Heuristics
Simple rules or shortcuts AI uses to make decisions or judgments quickly.
Heuristics are like mental shortcuts or rules of thumb that AI systems use to make decisions when they don't have enough time or information to consider every single possibility.
Think of it like this: if you're looking for a restaurant, a heuristic might be to always pick the one with the most people, assuming it's good. This isn't always the best choice, but it's a quick way to make a decision. AI uses similar shortcuts to process information and respond faster.
Heuristics are problem-solving methods that employ a practical approach, often involving a mental shortcut or a readily available approximation. They are used when an exhaustive search or a computationally intensive optimal solution is impractical or impossible. In AI, heuristics can guide search algorithms, inform decision-making in complex environments, or simplify complex probability distributions, trading optimality for speed and efficiency.
Heuristics are relevant as AI agents are being developed to handle complex tasks and games. Frontier labs are exploring how AI can learn or be taught to use these efficient decision-making strategies, moving beyond brute force to more adaptive and practical problem-solving.
- Can AI Agents Learn Their Way to the Top? Evaluating Heuristic Learning in a Long-Running Game Agent CompetitionAnthropic · arXiv cs.AI ·
- Teaching LLMs to reason like BayesiansGoogle · Google Research Blog ·
- AI's Energy & Water Demands: Sorting Fact from Fiction with Andy MasleyAnthropic · The Cognitive Revolution ·
- Anthropic Claude
- Anthropic is exploring how AI agents can learn heuristic strategies to improve performance in complex tasks like game playing.
- Google Gemini
- Google Gemini is investigating how to train large language models to use probabilistic reasoning, a form of heuristic, to enhance their learning and adaptation.
- Algorithms
- An algorithm is a step-by-step procedure for calculations or problem
- Bias
- Bias in AI refers to systematic errors or prejudices in the output, often
- Optimization
- Optimization aims to find the best possible solution, often throug
The term 'heuristic' comes from the Greek word 'heuriskein,' meaning 'to find.' It was popularized in the context of problem-solving by psychologist Herbert Simon. Source ↗
Ask a chatbot to recommend a movie. Notice if it asks clarifying questions or just gives a popular suggestion. Then, ask it to explain its reasoning. This can reveal the heuristics it used, like prioritizing widely acclaimed films.
The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.
- PromptRole-play a student struggling with math for AI tutor simulationSimulate a student's math struggles and uncertainty for AI tutor evaluation. · 5 min
- TemplateReusable workflow templates for healthcare teamsTeams can create and share templates for common workflows like documentation or prior authorizations. · 15 min
- TemplateTemplate for building an AI tool tailored to your workA structured approach to defining and building a custom AI tool for specific business needs. · 15 min
- TemplateEvals CSV prompt list for agent skillsA CSV file template for listing prompts used in evaluating agent skills. · 20 min
- HeuristicDouble coding agent task completion with docs linksProviding coding agents with a documentation link significantly increases their task completion rate. · 2 min
Questions people ask
- Are heuristics always bad for AI?
- No, heuristics are essential for AI to function efficiently in complex situations where finding the perfect answer is too slow or impossible. They are a trade-off between speed and accuracy.
- Can AI learn new heuristics?
- Yes, modern AI, especially machine learning models, can learn and adapt their internal heuristics through training on vast amounts of data and through techniques like reinforcement learning.
First seen on Subvolts Mar 4, 2026. Updated Oct 15, 2026. Written by an AI model from the items above and reviewed against them weekly; report an error.