Navigating Scientific Literature with ChatGPT: A Personal Guide

You and your AI sidekick

In a world overflowing with information, separating fact from fiction has become an essential skill. Misinformation, especially related to science, is not just misleading—it's dangerous. Traditionally, the role of distilling complex scientific information for the public has fallen on the shoulders of science journalists. However, the field of journalism is currently undergoing a crisis of credibility and accuracy, which exacerbates the risk of misinformation reaching the public. Science journalism is not exempt from this crisis; even reputable outlets like the BBC are not immune from publishing misleading articles based on otherwise valid research.

Given this precarious landscape, it's more important than ever for individuals to be equipped to directly access and understand primary scientific sources. Having a background in conducting technical due diligence for deep tech investments—a rigorous evaluation of a company's technology to assess its viability and potential risks—I've had to navigate through a maze of scientific literature. Often, this involves delving into subjects far removed from my own expertise, such as therapies for Autism Spectrum Disorder, advanced heart surgeries, and sound synthesis (to name just a few). This experience has turned into a rewarding hobby, and I've honed a strategy for dissecting scientific papers and materials effectively.

Here's my approach:

1. I start by gaining a broad understanding, consulting sources like Wikipedia and other introductory materials.

2. I then validate my understanding with experts in the field.

3. Finally, I delve into the specific materials I'm interested in.

ChatGPT, especially its more advanced versions like GPT-4, is incredibly useful for the first step of my approach. If you find its explanations unclear or too complex, you can easily ask follow-up questions. The model will adapt its responses to your level of understanding, providing explanations in a simplified manner if needed, all without any judgment and with endless patience.

Point 2 is obviously better left to humans – if you have access to them, that is. However, my experience is that experts in a field can be surprisingly generous with their time – especially if your questions pique their interest and you approach them respectfully.

Regarding point 3, when reading these papers, one of the most challenging aspects is determining which details warrant further investigation and which can be set aside as peripheral to the main argument. It's a delicate balancing act—overlook a key concept, and you risk misunderstanding the entire paper; focus too much on irrelevant details, and you can waste valuable time and mental energy. Making the right judgment on what to delve deeper into is essential for grasping the broader implications of the research.

The process involves frequent stops to understand complex passages, consuming significant mental energy. However, I've found a workaround using ChatGPT. It serves as an on-demand expert, aiding me whenever I encounter an unclear section. I use the iOS app for GPT to voice-record my queries, which are then transcribed into well-structured text by Whisper, an automatic speech recognition system. This approach not only allows me to maintain a consistent flow in my reading but also keeps me away from my computer and its myriad distractions, further aiding my concentration.

Here's an example from On Consciousness by Bernard Baars, a fascinating but challenging read. One issue is that Baars frequently cites intriguing research but rarely delves into the specifics. It's as if he assumes his readers have access to a well-stocked university library to chase down the original sources—something not applicable in my case.

Me:

Here is a fragment of a book I'm reading. Quote. Short-term memory is the domain of rehearsable, usually verbal, memory. It has been known since Wundt that people can keep in immediate memory only seven or so unrelated words, numbers, or even short phrases. If rehearsal is blocked, this number drops to three or four. Peterson and Peterson, 1959. End quote. Tell me more about rehearsal and in what ways rehearsal can be blocked.

Note that I am able to just say "quote" and "end quote," and the model figures out what is actually going on.

GPT:

Rehearsal in the context of short-term memory refers to the process of actively maintaining or repeating information to keep it in short-term memory. There are generally two types of rehearsal: maintenance and elaborative. Maintenance rehearsal involves simply repeating the information, while elaborative rehearsal connects the new information to existing knowledge, making it easier to remember.

Rehearsal can be blocked or disrupted in various ways:

1. Attention Diversion: Engaging in a different task that requires cognitive resources can limit the ability to rehearse.

2. Interference: Introduction of new, unrelated information can interfere with the rehearsal of the original information.

3. Time Constraints: Insufficient time to rehearse can affect memory retention.

4. Stress or Anxiety: Emotional factors can impede effective rehearsal.

5. Cognitive Load: High amounts of complex information can exceed the brain's capacity for rehearsal.

6. Dual Tasks: Performing two tasks that require rehearsal can cause conflict and hinder the process.

7. Auditory or Visual Disturbance: External sensory input can disrupt internal verbal rehearsal.

Understanding these factors can help in devising strategies to improve memory retention.

Me:

Please give some examples on how a researcher would block rehearsal.

GPT:

Researchers often employ specific techniques to block or limit rehearsal in memory studies. Some common methods include:

Brown-Peterson Task: Participants are given a set of items to remember and then immediately asked to perform a different task, like counting backward, to prevent rehearsal.

Filler Tasks: After presenting the information to be remembered, participants are given unrelated tasks, such as solving puzzles or answering questions, to divert attention.

Articulatory Suppression: Participants are required to continuously repeat an irrelevant word or phrase out loud (e.g., "the, the, the") to block verbal rehearsal.

Masking Sounds: White noise or other audio may be played to interfere with verbal rehearsal.

Dual-Task Paradigm: Participants are required to perform another demanding task simultaneously with the memory task, splitting cognitive resources and making rehearsal difficult.

Speed Constraints: Information is presented too quickly to allow for adequate rehearsal time.

Random Number Generation: Participants are asked to generate random numbers while trying to remember the items, creating cognitive load.

Immediate Recall: Asking for immediate recall of information after presentation, with no gap for rehearsal.

These methods aim to disrupt the rehearsal process, making it challenging for participants to maintain information in their short-term memory.

Another example from the same book:

Me:

Here is another fragment. Quote There is considerable speculation that inner speech may become automatic with practice. Some clinical researchers suggest that people who are depressed may have rehearsed depressive ideation to the point of automaticity so that they have lost the ability to control their self-denigrating thoughts, example, Beck 1976. End quote. Can you tell me more about this?

Note that Baars employs the APA citation style, like "(Baars, 2019)." While a quick Google search might sometimes yield the source, it often falls short, especially for common names. Since I'm reading the book on my Kindle, tracking down these citations is cumbersome. I have to memorize the reference, navigate to the bibliography, locate the specific citation, and then search for it on my phone or computer. This whole process severely disrupts my reading flow.

GPT:

Certainly. The idea that inner speech can become automatic with practice comes from cognitive psychology. In essence, it suggests that repetitive patterns of thinking, including self-talk, can become ingrained to the point where they happen without conscious effort.

The quote refers to Beck's work in 1976, which is likely Aaron T. Beck, a psychiatrist known for developing Cognitive Behavioral Therapy (CBT). His work emphasized how depressive ideation can become automatic if rehearsed enough. In this context, "automaticity" refers to thoughts becoming so habitual that they're triggered effortlessly, often without conscious awareness.

Beck's theory posits that individuals with depression have negative thought patterns ("cognitive distortions") that are automatic and self-reinforcing. These thoughts can revolve around the self, the world, and the future—the "cognitive triad." Because they become automatic, they're hard to interrupt, and often individuals don't even realize they're happening.

CBT aims to break this cycle by making patients aware of these thought patterns and giving them tools to challenge and alter them. Essentially, it aims to replace automatic negative thoughts with more balanced and constructive ways of thinking.

Would you like to know more about how CBT techniques work to combat this?

Here, not only I got some explanation, but also an educated guess on the full name of the researcher. This gives me a clearer direction for my further research, making it easier to find what I'm looking for.

While using ChatGPT offers a convenient method for understanding complex scientific materials, remember that it has its limitations. The model can assist you much like a knowledgeable human, but (similarly to your human expert) it's not infallible—especially concerning citations or specific details. So, you should treat whatever ChatGPT tells you as a stepping stone for deeper exploration into primary sources – not as the definitive answer to your question.

In this era of rampant misinformation, the ability to independently access and comprehend scientific literature is more crucial than ever. ChatGPT can be an invaluable ally in this quest, but it's essential to treat its insights with caution. Think of its input as an informed starting point for your investigations, rather than an irrefutable conclusion.

So, when you encounter baffling scientific jargon or intricate methodologies in your reading, use tools like ChatGPT to shed light on the complexities. However, always make the extra effort to corroborate the insights gained through additional trusted sources. By doing so, you are actively enriching your own understanding and contributing to a better-informed society. Happy reading!

---

Are you reading this? Do you find this interesting? I have no idea, unless you reply to this email or write a comment. I am mostly OK with sending these emails into the void, but I wouldn’t mind feedback. Also, if you know someone who may be interested, please forward this to them!