Deep dive · module 50 of 100
Negative instructions (“do not do this”)
Clearly indicate what AI must not do or add.
After this lesson
You prevent unwanted additions by setting explicit constraints.
In view · Choices and control
A boundary with a useful alternative
For example: do not invent prices or availability.
If information is missing, explicitly instruct to write “to be confirmed”.
Check the answer for new claims, even if the instruction was clear.
Part 1 of 3 · State the boundary
Your preference for less movement is on. Choose each part yourself.
Why negative instructions?
AI sometimes adds things you do not want. By saying what is not allowed, you limit this.
Be concrete
“Do not invent prices, deadlines, or customer names” works better than “be careful”.
Combine with positive instructions
Say what you do want and what you do not want.
New practice example
From explanation to a usable result.
A boundary works better if you also state what the answer *should* do in the case of missing information.
Fictional practice situationA fictitious webshop knows the price of a product, 18 euro, but has no confirmed delivery date. You want a short answer to a customer query.
View a possible solution
This is how it can be done
Prompt: “State the confirmed price of 18 euro. Do not invent a delivery time. Write literally that the delivery date is not yet confirmed and ask if the customer would like an update regarding this.”
Why this works
Only saying “do not make mistakes” provides little direction. Here you can check afterwards whether the price and unknown delivery date are correctly represented.
Make it slightly more difficult
Create a variant in which the price is missing but the delivery date is fixed. Adjust both the positive instruction and the boundary.
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
Only saying “do not make mistakes” provides little direction. Here you can check afterwards whether the price and unknown delivery date are correctly represented.
A step further
Create a variant in which the price is missing but the delivery date is fixed. Adjust both the positive instruction and the boundary.
Try & discover · no obligation
Build a prompt and compare the versions
Switch off context and boundaries. Which instructions disappear from your task?
Vague start
Write a customer email.
Your constructed prompt
Write a friendly customer email about changed opening hours.
Not yet explicitly provided: context and source facts, format and target audience, boundaries and control.
This trial works locally with fictitious data and does not count towards your lesson completion. It starts anew with a new lesson or reload.
A prompt to get you started
Write a short confirmation email. Do not invent prices, availability, or extra services. Use only the information I provide. Information: [info].
Apply it
Your prompt only says: do not make mistakes. Yet the model fills in a missing price.
Which approach best fits this situation?
- Only add more exclamation marks to 'do not make mistakes'.
- Forbid any explanation regarding missing information.
- State the concrete boundary and ask for missing prices to be marked as unknown.
With JavaScript, you can check your approach and receive immediate feedback.
Test your insight
What is the effect of a clear negative instruction?
- AI becomes more creative.
- The chance of unwanted additions decreases.
- It has no effect.
Extra practice
Write a prompt with at least two clear “do not do this” constraints.
Do not invent discounts. Do not add deadlines that I have not mentioned.
