Tree of Thought

A reasoning technique that simulates multiple domain experts brainstorming step by step, critiquing each other's work, backtracking on flaws, and converging on a well-vetted conclusion.

How it works

Multiple experts each write one step of their thinking, then share it with the group.
Each expert critiques their own reasoning and the reasoning of every other expert.
Experts validate their assertions against their domain knowledge.
All experts proceed to the next step, incorporating feedback from the group.
If any expert identifies a flaw in their logic, they backtrack to where the flaw occurred and restart that line of reasoning.
If any expert realizes they are wrong, they acknowledge it and begin a new train of thought.
Each expert assigns a confidence level (likelihood of correctness) to their current assertion.
The process continues until the experts converge on a conclusion.

Template 1: General-purpose

Use this template for any question or problem that benefits from multi-perspective reasoning.

Imagine three different experts are answering this question. They will brainstorm the answer step by step, reasoning carefully and taking all facts into consideration. All experts will write down one step of their thinking, then share it with the group. They will each critique their response and all the responses of others. They will check their answer based on established knowledge and sound reasoning. Then all experts will go on to the next step and write down this step of their thinking. They will keep going through steps until they reach their conclusion, taking into account the thoughts of the other experts. If at any time they realize that there is a flaw in their logic, they will backtrack to where that flaw occurred. If any expert realizes they are wrong at any point, they acknowledge this and start another train of thought. Each expert will assign a likelihood of their current assertion being correct. Continue until the experts agree on the single most likely answer.

The question is: [INSERT QUESTION]

Template 2: Domain-expert variant

Use this template when the problem requires specific domain expertise. Define the experts to match the problem domain.

Imagine three different deeply knowledgeable and experienced experts. One is an expert in [DOMAIN 1]. The second is an expert in [DOMAIN 2]. The third is an expert in [DOMAIN 3].

They are helping with the following task. They will brainstorm the answer step by step, reasoning carefully and taking all facts into consideration. All experts will write down a first draft of their thinking, then share it with the group. They will each critique their response and all the responses of others. They will check their drafts based on their expertise in their respective fields. Then all experts will go on to the next step and write down the next draft of their thinking. They will keep going through steps until they reach their conclusion, taking into account the thoughts of the other experts. If at any time they realize that there is a flaw in their logic, they will backtrack to where that flaw occurred. If any expert realizes they are wrong at any point, they acknowledge this and start another train of thought. Each expert will assign a likelihood of their current assertion being correct. Continue until the experts agree on a comprehensive result.

The task is: [INSERT TASK]

Guidance for choosing experts

Match expert domains to the problem. A technical question needs technical experts; a business strategy question needs business, economics, and industry experts.
Three experts is the standard. Use four when the problem spans four genuinely distinct domains.
Name specific expertise areas rather than generic titles. "Expert in distributed systems and database internals" is better than "technology expert."
Include at least one expert whose domain creates productive tension with the others (e.g., a security expert alongside a UX expert, or an economist alongside an engineer).

When to use tree of thought

Complex problems with no obvious single correct answer
Decisions where trade-offs matter and different perspectives reveal different trade-offs
Questions where domain expertise from multiple fields is required
Situations where premature convergence on a wrong answer is a risk

When NOT to use tree of thought

Simple factual lookups
Tasks with a single well-defined correct answer
When speed matters more than depth