Kobo’s AI Misstep

This was absolutely NOT on my bingo card for today.

I wasn’t planning on doing a blog post today. Sure, I wanted to get into posting to my blog more frequently, but I just did one yesterday. I didn’t need to do one today. And yet, Kobo kind of forced me into it, because now that I’ve started working on my blog again, I feel compelled to talk about this. Hell, if I was ready to start up my YouTube videos again, this would probably be a very ranty video I was recording right now.

And see, this was already a pretty shitty weekend for me. Yesterday was slow as fuck as my internet kept cutting in and out, making progress unbearably slow on the things I was working on (then I discovered my wireless router node was the culprit and I’m avoiding doing any more troubleshooting of it right now because it already gave me a headache once).

So What Happened?

So, yesterday I saw a YouTube video where the YouTuber, Paper Tiger, talked about Kobo’s embracing of AI. Their video was very thorough and went into a lot of detail on which of the features were perfectly okay (AI is not always gen AI) and which to be more concerned about. They showed a response they received from the director of Kobo Writing Life, Tara Cremin, which they seemed to have an overall positive interpretation of.

This video spurred me to send a message to Kobo about one of the features mentioned, specifically their gen AI recap feature they’re either about to beta or are already testing. You can find my email below:

I recently heard about this feature, and I wanted to provide my thoughts as both a reader on Kobo and an author on KoboWritingLife.

Specifically, I don’t want this.

As a Reader

I am fully aware of how inaccurate gen AI summaries tend to be.  I have installed plugins on my browser to turn off the mandatory AI summaries on google because they generally hallucinate details and contradict their source materials.  It was making it harder to do research online and making it easier to belief false information, which I abhor.

As a reader, I would want a recap feature to be accurate, and I know this would not be that.  I would much rather reread half the book than continue reading a book based on information that might be completely false.

Additionally, I am fully aware that ALL the licensable gen AI language models are trained on unethical or soon to be illegal data sets.  For example, I know that Meta’s model actually used a piracy website to pull data.  I also know that the big models are infamously cagey on their environmental impacts and that the better the outputs, the more energy is required to get those results, meaning that my options for these recaps are either recaps that are absolutely horrible or have devastating consequences for the environment, which it sounds like your company would actually care about, considering your environmental impact page.

As an Author

First, I don’t want my books associated with gen AI in any manner, not even recaps.  This defeats part of my stance when I put in my copyright page that gen AI is not used in any part of the production of the book.  It’s a terrible look, and it is a stance a lot of authors are taking.

Additionally, I don’t want readers to have bad experiences, and using a hallucinated/inaccurate recap to inform their experience of my books would inherently hurt their reader experience.  It would also have a deleterious effect on reviews as the reviews would be only partially on my writing and at least somewhat informed by recaps that might not have any relation to my writing at all.  It would make reviews useless for books that used recaps, and while I guess you could turn off the ability to review on books where recaps were used, that wouldn’t stop them from providing reviews on other sites.

Frankly, I expect better from Kobo.  Everything you’ve shown me so far makes me believe that you actually care about the experiences of writers and readers, and this contradicts that stance.

My Recommendation for an Alternative

I would suggest creating an “Author Generated Recap” feature where authors provide 1 sentence recaps of each chapter.  When a reader requests a recap, it concatenates those previously read chapters into one summary so they can get caught up.  This would be beneficial in a lot of ways:

• Improved reader experience because the summary would be created by the person who knows the book best
• Improved trust by the reader
• No gen AI licensing costs
• Reduced environmental impact
• Can maintain a no gen AI stance

The only downside is that adoption would likely be slow and not universal.  While you could require authors submitting new books to provide summaries, a lot of previous books would not have summaries, and while you could encourage authors to update those books, this would be time consuming for authors with a lot of books and some would not bother with their back catalog.

I hope that you found this helpful,

Danielle Forrest

Today, I received a response. This was their response:

Dear Danielle,

We understand that the mention of AI can raise questions, and we want to be transparent about how we’re exploring its potential to improve your reading experience. Rest assured, Kobo will never use your books to train AI models to generate new content.

We are not interested in doing whole summaries of books. Authors provide us with title descriptions – that’s what we rely on to describe the book to customers. We don’t want the risk of spoilers, plot details, or anything else being revealed that would ruin the experience of a book.

What we are interested in is making the reading experience even better for people who have already purchased the book, and finding ways to get more books finished. So, we are exploring a recap inside a book that has already been purchased. One of the problems we see is that, if someone steps away from a book for a period of time, they are much less likely to finish it. Picking up where you left off—getting back into a book—is a problem in both fiction and non-fiction.

As a potential solution, we will be testing this beta feature in the Kobo App:

Readers will be able to request a short, personalized recap of their recent reading activity. The amount of summary will be limited to their last reading session; it will be short and quick, viewable only by the reader and will include references to any highlights or annotations they’ve made. The recaps are short (~150 words), generated on the fly uniquely to the requesting user, and are not permanently saved. The creation of the summary does use a generative AI model, but neither the input or the output are used as training data.

While this feature uses an AI model to generate the recap, no author text will become training data for the model. It’s an application of generative AI, and we understand that some people have concerns about that. We hope you can see we’re trying to use it in a way that helps readers finish books, which is usually the step before they buy the next one—a benefit for everyone. Our ability to bring new readers to Kobo is partially based on our ability to make better, more interesting reading experiences, and this is one area where we compete to make customers happy and sell more books. We are committed to using technology responsibly, supporting authors, and enhancing the reading experience.

Sincerely,
Tara 

Director, Kobo Writing Life

If you watch Paper Tiger’s YouTube video, you’ll notice that this is, in fact, a form email. This is not a human responding, and they didn’t address any of my concerns, or really anything in my email at all.

I. Am. Livid.

I just got finished YESTERDAY talking about how I liked that you could actually reach a person at Kobo and they do this.

Now, generally, I don’t have problems with form emails. They can be very useful for Customer or Tech Support when dealing with common problems. But you have to actually make sure that the form message fits the problem in question. That’s why the HUMAN part is so important. The human has to read the message, decide if a form email is the right approach, then proceed accordingly. Not every issue can be solved with a form email, and this definitely wasn’t one of them. Or at least, that form email was not the right one. Maybe if it had been a form email saying they would forward my suggestion onto the appropriate team, that would be fine. I’d be okay with that. Really, that’s all I ever expected.

But this just got my back up. I just responded to their email right before starting on this blog post, and I really don’t know what to think. Like I said in my initial message to them, this is not a feature I would ever use. I would rather reread the entire book (and have done so before), and as an author, I’m now really concerned, especially since Amazon has already embraced something similar. I’m really worried about how such a feature could warp how readers view my books, and I don’t want that to happen, but I’m worried that there’s nothing I can do.

We’re increasingly moving into an online world that looks more and more like The Matrix. Our view of the things happening around us are being filtered by programs that alter reality for us when we are unable to view it with our own two eyes. Researching and fact checking are becoming increasingly impossible online. I even went so far as using a plugin to remove the AI summary on Google, but that seems to be broken, and the summaries are still popping through. Unfortunately, even the alternatives to Google search that I’ve tried include gen AI as well. When I do image searches, I can clearly see loads of obviously gen AI images, and even the ones that look like they might not be, I’m now questioning.

That isn’t really the point of this post, but it is so unbelievably related. I don’t know where things are going to go. I could certainly choose to remove my books from Kobo and Amazon to avoid this feature. That’s wouldn’t hurt my bottom line too much at the moment, though maybe that would be different with my next book launch. I do want to lean more into Indie spaces and owning the space where my books are featured and sold, but I also wanted to expand the reach of my books, to publish them on more and more vendors and formats.

This is so frustrating, and I kind of hate ending a post on such a weird note, but there it is. I don’t know what’s going to happen, and I did just get a response from “Tara.” It’s another email that could easily be a form email, this time thanking me for my feedback and assuring me that my emails are read and acknowledged and that they welcome my suggestions.

In spite of the fact that I basically just woke up, I’m feeling unbelievably tired right now, exhausted really. Like I said, this was not on my bingo card for today, and I just wish gen AI wasn’t getting pushed down our throats. I wish it was actually ready for the general public. I wish they’d spent more time figuring out how to handle energy requirements. I wish they’d thought more about how to ethically source their training data.

When I first heard about gen AI, I couldn’t help thinking of it as a proof of concept, that it was possible, but flawed. And there’s nothing wrong with that. It’s an important first step in any new technology. It boggles my mind, though, how much the landscape has changed in such a short time. It boggles my mind how much gen AI has been forced into artificial adoption in society. It’s, like I said, forced down our throats until young people see it as perfectly normal and most of the rest of us are screaming about it from whichever standpoint we’re objecting to it from.

It amazes me that so many companies, especially companies who cater to creative sorts, are even considering using it. We and environmentalists are the most vocal in opposition to it, and yet here we are.

Anyway, I hope they are sincere about appreciating my feedback. I hope they actually take it into consideration, and I hope my suggestion helps them in some way to find a more ethical and useful alternative. They don’t have to use my idea, but I hope it, along with any other ideas people might send them, helps to create a feature that actually will benefit their readers.

Anyway, that’s it for now.

Until next time,

Danielle