Lately, I’ve been talking to generative AI chatbots. No, this isn’t some AI girlfriend thing. I’m married, and though my wife wouldn’t necessarily be jealous of an LLM girlfriend, she’d criticize me for treading through such bizarre territory. What I’ve been talking to AI about is ideas. AI, specifically the ChatGPT iOS app, has become a virtual sounding board for my ideas about safety, writing, and life. This method of using AI has proven incredibly useful.

Striking up a conversation

I begin my conversations with ChatGPT by connecting my phone to my AirPods and opening the ChatGPT app. Next to the text input box is a small headphones symbol that, when clicked, allows you to speak to AI using your voice and hear audio responses in return. The voices are surprisingly realistic—not a Stephen Hawking-esque computer simulation of the past, but a smoother yet detectably non-human tone. This tone is an improvement, though it makes the conversation less weird than it ought to be.

Recently, I clicked the headset icon and said, “I want to write an article.”

ChatGPT said, “That's great! What's the topic or theme of the article you're planning to write? Are you looking for tips on how to get started, or do you need help with research or structuring your article?

I replied, “I want to write another article, but this one is about how the regulation is stifling companies by not allowing them to discover what sort of safety program would meet their needs without understanding the complexities of their own culture, but instead imposing upon them some different models that come from questionable science and evidence and track records. I know there's a need for regulation to impose some rules, but when it comes to governing the way and the methods used to manage safety, isn't it time that we recognize that they might not know what they're talking about?"

ChatGPT then spoke its response, a lengthy summation of where such an article might go,
“It sounds like you're interested in exploring a nuanced critique of regulatory approaches to safety in organizations. To frame your article effectively, you might consider focusing on several key points:

Regulatory Challenges: Begin by outlining the current regulatory landscape and how it attempts to govern company safety protocols. Highlight examples where regulation must be in touch with the practical realities of diverse organizational cultures and operations.

One-Size-Fits-All Approach: Discuss the limitations of a one-size-fits-all approach to safety regulation. You could explore case studies or theoretical models to show how such approaches might fail to account for the unique aspects of different company cultures or the risks various industries face.

Questionable Science and Evidence: Delve into the scientific and evidentiary bases of the prescribed safety models. Critique the quality and applicability of the research underpinning these models, emphasizing instances where the evidence is weak, outdated, or not universally applicable.

Impact on Innovation: Argue how overly prescriptive regulations can stifle innovation within companies. Offer examples of organizations that have developed highly effective, tailored safety programs that deviate from standard regulatory models.

Alternative Approaches: Suggest alternatives to the current regulatory paradigm. Alternatives could involve a more collaborative approach between regulators and industry, using adaptable frameworks that companies can customize or incorporating feedback loops that allow for improving safety protocols based on real-world outcomes.

Call to Action: End with a call to action for regulators, companies, and safety professionals. Encourage a dialogue that respects businesses' expertise and autonomy while ensuring the safety program's primary goal—protecting employees' well—being—is met.

Throughout your article, maintain a balanced perspective by acknowledging the importance and achievements of regulation in enhancing workplace safety historically. Your critique will be more robust if you recognize the complexities of regulating safety across diverse industries and suggest constructive pathways rather than merely pointing out flaws.

Would you like to delve deeper into any specific aspect of this topic, or do you need assistance finding sources or case studies to support your arguments?”

I didn't unquestionably accept ChatGPT's input. Instead, I continued back and forth with the AI until I added all the relevant points I wanted to get across in the piece, such as talking about a specific regulation, referencing the work of authors Malcolm Sparrow and Sidney Dekker, and changing aspects of how the article would be structured.

The Final Product

Ultimately, ChatGPT didn't provide me with a finished piece of work. The outputs from generative AI are flat, overly effusive at times, and lacking in sincerity. AI-generated work has a tone and feel that is detectable to the frequent user. Instead, I received a well-structured draft—a starting point from this exercise. When I return to it, I will undoubtedly find many ways to rewrite it, hone its message, adding more of my personal, human touch to it.

Applications in Construction Safety

Suppose you are out in the field and see a work activity for which you need to assess the risks more adequately. One might open the ChatGPT app, add a photo to the chat, and say, "I want to create a JSA (Job Safety Analysis) for this task." Next thing you know, a smooth, well-spoken, virtual voice will respond positively with a call to action. It's like having an intelligent writing partner egging you on to think of things you might not have thought of--a partner that is happy to take the piece of writing wherever you want it to go, perform research on your behalf, or flesh out ideas that are still emerging in your mind. 

I've found this method of using AI to be particularly beneficial. It could be because I like to vocalize ideas more than I enjoy writing them down. Having a conversational AI is like a propellant that keeps nudging me along until I've created the document or written the article that captures the ideas I want to express.


P.S. This article was not written using this technique. ;-)