Deep dive · module 72 of 100
Chain-of-thought and guiding reasoning
Ask the model to reason step-by-step.
After this lesson
You can write a prompt that makes the model think in steps.
In view · Step by step
An explanation you can verify
State the data, assumptions, and the desired result.
Ask for important intermediate results, assumptions, or a short explanation of the conclusion.
Recalculate figures or check sources; a plausible explanation can be wrong.
Part 1 of 3 · Define the problem
Your preference for less movement is on. Choose each part yourself.
What is chain-of-thought?
The model is asked to show the steps before arriving at an answer.
When is it useful?
For complex questions, calculations, or decisions with multiple factors.
Note
The reasoning may sound convincing but can still contain errors. Check the steps.
New practice example
From explanation to a usable result.
Ask for verifiable substantiation and intermediate results; a long chain of thought does not prove accuracy.
Fictional practice situationA fictitious answer claims that three sessions of 45 minutes together last two hours. Below it is a long explanation that sounds convincing.
View a possible solution
This is how it can be done
Ask for the calculation with units: 3 × 45 = 135 minutes, so 2 hours and 15 minutes. Check this yourself. You do not need to obtain the model's hidden internal reasoning steps to discover this error.
Why this works
The visible calculation is testable. More text about the thought process is no replacement for checking the result.
Make it slightly more difficult
When creating a summary, consider which intermediate result is verifiable, for example, a list of claims with corresponding source passages.
Extra practice, at your own pace. You do not need to submit anything. This does not count towards an automatic assessment.
Your workplace · learning by doing
Make something. See what changes.
Checking the AI connection… You can already view the written example.
Present your assignment to Ende.
Read the answer, change one thing, and compare.
Previously held
New version
You remain in control
What do you want to improve in your next attempt?
On this device only. AI answers remain only in this page memory. This is self-review, not a grade.
Why this approach works
The visible calculation is testable. More text about the thought process is no replacement for checking the result.
A step further
When creating a summary, consider which intermediate result is verifiable, for example, a list of claims with corresponding source passages.
A prompt to get you started
Solve this problem step-by-step. Show each step before reaching a conclusion. Problem: [problem].
Apply it
You are comparing two fictional options. You want to be able to check the result without asking for a long hidden chain of reasoning.
Which approach best fits this situation?
- Ask for the data used, assumptions, brief substantiation and verifiable calculations.
- Consider a long line of reasoning as proof that the answer is correct.
- Accept only a conclusion without data or assumptions.
With JavaScript, you can check your approach and receive immediate feedback.
Test your insight
What is the advantage of chain-of-thought?
- The model never makes mistakes again.
- You see the reasoning and can verify it.
- It is always faster.
Extra practice
Write a prompt that asks for step-by-step reasoning for a decision or calculation.
“Determine step-by-step whether this quote fits our budget. Show every consideration.”
