‘Hey ChatGPT, Which Hernia Mesh Should I Use?’

“It ain't what you don't know that gets you into trouble. It's what you know for sure that just ain't so.”

—The Big Short (2015), attributed in error to Mark Twain

With the recent advances in generative artificial intelligence tools, such as ChatGPT, there is a need to understand what AI can and cannot do. Generative AI tools can only reflect the current state of human knowledge and available information. For a subject that is not controversial—examples are the results of every Super Bowl or the details and structure of a specific type of legal agreement—the query results are accurate. I think of it as Google on steroids. But when the topic is controversial or in evolution, the generative AI tool can exhibit the same cognitive biases and errors that we exhibit as human beings. To demonstrate this point, I wanted to explore the application of science and healthcare that I have been passionately studying over the past 20 years.

My motivation comes from learning that our global healthcare system is unsustainable. After the first decade of my studies, I learned that our healthcare system is unsustainable because it is founded on the principles of reductionist science. Over the past century, we have learned that reductionist science is not a valid scientific paradigm. Fortunately, a newer paradigm based on systems and data science is available to apply to healthcare and our world.

As an introduction, I first questioned ChatGPT to learn how it was created and its benefits and weaknesses in its application to science and healthcare. The answer describing how it was created is simple and understandable. However, it does admit a potential for errors.

How were you created?

I am an AI language model called ChatGPT, developed by OpenAI. My training involves exposing me to a large amount of text data from diverse sources, such as books, articles, websites and more. These training data help me learn the statistical patterns and relationships between words and generate human-like responses based on the input I receive.

The training process involves using a deep learning architecture called a transformer neural network, specifically the GPT (Generative Pre-trained Transformer) architecture. During training, the model learns to predict the next word in a sentence given the context of the previous words. This enables me to generate coherent and contextually relevant responses when given a prompt or question.

It’s important to note that while I strive to provide accurate and helpful information, I’m still an AI model and may not always have the most up-to-date information or be completely error-free.

I then started with a series of questions about our current scientific paradigm shift from reductionism to systems science. The first question is to clarify the fact that randomized controlled trials (RCTs) are a reductionist science tool.

Isn’t the RCT a reductionist science tool?

Randomized controlled trials can be considered reductionist in the sense that they aim to isolate the effects of a specific intervention on an outcome of interest, while controlling for other factors that may confound the results.

Then, I confirmed the assumptions that are required for RCTs to be valid.

Aren’t assumptions for RCTs to be valid: nothing can change, variables must be known and controllable, and results must be generalizable?

Yes, these are some of the assumptions that are commonly made for RCTs to be considered valid.

Clearly, none of these assumptions are true. So, after confirming that RCTs are not valid, I wanted to confirm that we are going through the first scientific paradigm shift since the Renaissance, when reductionism was discovered.

Isn’t our world going through a paradigm shift from reductionist science to systems science?

Yes, the scientific community is currently undergoing a shift in thinking, with a growing recognition of the limitations of reductionist approaches and a growing appreciation for the complexity and interconnectivity of systems.

Remember that the ChatGPT answers are only patterns based on predictions from all available information humans create. In this case, there is a lot of denial and cognitive dissonance about letting go of the principles of reductionism and tools like RCTs. I remember how uncomfortable I felt as I learned that the scientific tools of reductionism I was taught and believed in were not valid. So, after answering in the affirmative that our world is going through a scientific paradigm shift, the ChatGPT cognitive dissonance kicked in.

It’s important to note that the shift from reductionist science to systems science is not a replacement of another approach but rather a complement. Both approaches have their strengths and weaknesses, and they can be used in combination to provide a more comprehensive understanding of complex systems.

After spending nearly 20 years studying this scientific paradigm shift, I knew this last statement was false, so I asked a straightforward question.

Isn’t a paradigm shift when one scientific paradigm replaces another one?

Yes, you are correct. In the context of science, a paradigm shift refers to a change in the dominant or prevailing scientific worldview or framework, often accompanied by a shift in the way that problems are approached and solved. This can happen when one scientific paradigm replaces another one, due to new evidence, new theories or new methods that challenge the existing paradigm and provide a better way of understanding the world.

With this more direct question, the ChatGPT answer was correct, directly contradicting the previous statement.

Finally, I asked if it could recommend or predict the best value treatment or product for a specific patient problem.

What is the best-value hernia mesh for my inguinal hernia repair?

As a language model, I don’t have the ability to diagnose or provide specific medical advice. … Ultimately, the best way to determine which mesh is the best value for your inguinal hernia repair is to consult with a qualified healthcare provider.

With a clear understanding of the limitations of generative AI, we should recognize the importance of a human-computing symbiosis that is the foundation of systems and data science principles that can produce AI that we desperately need in healthcare—that is, actual intelligence. Applying the principles of systems and data science using human-computing symbiosis is how innovations and highly accurate predictive algorithms can be generated to lead to a sustainable healthcare system and improve our world.

For healthcare to be sustainable, we must change our thinking and break free from our reductionist mindset that craves certainty and control. The concepts of certainty and control are harmful illusions. The mindset of systems science can overcome this reductionist thinking and allow our human potential for empathy, creativity, discovery and innovation to be unleashed.

We’re all in the global healthcare system and we need one another. Learning together with the incredible computing capabilities of a variety of analytical tools applied to real patient care and networking those learnings and algorithms will allow for a sustainable healthcare system based on value for patients to emerge. Who doesn’t want this? Why wouldn’t they?

This article is from the August 2023 print issue of General Surgery News.