What is reinforcement learning?
The system chooses actions and gets feedback on how well they worked.
What is it?
In reinforcement learning a system tries actions and gets a reward for good outcomes and a penalty for bad ones. After many attempts it learns a strategy that earns the most reward. That is how computers learned to play Go, and how language models are fine-tuned further.
An everyday example
A robot vacuum that learns to get round the furniture faster with each clean follows a similar principle: try, see the result, adjust.
Why does it matter?
The method can learn complex sequences of actions, but it depends on how well the reward is defined.
A common misconception
That the system does what we meant. It does what the reward actually rewards, and sometimes finds shortcuts nobody anticipated.
In more detail
What is the core idea?
An agent tries out actions and learns a strategy that earns a bigger reward over time.
Where is it used?
In games, robotics, controlling systems and shaping the behaviour of some AI models.
Is a reward the same as “good”?
No. A badly defined reward can encourage the wrong shortcuts.
Is it the same as an agent?
No. An agent is a way of organising a system; reinforcement learning is one method of learning.
Remember: A system can do an excellent job of optimising the wrong goal.
Reliable sources
Last reviewed 1 November 2026. An explainer, not professional, legal or medical advice. For rules and figures that change, check the official version.