There is a need to return to consideration of some aspects of science that should be categorised as “dodgy”.
Dodgy in the sense of dishonest but not at a high level of seriousness. Artificial intelligence (AI) in the form of ChatGPT, launched in 2021 by Open AI, can be used by anybody to write “expert” articles in the Stem (science, technology, engineering and maths) area.
Moreover, there are reports that a significant proportion of research published on sustainability and energy transitions contribute nothing to the advancement of knowledge. Such articles have been described as “scholarly bullsh*t”.
ChatGPT interacts conversationally with users using advanced AI technology. The name ChatGPT stands for “chat generative pre-trained transformer”. There is merit in defining its elements. Chat means users ask questions and receive responses in a conversational format.
Turnkey townhouse likely to tempt downsizers in Blackrock for €525,000
After the Hunt review: This campus drama makes ill use of some very fine actors
‘I’ll have to tell my husband I’ve lost all our money’: Calls reveal trauma of Irish scam victims
‘Isn’t that the horror guy?’ - Kildare author tackles another Hollywood boogeyman
Generative means ChatGPT generates text based on the input received. Pre-trained means it is built on a huge data-set and undergoes massive training. Transformer means the underlying transformer architecture is a neural network transforming sequential data and is particularly good at processing natural language.
I have known about ChatGPT for some time without looking closely at it. A friend recently asked me about it and demonstrated its workings to me by asking the ChatGPT app on his mobile phone to write an essay on why humans need sleep.
In earlier more orderly times published research papers had to first pass rigorous prepublication ‘peer-review’
Within about 10 seconds the app responded with an essay about 600 words long. It was clearly and coherently written and I could see little wrong with the information. So, in preparation for writing this article I asked ChatGPT to “write an essay on fraud in science in the style of William Reville”.
Again, ChatGPT responded within 10 to 15 seconds, producing a nicely-written essay and in my general style. There was little to quibble with in the actual content.
It was preceded by a short blurb about me personally: “William Reville, an emeritus professor at University College Cork, is known for his clear, balanced, and accessible science communication style, particularly through his Irish Times column The Science Notebook. He typically writes in a measured tone, often using historical context, light analogies and a focus on public understanding rather than sensationalism.”
This description is rather endearing – ChatGPT is a real old charmer! It should be noted that this description of me, whatever about any errors of judgment it may be making about my personal abilities, contains a factual error in calling my Irish Times column The Science Notebook.
Julian Kircherr, assistant professor in geosciences at Utrecht University, studies the circular economy in which raw materials and manufactured products are used, repurposed and remanufactured/recycled as much as possible before finally discarding.
He noticed this popular area of research has attracted increasingly more worthless research articles over the years. He describes such research papers as “scholarly bullsh*t” ie “scholarship so pointless and unnecessary that even the scientist producing it cannot justify its existence”.
Kircherr categorises such bullsh*t articles into five categories:
1. Boring question scholarship: These articles all address the same question but analyse different data-sets. Kircherr claims such papers simply replicate previous work and are mostly redundant since the topics are already sufficiently validated.
2. Literature reviews of literature reviews: This circular economy area has become very popular, indeed “trendy”, generating such a high volume of research papers that periodic reviews of prior literature reviews can be justified. But these reviews simply summarise knowledge already known.
3. Recycled research: This is where scholars reproduce prior economic research but rebrand it with the term “circular economy”. Editors of many journals are very slipshod about checking if the submitted research is already in the literature.
4. Master-thesis madness: The practice whereby professors routinely put their names to masters students’ thesis papers to boost their own perceived academic productivity. This encourages quantity of papers over quality.
So, why such pressure to publish ever more papers? Even worthless papers? Kircherr’s answer is the academic culture of “publish or perish”. Academic productivity/worth, on which awards of research support and personal career promotions are made, is measured largely in terms of research-paper output.
In earlier more orderly times published research papers had to first pass rigorous prepublication “peer-review”. But there is now such a proliferation of research journals, some enforcing minimal standards, that very shoddy work can get published. The scientific enterprise is finding it very difficult to rid itself of this malignancy.
William Reville is an emeritus professor of biochemistry at UCC
