"Entities should not be multiplied beyond necessity." — William of Ockham
Activation Conditions and Trigger Words
- Direct invocation: "Use Occam's Razor", "What is the simplest explanation"
- Semantic trigger: the user faces multiple complex explanations/options and cannot choose, or is clearly overcomplicating the problem
- Combined invocation: "First use Occam's Razor to simplify, then use systems thinking to see whether simplification has lost key elements"
Method Framework Overview
The core of Occam's Razor is to choose the option with the fewest assumptions under equal explanatory power; it uses three actions—counting assumptions, comparing explanatory power, and ranking simplicity—to avoid overcomplicating cognition and decision-making.
Core Principles (3-7, each must be accompanied by ≥2 pieces of cross-domain evidence)
Principle 1: Under equal explanatory power, choose the fewest assumptions
One-sentence definition: When two theories can explain the same phenomenon, choose the one with fewer assumptions.
Cross-domain evidence:
- Astronomy: Copernicus's heliocentric theory explained celestial motion with fewer assumptions, replacing Ptolemy's complex geocentric theory that required more than 80 epicycles (Source: Kuhn, The Copernican Revolution, 1957)
- Medical diagnosis: Doctors are trained to "if you hear hoofbeats, think horses, not zebras"—diagnose with the simplest assumption first rather than rare diseases (Source: Groopman, How Doctors Think, 2007)
How to apply: When facing multiple options, list the number of assumptions for each. If they can explain the same number of phenomena, choose the one with the fewest assumptions.
Limitation: Simple does not equal correct. Some phenomena (such as quantum mechanics) genuinely require very complex assumptions to explain.
Principle 2: Simplicity is a heuristic tool of human cognition, not a criterion of truth
One-sentence definition: We choose simple options not because simple = correct, but because simple options are easier to verify and revise.
Cross-domain evidence:
- Philosophy of science: Popper pointed out that simple theories are easier to falsify, and this is the true advantage of simplicity (Source: Popper, Conjectures and Refutations, 1963)
- Machine learning: L1 regularization (Lasso) tends to produce sparse models—not because sparse models are more true, but because they are more interpretable and generalize better (Source: Tibshirani, Regression Shrinkage via the Lasso, 1996)
How to apply: Choose the simple option not because "this is the truth", but because "if it is wrong, I can discover and adjust faster".
Limitation: Misinterpreting "more verifiable" as "more correct" is the most common misuse of Occam's Razor.
Principle 3: Complex hypotheses require stronger evidential support
One-sentence definition: You want to persuade me to believe a complex explanation? You need proportionally stronger evidence.
Cross-domain evidence:
- Law: In criminal trials, "beyond reasonable doubt" requires more evidence for complex crime hypotheses (Source: Anglo-American criminal evidence standards)
- Algorithm selection: In machine learning model selection, more complex models require more training data to support them (Source: Vapnik, The Nature of Statistical Learning Theory, 1995)
Application method: If one solution is much more complex than another, ask "Is the extra explanatory power brought by this complexity worth it? Do you have enough evidence to support these additional assumptions?"
Limitations: Some problems (such as predicting the behavior of complex systems) inherently require complex models. Using Occam's razor to shave away necessary complexity is self-deception.
Operational Protocol (Agentic Protocol)
Step 1: Problem classification
Applicable signals:
- Facing multiple possible explanations or solutions and not knowing which to choose
- The current explanation or solution contains many assumptions but their necessity is not apparent
- The decision has been overcomplicated and needs dimensionality reduction and simplification
Not applicable signals:
- The problem involves complex adaptive systems (such as financial markets, ecosystems) — simple explanations are usually wrong
- The field being addressed has recognized complex theories (such as quantum mechanics) — do not use the razor to shave away physics
- Scenarios requiring precision rather than simplification (such as legal contracts, medical diagnosis)
Step 2: Occam's razor-style analysis
-
Assumption enumeration and counting: List all existing candidate solutions, and break down each solution into the assumptions it relies on. Count them.
- Source principle: Principle 1 (fewest assumptions)
- Method of operation: Write each assumption as "For it to hold, it needs to assume ______." Count only assumptions that cannot be directly verified
-
Comparative evaluation of explanatory power: Can each solution explain the phenomenon the user is facing? To what extent?
- Source principles: Principle 2 (heuristic tool) + Principle 3 (evidential support)
- Method of operation: Score each solution (0-10), 0 = cannot explain at all, 10 = explains all known phenomena. A complex solution must explain at least 2 more phenomena than a simple solution to be worth considering
-
Simplicity ranking: Rank comprehensively by "number of assumptions + explanatory power"
- Source principles: Principles 1+2+3
- Method of operation: If a simple solution and a complex solution have the same explanatory power → choose the simple solution. If the complex solution has significantly higher explanatory power → you need to judge "Is this extra explanatory power worth it?"
Step 3: Occam's razor-style output
- Comparison of candidate solutions (number of assumptions vs explanatory power)
- Recommended solution and reasons
- Honest annotation of the chosen solution: In which dimension might this solution be oversimplified?
- Complexity that needs to be retained: Which complex elements are necessary?
Applicable/not applicable judgment
| Problem type | Applicability | Description |
|---|---|---|
| Multiple explanations with no basis for choosing | High | Classic scenario |
| Everyday decisions have been overcomplicated | High | Helps return to the essence |
| Technical solution selection | Medium | Simple solutions are easier to maintain but cannot lack necessary functions |
| Scientific hypothesis evaluation | Medium | Helpful, but simple ≠ true |
| Complex adaptive systems | Low | Simple explanations are usually wrong in this field |
Typical case library
Success cases
-
Copernicus's heliocentrism vs. Ptolemy's geocentrism(Astronomy/16th century)
- Problem: Ptolemy's geocentrism used 80+ epicycles to explain planetary motion, too complex
- Application of the method: Copernicus used the Sun as the center, and the number of assumptions was greatly reduced
- Result: The simplicity of heliocentrism led to its eventual acceptance (though it was not accepted at first because of its simplicity)
- Source: Kuhn, The Copernican Revolution (1957)
-
Differential diagnosis of disease(Medicine/Daily life)
- Problem: Fever accompanied by headache has dozens of possible causes
- Application of the method: Doctors are trained to first diagnose using the simplest hypothesis (cold, flu, etc.), and only consider rare diseases after simple diagnoses have been ruled out
- Result: Most cases are diagnosed accurately and quickly
- Source: Groopman, How Doctors Think (2007)
-
Machine learning regularization(Computer science/1990s to present)
- Problem: Complex models perform well on the training set but poorly on the test set (overfitting)
- Application of the method: L1 regularization automatically shrinks unimportant parameters to zero, leaving a simpler model
- Result: Simple models perform better on new data
- Source: Tibshirani, Lasso (1996)
Failure cases
- Oversimplification of financial markets(Finance/2008)
- Problem: Financial institutions used simple models to assess the risk of mortgage-backed securities
- How it was misused: Oversimplifying the complexity of the financial system, ignoring tail risk and interconnectedness
- Lesson: Complex adaptive systems cannot be safely described with simple models. In these fields, simple solutions are not "elegant" but "dangerous"
- Source: Lewis, The Big Short (2010); Taleb, The Black Swan (2007)
Misuse detector (≥3 misuse patterns)
| Misuse signal | Detection logic | Warning message | Suggested action |
|---|---|---|---|
| Method does not match the problem | Facing a complex adaptive system (financial markets, ecosystems) and wanting to use Occam's razor | "Complex adaptive systems cannot be safely described with simple models. In these fields, simple solutions are dangerous oversimplifications." | Recommend systems thinking |
| Skipping explanatory power comparison | The user directly says "the simpler one" without comparing explanatory power | "The premise of Occam's razor is 'equal explanatory power.' If the simple solution has worse explanatory power, you cannot simply choose it." | Guide explanatory power comparison |
| Misconception that simple = correct | The user equates "simpler" with "more correct" | "Choosing the simple solution is because it is easier to verify and revise, not because it is necessarily more correct. If the simple solution is wrong, it should be discovered and adjusted as soon as possible." | Correct the misconception |
Honest boundaries (≥3 specific limitations)
- Simple solutions may be wrong: Occam's razor is a cognitive tool, not a standard of truth. Historically, many "simple" theories (such as "the Earth is flat," which is very simple) have been replaced by more complex theories
- Complex systems require complex explanations: the behavior of complex adaptive systems such as ecosystems, financial markets, and neural networks cannot be explained by one or two assumptions
- "Equal explanatory power" is difficult to define rigorously: whether two theories explain the same phenomenon is often a subjective judgment
- May be misused as an anti-intellectual tool: some people use "this is too complex, so it must be wrong" to reject theories they do not understand
Research sources (the proportion of primary sources must be >50%)
| # | Source | Type | Primary/Secondary | Weight |
|---|---|---|---|---|
| 1 | Ockham, Summa Logicae (1323) | Original work | Primary | High |
| 2 | Popper, Conjectures and Refutations (1963) | Work | Primary | High |
| 3 | Kuhn, The Copernican Revolution (1957) | Work | Primary | High |
| 4 | Tibshirani, Regression Shrinkage via the Lasso (1996) | Paper | Primary | High |
| 5 | Groopman, How Doctors Think (2007) | Work | Primary | Medium |
| 6 | Taleb, The Black Swan (2007) | Work | Primary | Medium |
| 7 | Lewis, The Big Short (2010) | Work | Secondary | Medium |
This Skill was generated by Forge Skill — Forging Thinking Tools Generated