# Exploration-Exploitation Trade-off

> Source: https://sukruyusufkaya.com/en/glossary/kesif-somuru-dengesi
> Updated: 2026-05-13T21:08:07.256Z
> Type: glossary
> Category: yapay-zeka-temelleri
**TLDR:** The balance problem between trying new options to gain information and using already known good options.

<p>The exploration-exploitation trade-off is a fundamental tension in reinforcement learning and decision systems. On one side, the system wants to try new options in order to discover better strategies; on the other, it wants to use options already known to work in order to preserve short-term value. Too much exploration can create inefficiency, while too much exploitation can prevent the discovery of better opportunities. This balance sits at the center of good agent design. Learning systems improve not only by relying on what they already know, but also by carefully testing what they do not yet know.</p>