The Peer Review Process and Systems Thinking
Process Safety Systems Thinking
Ugo Bardi has just published a post entitled Carbon Dioxide as a Pollutant. In the post he discusses the problems of publishing scientific papers, and the peer review process in particular.
He says,
. . . it is another example of the incredible inefficiency of the current scientific system. It took six months (!!) to receive an answer from the editors of the journal, with two reviewers requesting just a few marginal changes. Six months for a work that could have been done in a few hours! And we were lucky, because we were trying to publish something new and interdisciplinary and that's notoriously a no-no in modern science.
Since I review potential articles for a professional journal I responded as shown at the end of this post.
Process Safety
Although Ugo’s post was mostly to do with the peer review process and the role of CO2 in the atmosphere, he also alludes to the difficulty of publishing anything that is inter-disciplinary. He is right, and the problem is evident when it comes to articles and papers to do with process safety management. For example,
Safety culture draws on organizational psychology, behavioral science, and change management. It can’t be separated from leadership practices and frontline decision-making.
Management of Change (MOC) is a procedural control that requires understanding of workflow systems, instrumentation, communication theory, materials science, and project management.
Incident investigation and root cause analysis often use tools from forensic science, cognitive psychology, and legal analysis — especially when human error is involved.
At this site I frequently review incident reports from agencies such as BSEE (Bureau of Safety and Environmental Enforcement) and the CCPS (the Center for Chemical Process Safety). (BSEE Safety Alert No. 459: Training, JSAs and Stop Work Authority is an example.) One of my reasons for writing these posts is that the agencies themselves often fail to highlight the systems nature of the causes of the incidents that they describe. The same critique can be applied to many of the articles and papers to do with process safety management.
My Response to Bardi’s Post
I work on the other side of the review process; I am asked to review potential articles for a chemical engineering journal. I agree that the review process is both slow and unreliable. Which begs the question: why?
Here are some thoughts.
The reviewers are not paid in most cases. Therefore, the review process drops to the bottom of the priority pile.
The reviewers are, by definition, leaders in their field, so they are busy people. Reviewing someone’s paper is yet another task.
The papers are often tendentious. In my area the first few paragraphs are usually about the importance of safety. We know that ― get to the point.
The papers are often written by authors whose first language is not English, so they can be difficult to follow at times. This is not a criticism of the authors, but I do wonder why they don’t run the draft through ChatGPT first.
The papers are often highly specialized, so the reviewer will feel obliged to take a pass. For example, my journal has been offered a paper about the chemical industry in China. No one in our group has the knowledge and experience to make constructive comments on this topic.
I have found that the review process is a natural application for Large Language Models. Given a prompt based on the above concerns, these models can be extremely helpful, and they can turn the six-month turnaround to six minutes. (One difficulty is that putting the paper into the LLM provides training material for the LLM itself. However, some of them have ‘private’ switches.)
Actually, I think that the problem is solving itself. If I want to know about a topic in my field I go to ChatGPT or Google. It rarely crosses my mind to read a paper or book. I realize that these engines derive their authority from peer-reviewed information, so we will continue to need professional reviewers.
Note: This comment was reviewed by ChatGPT, and it was helpful.
