Considering the key skills required for effective research communication, Andy Tattersall, discusses how he has used AI tools to augment his work and what this might mean for academics and professionals looking to engage wider audiences.
Depending on what you read, AI is either coming to take your job (and life), or heralding a new dawn for creativity and productivity. The idea that so many processes that swallow up academic time could be offset by the smart use of AI is appealing, but is not without caveats and issues relating to ethics and information governance. Taking a positive angle, this time saved and the input from AI tools themselves could be invested in better research communication and meeting rising demands from funders to produce accessible research outputs. However, even with AI, there are still barriers to communicating research, around time, finances, confidence and knowledge. This raises questions as to where and how AI might benefit research communication.
What do researcher communicators actually do?
From my own experience, a key skill involved in research communication is translation. Primarily, the translation of complex academic writing into ‘simpler’ media and formats accessible to different audiences, eg. lay summaries, press releases and blog posts, or for the public affairs oriented, policy briefings.
Most researchers are taught to write in a particular style, with the end goal of publishing journal papers. This form of peer to peer writing, (eg. peer review, funding bids, editorials and conference posters) often comes at the expense of the informal, less technical and jargon-free. As academics largely write for other academics and funders, naturally they tailor their writing to make sense to this audience.
a key skill involved in research communication is translation. Primarily, the translation of complex academic writing into ‘simpler’ media and formats accessible to different audiences
However, these ‘simplified’ formats entail significant work. Lay summaries and press releases can have multiple authors and reviewers and a 300 word document can receive many comments and go through numerous iterations. This can be difficult for first authors with other priorities. It is easier with experience, or when you have a communications expert on board, but this is a luxurious exception. Until recently, if you did not have the time to write a press release, you paid someone, if you did not have the money, you could draft it yourself. Now you might simply ask an AI to do it.
Fully automated luxury research communication
Does this work? I explored using ChatGPT to assess whether it could produce a good standard first draft of a press release (Disclosure, ChatGPT contributed nothing to this blogpost). I instructed it to write a 300 word lay summary that the general public would understand from a research paper published by my colleagues titled ‘Why do ambulance services have different non-transport rates? A national cross sectional study’.
I pasted the abstract, introduction and conclusion of the paper and received a 300 word press release with the title: ‘New Study Investigates Non-Transport Rates in Ambulance Services and Identifies Factors Influencing Discharge at Scene.’ The summary was well formatted with prompts for a date, contact details and other elements you would expect to see on a press release. The text was rather vanilla and quite dry, but it was still a decent attempt at a first draft.
For an academic to have drafted this it could have taken them several hours, perhaps spread over several days or weeks.
I then instructed ChatGPT ‘Could you rewrite this so it’s 250 words long and a 14-year-old could read and understand it?’ The response was far better and returned the following headline: ‘Study Reveals Why Some Patients Are Not Taken to the Hospital by Ambulances’. The subsequent text was equally straightforward, informative and covered key elements in a clear and concise manner. The lead author acknowledged that it was a very good lay assessment of the paper.
For an academic to have drafted this it could have taken them several hours, perhaps spread over several days or weeks. All that was lacking was a quote or two, contact details and a link to the paper. From this starting point, the document could be shared with co-authors and other stakeholders for review. In addition with the same input it can write a series of threaded social media posts, which it does well, although it does try to shoehorn in a few too many hashtags.
Tools like ChatGPT have real potential for the dissemination of research, as a starting point for creating lay summaries, as well as saving time. It is not without its problems though, as a subsequent test presented the common problem of AI hallucination. The press release generated fictional quotes attributed to an anonymous academic. The quotes were not plagiarised text from the research paper, thus highlighting the need for human peer review.
as major academic search engines use AI tools to read and parse the literature, authors themselves may even find themselves disintermediated in this process
However, this is only the beginning. AI driven translation tools, such as Microsoft Bing’s translation tool, could open up lay summaries to global audiences. Beyond the written word, creative visual platforms such as Adobe, Biteable and Canva now employ AI. The second giving users the ability to generate a short animation on a topic within a matter of seconds. This has huge potential for research subjects that are hard to communicate visually through traditional and creative commons image banks (although copyright legislation remains an ongoing issue). Similarly, straightforward AI-enhanced tools can improve the audio quality of lo-fi research podcasts using Adobe’s AI speech enhancement tool Adobe Podcast. Gamma can help users create a presentation in a matter of minutes. And if you are unhappy with your profile image you can use pfpmaker’s AI tool to generate professional looking headshots. Finally, as major academic search engines use AI tools to read and parse the literature, authors themselves may even find themselves disintermediated in this process, as user interfaces automate much of the explaining and interpretation of papers. This issue is not exclusive to research communications as Wikipedia editors will testify as AI encroaches further into their territory.
Just prompt engineers?
Key to leveraging AI tools is learning how to write effective prompts, the better the prompt the better the results. This becomes another writing skill academics will need to master if they, albeit an easier one. However, to succeed, this still requires knowledge as to what good communications outputs look like and how to use them. If something makes sense to you, it still might not make sense to external audiences, even with AI intervention. Audiences vary, so understanding them remains core to any activity, a factor that AI may struggle with. The effectiveness of research communications is not just the medium and message, but also the network communicators interact with. Effective networking remains a human process, AI might suggest who to network with via social media, but these interactions remain for now largely human.
Effective networking remains a human process, AI might suggest who to network with via social media, but these interactions remain for now largely human
Whilst it may be tempting to imagine AI as a single fix for the problems academia faces relating to time, skills and funding and a means to outsource academic tasks wholesale, this is not yet the case. As AI tools become integrated into academic working practices, we should also be attentive to issues around ethics, ownership and the way AI tools use data. Many of these tools are free now, but there is no guarantee they will be tomorrow. Something early adopters of Web 2.0 academic tools discovered as the developers they relied disappeared or transitioned to costly subscription models. Platforms such as Midjourney and ChatGPT already have subscription models in place, new startups may begin offering their services for free leading to new resource inequalities for those who start to use AI tools. Forward thinking researchers, who have found ways to leverage AI, are likely already factoring funding for this into their grant writing.
The content generated on this blog is for information purposes only. This Article gives the views and opinions of the authors and does not reflect the views and opinions of the Impact of Social Science blog (the blog), nor of the London School of Economics and Political Science. Please review our comments policy if you have any concerns on posting a comment below.
Image Credit: Victor Freitas via Unsplash.