Leveraging Brainstorming with ‘Theory of Mind’ to Enhance Cognitive Output from GenAI

Generative AI tools like ChatGPT and Gemini have transformed how we approach complex cognitive tasks. While simple prompts often yield satisfactory responses, unlocking the full potential of these tools requires more sophisticated strategies.

Let’s explore how blending real-world brainstorming techniques with psychological principles, such as the Theory of Mind (ToM), can elevate the depth and diversity of outputs from generative AI.

The Limitations of Simple Prompts

When dealing with a complex topic like immortality, a simple prompt might look like this:

1. Basic Prompt:
   “Describe immortality.”
2. Persona Prompt:
   “You are a philosopher. Explain immortality.”
3. Combination of Persona and Audience:
   “You are a philosopher. Explain immortality to a high school student.”

While these prompts provide useful responses, they are often linear and may miss the multi-dimensional perspectives that make brainstorming so effective in real-world settings.

Introducing Multi-Agent Prompting

One way to mimic real-world brainstorming is by using multi-agent prompting, where the AI assumes multiple personas to engage in a simulated debate or discussion.
Here’s an example prompt for analyzing the concept of immortality:

“There are three personas: a scientist, a student, and a mother. The scientist will start by introducing the topic. The student and the mother will debate with the scientist, sharing their perspectives. Continue the discussion until I say stop. The topic is immortality.”

Sample Output (ChatGPT):

This approach generates a rich and diverse discussion, offering insights that a single perspective cannot.

---

Enhancing with the ‘Theory of Mind’

The Theory of Mind (ToM) is a concept from psychology that refers to the ability to attribute thoughts, emotions, and intentions to oneself and others. By incorporating ToM into AI prompts, we can enrich the personas involved in the discussion.

Step 1: Analyze AI’s ToM for Each Persona

After the initial brainstorming session, ask the AI to describe the underlying assumptions it used for each persona. For example:

“Describe the personas you assumed based on the Theory of Mind.”

Response Example ( In continuation to earlier ChatGPT discussion thread):

Step 2: Refine and Reintroduce Enhanced Personas

Once you understand the AI’s ToM assumptions, you can refine them further. For instance:

“The mother is not just empathetic but also pragmatic, considering economic and cultural challenges. Adjust this perspective and continue the discussion.”

With this feedback, the AI adapts its personas, resulting in deeper and more nuanced interactions.

---

Applications for Multi-Agent Prompting with ToM

This technique is not limited to philosophical debates. It can be applied across various domains:

1. Corporate Decision-Making:
   A bank is introducing AI tools into its operations. Use three personas—CIO, CEO, and Cybersecurity Head—to debate key challenges.
   Example insights:
   • CIO: Focuses on implementation and scalability.
   • CEO: Considers ROI and organizational alignment.
   • Cybersecurity Head: Highlights risks and compliance concerns.
     
2. Product Development:
   Debate the features of a new mobile app using three personas—developer, marketer, and end-user.
   Example insights:
   • Developer: Prioritizes functionality and technical feasibility.
   • Marketer: Advocates for user-friendly design and branding.
   • End-User: Shares feedback on real-life usability.
     
3. Cultural Analysis:
   Discuss the impact of globalization with three personas—a historian, a local artisan, and a global entrepreneur.
   Example insights:
   • Historian: Reflects on historical patterns and cultural preservation.
   • Artisan: Highlights the impact on traditional crafts and livelihoods.
   • Entrepreneur: Focuses on scaling and market opportunities.

---

Benefits of Combining Brainstorming and ToM

• Diverse Perspectives: Multi-agent prompts simulate real-world debates, providing well-rounded insights.
• Enhanced Context: Incorporating ToM allows for deeper, more empathetic discussions.
• Practical Solutions: This approach mirrors brainstorming sessions, making it highly effective for tackling complex topics in business, education, or creative projects.

---

Conclusion

By merging multi-agent prompting with the psychological principle of Theory of Mind, you can unlock the true cognitive potential of Generative AI tools. This approach is particularly valuable for exploring intricate topics, generating innovative ideas, and simulating real-world problem-solving.
The next time you face a challenging question, try this technique—and watch as GenAI delivers insights that are both profound and actionable.