What is Heuristic?
Heuristic — A practical, rule-of-thumb approach to problem-solving within algorithms that may not be optimal but is sufficient.
Heuristics are practical shortcuts that produce good-enough solutions without exhaustive computation. In AI, heuristics guide search algorithms, set default behaviors, and provide fallback logic when models are uncertain. They trade guaranteed optimality for speed and simplicity.
Frequently Asked Questions
How are heuristics used in AI systems?
As decision rules that complement model predictions. For example, a heuristic might flag any transaction over $10,000 for review regardless of what the fraud model predicts.
Are heuristics better than machine learning?
For simple, well-understood rules, heuristics can outperform ML with zero training data needed. For complex patterns, ML is superior. Many production systems combine both approaches.
When should I use a heuristic instead of a model?
When the rule is simple and well-defined, when you have limited training data, when you need full explainability, or as a safety layer alongside model predictions.