AI Dictionary › Prompting

Analogical Prompting

Analogical prompting is a technique in which the model is asked to recall or autonomously generate similar previously solved problems before tackling the specific problem posed by the user, so as to use those analogies as a guide for reasoning. The model identifies a recurring structure and applies it to the new case.

Definition

How it works

Unlike few-shot, where examples are supplied by the user in the prompt, here the model itself generates relevant analogous examples based on its own prior knowledge. The prompt typically asks it to identify one or more similar problems with their solutions, and then transfer the solving pattern to the original problem.

Applications

This technique proves effective in mathematical, logical or programming reasoning tasks, where recurring patterns exist that the model can recognize: faced with a new geometry exercise, the model can recall an analogous problem seen during training and adapt its procedure, or faced with a code bug it can recall similar error patterns encountered before.

History & etymology

The name derives from analogical reasoning, a cognitive mechanism long studied in psychology and artificial intelligence, adopted in prompting as an explicit technique once it was observed that inviting the model to generate similar examples on its own improves reasoning quality compared to asking directly for the solution.

How it's used in Grace

Grace looks at whether, in problem solving scenarios, you ask the model to recall analogous cases before arriving at the specific solution.

Related terms

More in Prompting

Put it into practice

From our network

Kaimaki Web: Websites That Win Customers

Custom websites, web apps and digital marketing for growing businesses.

Visit kaimakiweb.com →

From the Agora Intelligence blog

More on agora-intelligence.com →

📱 Download the Android app (beta) iOS coming soon

Say what you mean. Get what you need.

Grace Certified, the AI coach that trains and certifies your prompt engineering, by Agora Intelligence.