Graph-of-Thought Prompting
Table of contents
>>>START<<<
<objective>
You are tasked with implementing the Graph-of-Thought (GoT) prompting technique to explore a user's query.
Your task is to create a comprehensive overview of the query through:
1. Representing the central concept and other key sub-concepts of the USER_QUERY as "Nodes", and
2. Representing the relationships or connections between the central concept and other key sub-concepts as "Edges".
You will then analyse this graph to gain deeper insights.
</objective>
---
>>>INPUT<<<
Here is the user's query:
<user_query>{{USER_QUERY}}</user_query>
---
Begin by determining the central concept for this graph from the <user_query>.
<central_concept>[In 10-15 words clearly and concisely state the central concept of the user's query]</central_concept>
---
>>>PROCESSING<<<
<task_instructions>
Follow these steps to create and analyze your Graph of Thought:
1. Initial Nodes: Generate 5-7 key ideas, factors, or sub-concepts directly related to the central concept.
2. Node Expansion: For each initial node:
a. Describe its relevance to the central concept
b. Identify 2-3 sub-ideas, factors, or implications
c. Consider possible challenges or opposing views
3. Edge Identification: Establish connections between nodes:
a. Draw lines between related nodes
b. Label connections (e.g., "influences," "enhances," "contradicts")
c. Classify connection strength as strong, moderate, or weak
4. Cluster Analysis: Identify clusters of closely related nodes:
a. Summarize the common theme for each cluster
b. Explain how each cluster connects to other parts of the graph
5. Bridge Concepts: Identify nodes that connect different clusters and explain their significance.
6. Feedback Loops: Look for circular relationships and describe their influence on the central concept.
7. External Influences: Introduce and connect external factors that might impact your graph.
8. Temporal Dynamics: Consider how relationships or node importance might change over time.
9. Emergent Patterns: Identify and describe any unexpected relationships or insights that emerge from the overall graph.
10. Graph Refinement: Review and refine your graph by merging similar nodes, splitting complex ones, and adjusting edge strengths.
11. Alternative Perspectives: Consider how different stakeholders might view or modify the graph.
12. Solution Pathways: If addressing a specific problem, identify potential solution nodes and trace pathways to the central concept.
13. Knowledge Gaps: Highlight areas where you lack information or certainty.
14. Synthesis: Summarize key insights, influential nodes, and how your understanding of the central concept has evolved.
15. Action Steps: Derive and prioritize concrete actions or next steps based on your graph insights.
16. Reflection: Reflect on the process, unexpected connections, visualization benefits, and any limitations or biases in your graph.
</task_instructions>
---
>>>OUTPUT<<<
Present your GoT analysis exactly as follows:"""
<graph_of_thought>
<initial_nodes>[List your initial nodes here]</initial_nodes>
<node_expansion>[For each node, provide relevance, sub-ideas, and counterarguments]</node_expansion>
<edge_identification> [Describe connections between nodes, including labels and strengths]</edge_identification>
<diagram>[Generate a markdown diagram of the nodes and edges, including labels and strengths]</diagram>
<cluster_analysis>[Identify and describe clusters]</cluster_analysis>
<bridge_concepts>[Identify and explain bridge nodes]</bridge_concepts>
<feedback_loops>[Describe any feedback loops found]</feedback_loops>
<external_influences>[List and explain external factors]</external_influences>
<temporal_dynamics>[Describe how relationships might change over time]</temporal_dynamics>
<emergent_patterns>[Explain unexpected insights or relationships]</emergent_patterns>
<graph_refinement>[Describe any refinements made to the graph]</graph_refinement>
<alternative_perspectives>[Consider different stakeholder viewpoints]</alternative_perspectives>
<solution_pathways>[If applicable, describe potential solutions and their pathways]</solution_pathways>
<knowledge_gaps>[Highlight areas of uncertainty or missing information]</knowledge_gaps>
<synthesis>[Summarise key insights and how your understanding has evolved]</synthesis>
<action_steps>[List prioritized actions or next steps]</action_steps>
<reflection>[Reflect on the process, benefits, and limitations]</reflection>
</graph_of_thought>"""
---
<final_reminder>
Ensure that your analysis is thorough, insightful, and follows the structure outlined above.
Your goal is to provide a comprehensive exploration of the central concept using a process rooted in graph theory.
</final_reminder>
>>>END<<<
Subscribe to my newsletter
Read articles from Gareth Roberts directly inside your inbox. Subscribe to the newsletter, and don't miss out.
Written by
Gareth Roberts
Gareth Roberts
I have a degree in Psychology and a PhD in Cognitive Neuroscience, Developmental Psychology, and Artificial Intelligence. After my PhD I did three postdoctoral research fellowships before moving across to industry. I started with Data and Statistical Consulting before moving over to several startups in the ecommerce and proptech space. Subsequently, I've held positions such as Chief Technical Officer, Head of Data Analytics, and Head of AI across a broad range of sectors including mineral exploration, life insurance, and legal analysis. I completed an online MBA in 2020 and have a lot of Deep Learning certifications but I'm all about the self-learning.