The Doug Macy

We all know that Large Language Models (LLMs) are powerful tools. Give them a task, and they can generate text, translate languages, and answer questions with astonishing accuracy. But what if there’s a way to go beyond simply tasking an LLM and instead unlock a truly collaborative partnership?

Think about brainstorming with a brilliant but slightly eccentric coworker. At first, your ideas might not fully mesh. You ask them clarifying questions, they make surprising connections, and something far better emerges than either of you imagined at the outset. This is the essence of Conversational Scaffolding.

Speaking to an LLM: Like a Person, But Not

LLMs, like humans, possess a vast store of knowledge. They can adapt to different conversational paths and introduce novel ideas. But they differ in key ways too. Their scope of knowledge is greater, yet they can lack the nuanced understanding of context we take for granted. This difference is why a conversational approach is so powerful.

Conversational Scaffolding: Your New AI Co-Pilot

This framework isn’t about ‘better prompts’ but a mindset shift. Think of interacting with an LLM as guiding an exploration, not issuing a command. The process can be broken down into types of questions and prompts:

• Foundational: Start by activating what the LLM already knows, aligning your understanding.
• Open-Ended Exploration: Encourage broad thinking, let the LLM make unexpected connections.
• Dynamic Interaction: Ask probing questions, refine ideas together as new possibilities emerge.
• Contextual Refinement: Shape the focus, zeroing in on the ‘why’ behind your desired output.
• Targeted Output: Collaboratively draft text, adjust, and iterate until the shared vision is realized.

The Potential: Enhanced Human-AI Cognition

Conversational Scaffolding can help us tap into a unique form of hybrid intelligence. An LLM’s ability to connect and recombine ideas, paired with human experience and judgment, leads to solutions neither could achieve alone. It’s not just about getting better results from an AI – it’s about evolving how we think alongside AI.

This is just the beginning…

Conversational Scaffolding unlocks LLM potential far beyond simple queries. It’s a framework for fostering a collaborative problem-solving space – one with the potential to transform our relationship with this incredible technology. The future isn’t about humans versus AI, but about humans working with AI to build, create, and discover in ways we can’t yet fully imagine.