December 17, 2023
For our final webinar of 2023, we hosted a discussion about Artificial Intelligence on the anniversary of ChatGPT’s release! For the first time we switched up the format and jumped straight into a discussion. Our Head of Publisher Relations, Romy Beard, hosted an interesting panel of experts from inside and outside of scholarly communications, including our own Co-CEO and Founder Christian Grubak. We heard from Wes Beard, Founder and Managing Director at Konversable, Avi Staimann, Founder and CEO at Academic Language Experts, and Dustin Smith, President and Co-Founder at Hum.
To start with Dustin identified that the first big shift in publishing was from print to digital. Now that has mostly been completed, what’s coming next? On publisher websites every user gets the same experience. This means that nothing is personalized, and when an author accesses a publisher website, they must spend time looking for what they need. A personalized author experience is top of the list for what’s next in publishing. The panel agreed that AI can play a huge role in delivering a bespoke experience for authors.
One solution would be to utilize AI to generate richer insights about each author, so that they are immediately offered the services most relevant to them. By understanding the author’s content consumption up front, you can understand their needs better and provide them with a tailored service. This extends from recommended content on a platform to bespoke emails. For Dustin, using AI to gain insights is all about understanding the individual and moving away from the mass marketing approach of the past.
“Platforms need to utilize AI to understand who you are, where you came from, what you intend to do, how far down the funnel you are, and decide what is the language we use when we address you at this point.” Christian Grubak
It is known that users can get fatigued from looking at lots of different services. This is no different in publishing where authors must choose a journal from hundreds or thousands of options. Can you relate to having multiple tabs open in a browser? If you can, then this is one problem the panel believed AI could solve. Wes and Konversable have managed to overcome this kind of fatigue by having what he termed a “single source of truth”. In practice this means meeting the user where they are at.
Konversable have integrated their chat bot product with Facebook marketplace and WhatsApp for the UK market, because they are the dominant communication platforms in the UK. This makes it much better for the end user according to Wes, “because all the data points have been aligned before the information has been delivered to them. They don’t have to sift through everything.” It is also delivered by a service that they are accustomed to using. Avi agreed, “we can create something really powerful, but if the users don’t understand how to use it or why it is useful for them, then it doesn’t work in reality.”
We are going to take a little detour here, because we know chat bots can sometimes turn users off from engaging with the technology. For Wes, understanding user intention is key for creating an experience that is meaningful and AI supports this. One of the key developments they are working on is making the most of inventories, in this case it could be an inventory of products for a car manufacturer. As a user, you will be shown suggestions based on what you want to see. Essentially, chat bots can be used as a recommendation engine, and this approach can be applied in scholarly publishing.
To get to the stage of being able to offer great recommendations, there needs to be a good data structure behind it. ChronosHub has put a lot of emphasis on using AI to extract metadata from manuscripts and enriching it. Extracting the data has consequences further along the publishing process. All the information will be available to automatically assign articles to read and publish agreements, for example.
Christian told the panel that the AI was trained on actual submissions. “We have been able to compare the submissions to acceptance data. It gave us a chance to look at a submitted manuscript and something that ended up being accepted. As a result, this opens things up for journal matching, acceptance probability, journal recommendation and so on.”
Making the most of integrations is a huge part of this. If information is not shared, then it is hard to get to know the author. If data can be referenced across systems, then it empowers recommendations and the platforms referencing data cross sources become more dynamic. This makes for an improved and more personalized author experience.
As Dustin from Hum said: “You can have a super smart brain but if it is starved of sensory experiences, then it’s not going to be very useful.”
The conversation also brought up localization as something that could be powered by AI. As Avi’s company focuses on providing author services for researchers who need support with translation and academic English, he wanted the panel to consider not just language, but cultural attitudes too. This applies to trust in AI and Avi posed the important question: “How much level of trust is there in the AI tool? India, China are early adopters of AI and are more trusting of the tool. Europeans are more cautious and worried about how their data is being used. Are the audience going to see AI as an opportunity or a threat?”
With the introduction of AI into the publishing flow, publishers will likely want to highlight that something has been AI generated and/or verified by a human. Avi pointed out that, as his company can utilize AI to improve the editing and translation element of publishing, so too is there a risk that AI could lead to more paper mills. This means it is probable that there is going to be a shift to quality and validity in scholarly publishing and services created that provide verification. Wes pointed out that it is more than likely it will be AI that provides these checks!
We are only starting to see how AI can be integrated into author experience. Wes mentioned the latest demo of GPT 4. Someone was able to get help to fix their bike by taking a photo and sending it to Chat GPT. The ramifications of this across retail and healthcare are huge, not just in publishing!
For ChronosHub the focus is making sure researchers are the experts in their research field and not experts in publishing – that is for the publishers! Christian describes it like this: “if we can collect all the data and use other tools to contextualize and segment it, then that is a success. The author goes through a seamless flow, gets what they need and understands that they have completed all the tasks they need to get published successfully.” Using AI to underpin this process, without being detected, is the ideal implementation of the technology in the workflow. We can definitely say that the future looks bright for the use of AI in publishing. Alongside other measures, such as integrations, it can really add value to the author experience.
“We are putting a lot of weight on researchers to comply with their funder. They could ask the publisher, but they might not know, or they could ask their funder. If we could give them the key to some of these answers, so they don’t need to know everything.” Christian Grubak