🔮 Using AI for analytical thinking

How to turn ChatGPT into a powerful thinking partner

Hi, Azeem here.

I am travelling this week — I was in Estonia for a meeting with Prime Minister Kaja Kallas — so I asked EV member Marshall Kirkpatrick to step in and share with us his ways of using ChatGPT for analytical thinking. Marshall presented this to the EV community last month, and our team found it so insightful that we decided to bring it to you. We’re hosting another AI in Practice session for members of Exponential View on 27 July — if you have a practice you’d like to share, fill out this form.

Enjoy,

— A.

---

Hi,

I am Marshall Kirkpatrick, a tech journalist-turned-entrepreneur, and the publisher of SunflowerNews.org, a forthcoming climate change & solutions site. I’ve long experimented with innovative uses of technology, including AI, in strategic work on global challenges.

The internet is chockfull of articles about AI prompts you can use for various purposes. Most of them are simple ways to prompt a large language model to focus its attention on a part of its map of the world most relevant to a given task. That way the language and concepts it recommends will be consistent with your expectations. 

More high-leverage than prompts for content production or translation, though, is using an AI system of your choice to ponder strategic opportunities and facilitate thinking for innovation. I want to share some strategies and prompts you can use to support strategic thinking and innovation in your own mind and work. I’ll cover:

• Analysis and synthesis,

• Problem formation,

• Probabilistic analysis,

• Foresight models,

• And four tips you can implement right away to make the most of your favourite AI at work.

I’m going to explain it to you like you were 5 years old. Just kidding — though that’s a great prompt! I love finding a concept that I want to understand better and asking ChatGPT to “explain it to me like I’m 5, a high school senior, and a college graduate student.” You’ll get progressing levels of complexity, like Hilary Mason explaining machine learning to children through adults.

Here are some of my current favorite ways to examine concepts and opportunities using ChatGPT or other AI models.

Analysis and synthesis

Breaking things down to their constituent parts can be very helpful in exploring an issue or making a plan. For example, take any specific opportunity in front of you. Tell the machine about it and ask it to “break this down for me into its constituent parts that make it a meaningful opportunity.” You could do this yourself, but an AI system will do it faster and it may think of some things you wouldn’t think of. Example: “Explain the Curse of Expertise in 5 parts, using sophisticated academic models as appropriate”. And there’s no reason to stop at a single prompt, you and the AI can think it through together just like you and a person could.

Marshall: Explain the Curse of Expertise in 5 parts, using sophisticated academic models as appropriate.

ChatGPT: The “Curse of Expertise” is a cognitive bias that affects those who have become so skilled or knowledgeable in a particular field that they struggle to view problems from the perspective of less-informed individuals, sometimes leading to ineffective communication or inefficient problem-solving. Here, I will explain the concept in five parts using relevant psychological and cognitive models.

1. Development of Expertise (Dreyfus Model of Skill Acquisition): The first step in understanding the Curse of Expertise is to understand how expertise is developed. The Dreyfus Model of Skill Acquisition proposes five stages of learning: novice, competence, proficiency, expertise, and mastery. As individuals progress through these stages, they begin to rely more on intuition and less on rigid guidelines. By the time they reach the stage of expertise, their knowledge is highly intuitive and integrated. This deep integration of knowledge can sometimes make it difficult to communicate their knowledge to less-informed individuals who are still reliant on structured, step-by-step guidelines.

2. Knowledge Integration and Cognitive Structures (Schema Theory): As per Schema Theory, experts have highly organized cognitive structures related to their field, which enables them to efficiently process and store information. However, this very organization can create challenges when it comes to interacting with non-experts. Experts tend to forget what it is like not to have these schemas, which can result in them glossing over information that may be crucial for a novice’s understanding. This is often referred to as the "Curse of Knowledge".

3. Perspective Taking Failure (Theory of Mind): The Theory of Mind refers to the ability to understand that other people's mental states can be different from one's own. This involves empathetic skills and perspective-taking. Experts often struggle with this because they are deeply entrenched in their own perspective and find it difficult to imagine the state of knowledge of someone less familiar with their field. This is a central part of the Curse of Expertise and leads to ineffective communication and teaching.

4. Overconfidence and Bias (Dunning-Kruger Effect): The Curse of Expertise can also be understood in terms of cognitive biases. One such bias is the Dunning-Kruger Effect, which suggests that those with low ability at a task overestimate their ability, while those with high ability underestimate their own competence. The latter effect can contribute to the Curse of Expertise by making experts overly confident in their ability to communicate complex concepts or solve problems in their field, leading to potential miscommunication or oversight.

5. Mitigating the Curse of Expertise (Instructional Design Theory): Several educational theories offer insights into how to overcome the Curse of Expertise. For instance, Instructional Design Theory suggests that experts should take a structured, systematic approach when designing learning experiences for novices. This involves conducting a needs assessment, defining learning objectives, and creating instructional materials with the learner's perspective in mind. By consciously adopting the perspective of the learner and structuring information in a way that is accessible to them, experts can mitigate the impacts of the Curse of Expertise and effectively convey their knowledge.

Taking things apart is powerful, but putting things together is too. You can put anything together with anything else, from the clearly related, such as “Articulate 3 key similarities and differences between innovation thinking in the 20th and 21st centuries.” Then, “apply those insights to the following situation”, to the not apparently connected at all like: “Write a list of 3 things this process has in common with rehabilitation of injured wild raptors.” You can ask for an analogy about anything. “Explain this using a baseball analogy,” is one of my favorites.  It will often be a simple analogy, and that’s ok. It is often helpful anyway.

Sometimes I like to open a page at random in an old paper book, use my iPhone camera’s OCR to copy and paste the text into the ChatGPT mobile app, and ask “How does THIS relate to what we’ve been talking about?” The connections it can draw are often amazing! It can certainly prompt some unexpected thinking.

As Shawn Wilson explains in this inspiring, interactive talk about his book Research is Ceremony: Indigenous Research Methods, there are powerful worldviews that believe everything is connected to everything and it’s the relations between things that make up the universe. I believe that language sourced from the internet, and played with in dialogue with a large language model, can be another tool to help illuminate some of those connections.

Problem formation

As Oguz A. Acar wrote this month in HBR (AI Prompt Engineering Isn’t the Future), problem formation is where there’s a lot of leverage. Remember how people have said for years, “in the future, success won’t be about knowing the answers - the key to success will be knowing how to ask the right questions.”

That time is clearly here.  

Acar writes that there are four key components for effective problem formulation:

1. problem diagnosis,

2. decomposition,

3. reframing,

4. and constraint design.

Framing, multi-step processes, upstream, downstream, system messages. One way to prepare AI to be your partner in problem formulation is to prompt it with: “You are a strategic foresight consultant, and you know that the future is both foreshadowed by early signals in the present and cannot be assumed to be just an extrapolation of present trends”. 

The opportunities presented by focusing on problem formation in your interaction with AI are nearly limitless. They are certainly overwhelming.  But the opportunities are different for everyone; ask your favorite AI about it.

Probabilistic analysis