning. Remember GoldenPalace.com?
The ChatGPT rewrite speaks to the same notion, but talks about digital influencers collaborations instead of walking billboards.
The section that is lightly touched is the general definition of advertising, which really hasn’t changed. Today, advertising is still a form of communication and continues plays a role in informing, persuading, and reminding consumers.
This was encouraging to see that ChatGPT truly was reading my prompts and recognizing that I am only wanting to update outdated references.
But it’s not perfect, and I would continue to remind it of the job at hand. I prompted it to eliminate references to campaigns from 2007 and suggested that it only reference campaigns from 2015 on. In another prompt, I asked ChatGPT to update references to outdated technologies such as MP3 players. Here is how it responded after the rewrite:
Note: In this revised section, I updated the references to technology, replacing outdated examples like MP3 players with more contemporary examples like streaming platforms and digital billboards. The examples provided for campaigns are the “Share a Coke” campaign by Coca-Cola and the “Real Beauty Sketches” campaign by Dove, which still serve as relevant and impactful examples.
It also would tended to leave any actual reference articles in the chapter. I am not sure why, but it seemed to place higher value on actual article citations. If I was doing this in production, it would be wise to go back and find new papers and articles to cite.
And while it was very good at doing simple rewrites, I really struggled with getting good learning objectives out of it. At the very end I asked it to rewrite learning objectives for the chapter. ChatGPT wanted to write LOs for information that was clearly not in the chapter. At one point I asked it why it did this and it apologized and somehow continued the error.
While it’s not perfect, I am still deeply impressed with its ability. It took me only a couple hours (I think? Why does ChatGPT not provide you with timestamps?) and I imagine I would only get faster over time. Later in the afternoon, I discussed with a colleague the potential for it to be a more seamless process if I would have just used a Chrome plugin to rewrite the page directly on the website rather than going back-and-forth with ChatGPT.
ChatGPT is as an accelerant seems like the right mindset. Which is how we should look at these tools in general though, yes? It’s very possible that within a weeks time I could now do what, say, might have taken a semester’s time before. In my books (pun intended), that seems like a pretty big productivity win.
On the other hand, how would a student feel if they knew that their textbook had been revised by man and machine?
Also how do you license the newly remixed content?
But I continue to think that this a best use case for generative AI. I could imagine an entire class project where you have students update a chapter of a textbook and do a similar analysis looking at the final product and reflecting on the changes the tool did or did not make. It reminds a lot of some of the projects I used to closely follow like Robin DeRosa’s open textbook for Early American Literature that was crowdsourced by undergraduate summer assistants.
But is it appropriate to use the final product in the classroom?
These are the questions I continue to wrestle with as I play with these tools, and I would love some thoughts. I’m including below a list of links for those who want to explore the rewritten book:
- Old textbook (webpage, Github repo)
- New textbook (webpage, Github repo) *note only Chapter 2 was rewritten
Featured Image: “A digital image of a dishelved author shown from the back struggling to rewrite a chapter on his laptop in dimly lit office.” Created with DALL·E, an AI system by OpenAI
