What Is Reinforcement Learning?
Reinforcement Learning is a machine-learning paradigm in which an agent learns to make decisions by taking actions in an environment and receiving rewards or penalties that it seeks to maximise over time.
Reinforcement Learning — a machine-learning paradigm in which an agent learns to make decisions by taking actions in an environment and receiving rewards or penalties that it seeks to maximise over time.
Unlike supervised learning, which learns from labelled examples, reinforcement learning learns from trial-and-error feedback. It underpins reinforcement learning from human feedback (RLHF), used to align large language models to human preferences. From a governance view, reward design is a key risk: poorly specified rewards can produce unintended or unsafe behaviour ("reward hacking").
Source: ISO/IEC 22989:2022 (AI concepts and terminology)
Plain-language explanation
Unlike supervised learning, which learns from labelled examples, reinforcement learning learns from trial-and-error feedback. It underpins reinforcement learning from human feedback (RLHF), used to align large language models to human preferences. From a governance view, reward design is a key risk: poorly specified rewards can produce unintended or unsafe behaviour ("reward hacking").
Related terms
See where you stand on AI governance
Take the free 7-question maturity assessment and get a personalised action plan.
Free assessment — 3 minutes →