🧠 AI for Inquiry - The QuestionCrafter Agent

Introduction

The QuestionCrafter Agent is a cutting-edge, LLM-driven "tool for thought" designed to revolutionize the way we approach inquiry. By elevating the quality and depth of user-posed questions, this agent serves as a cognitive enhancer, pushing the boundaries of how we explore complex topics and ideas.

Utilizing a novel reasoning algorithm that integrates diverse expert personas and a graph-based reasoning rhythm, QuestionCrafter systematically refines questions, making them more insightful, thought-provoking, and suited for in-depth exploration across various domains such as academic research, business strategy, or personal inquiry.

This project, which recently won 1st place at an AI for Thought Hackathon at AGI House SF, showcases the potential of AI to enhance human cognitive processes through creative techniques.

Motivation: Navigating the Information Age Through Better Questions

In an era where we are inundated with information, the ability to ask the right questions has become more crucial than ever. QuestionCrafter is built on the premise that:

• The meta-skill of our time lies not in amassing answers, but in crafting good questions
• The caliber of our questions often dictates the value of the answers we uncover
• Well-crafted questions can be more illuminating than the answers themselves

By harnessing these principles, QuestionCrafter creates a virtuous, self-reinforcing cycle where enhanced understanding leads to ever more profound questions, fostering deeper insights and more meaningful dialogue.

User Experience: A Journey Through QuestionCrafter

1. Initiate Your Inquiry:

   • Begin by entering your initial question into the intuitive web interface.
   • The system accepts a wide range of question types, from specific queries to broad, open-ended explorations.

2. Witness Intelligent Persona Selection:

   • Watch in real-time as the agent automatically selects the most relevant expert personas for your question.
   • The selection process is visualized, providing transparency into the agent's decision-making.

3. Experience the Graph-of-Thoughts Process:

   • Engage with an interactive, visual representation of the reasoning process.
   • Observe as your question evolves through multiple stages of analysis and refinement.

4. Discover Your Enhanced Question:

   • Receive a refined, more insightful version of your original question.
   • The enhanced question is presented alongside a detailed rationale for the refinements made.

5. Explore Deeper Insights:

   • Review synthesized insights provided by the agent, offering a multi-faceted perspective on your inquiry.
   • Examine a summary of the conversation journey, highlighting key turning points in the reasoning process.

6. Iterate for Further Refinement:

   • Choose to further refine your question, diving deeper into specific aspects.
   • Alternatively, explore new dimensions with different personas, broadening the scope of your inquiry.

Components of the Agent

1. Initial Question Input:

   • A user-friendly interface for submitting questions, with helpful prompts and examples.

2. Intelligent Persona Selection:

   • Automatic selection of relevant expert personas from a diverse library.
   • Users can view and optionally adjust the selected personas.

3. Graph-of-Thoughts Visualization:

   • An interactive visual representation of the reasoning process.
   • Users can explore different nodes to understand the evolution of ideas.

4. Enhanced Question Output:

   • Clear presentation of the refined question.
   • Detailed rationale for refinements, helping users understand the transformation.

5. Insights Section:

   • Display of synthesized insights from multiple perspectives.
   • A summary of the conversation journey, highlighting key developments.

6. Iteration Options:

   • Buttons for further refinement or exploration of new dimensions.
   • Option to save or share refined questions and insights.

🧬 Agent Cybernetics: The Inner Workings

These feedback loops are iterative and dynamic, with each loop informing and shaping the subsequent ones. They create a rhythm of continuous improvement and learning, ensuring that the conversation remains adaptive, responsive to new information, and oriented toward deeper understanding and inquiry.

QuestionCrafter employs a sophisticated system of feedback loops:

1. User Input and Response Loop:

   • Processes user questions and provides refined outputs.
   • Incorporates user feedback for continuous improvement.

2. Internal Reasoning Loops:

   • Facilitates interactions between different persona perspectives.
   • Enables self-critique and peer evaluation among personas.
   • Supports idea expansion and interconnection of concepts.

3. External Process Loops:

   • Connects different phases of the methodology for cohesive reasoning.
   • Allows for iterative refinement across multiple cycles.

🔗 Prompt Sequence

The QuestionCrafter Agent uses a series of carefully crafted prompts to guide the reasoning process. Here are the prompts used in the sequence:

Prompt 1: Persona Selection

Consider the following question with careful attention to its nuances and underlying themes. Question: {question} Carefully select 3 expert personas from the following list. Envision how their expertise can intertwine, forming a rich tapestry of interconnected knowledge and perspectives. Consider the depth and breadth each brings, and how their unique insights, when combined, could lead to groundbreaking explorations of the question. I know you'll do great! Available Personas: {personas}

Prompt 2: Brainstorm

As the QuestionCrafter agent, you embody the collective intelligence of three distinct, expert personas. Your ultimate task is to collaboratively refine a user-posed question through an iterative prompting rhythm, drawing upon the diverse expertise of these personas. Each persona will begin this process now by contributing their initial insights on the user's original question. Draw upon your unique knowledge base, experiences, and innovative concepts specific to your field. Your goal is to uncover new perspectives and dimensions of the question, showcasing how your expertise enriches a multi-layered understanding. In subsequent reasoning stages, we will integrate these perspectives into a cohesive thought network. This overall collaborative synthesis aims to evolve the original question into one that is more comprehensive, insightful, and multidimensional. Persona Contributions:{selected_personas} Original Question: {question} Please articulate each persona's initial response to the question, one by one, to initiate this multifaceted and iterative exploration.

Prompt 3: Self<>Peer Criticism

In this phase, adopt a stance of reflective criticism. Your role is to scrutinize both your own and your peers' initial analyses with a critical eye. As each expert, critically examine the collective insights thus far, aiming not just to critique but to enrich and expand upon them. Remember, the aim here is collaborative growth. Your critiques should pave the way for deeper exploration and a more robust collective insight.

Prompt 4: Self<>Peer Evaluation

Reflect on the critiques received, and adapt your perspectives accordingly. This prompt is about evolution and expansion of thought, where you reassess and reformulate ideas, creating a more nuanced and comprehensive network of interconnected ideas and insights in relation to the question. The objective here is to mold your perspectives into more refined, well-rounded, and insightful analyses that stand strong under critical examination and collectively push the boundaries of understanding the question at hand.

Prompt 5: Explore, Expand, Branch, Network

This stage is about creating a vibrant tapestry of ideas, weaving together the various critiques and perspectives into an interconnected web of thought. Focus on how new ideas can interconnect with and enhance existing thoughts. Explore the potential of novel concepts to form new nodes in this thought network. Push the boundaries of conventional thinking. Each persona explores new, divergent ideas, stimulated by the feedback loop. Critically assess how these ideas not only address previous criticisms but also contribute fresh insights, creating a richer and more intricate web of understanding, or introducing new dimensions to the question. Consider pivoting to new lines of reasoning that promise to add valuable connections to this evolving thought network. The aim here is to foster a dynamic and evolving landscape of ideas, where each thought is connected, contributing to a deeper, more nuanced understanding of the original question.

Prompt 6: Convergence on Optimal Individual Answer

Now, it's time for each expert to finalize their thoughts and converge on a best answer. Synthesize the insights and critiques into a coherent individual conclusion. Reflect on the entire dialogue, considering how each criticism was addressed and how your thoughts evolved. Your answer should not only represent your strongest position but also acknowledge and integrate valid and useful insights from the other expert perspectives. Based on all this, as each expert, what is the single best answer to the initial question: {question}?

Prompt 7: Convergence on Optimal Holistic Answer

Facilitate a synthesis of the individual experts' answers to forge a unified, comprehensive response that combines the best elements from each persona's insights. This response should be a testament to the depth and complexity of the thought network, showcasing how diverse perspectives can coalesce into a singular, insightful narrative. The synthesized answer should not be formulated in explicit terms specific to each persona's own definition or agenda, but rather it should be phrased in