Wisdom and philosophy
Occam's razor
When several explanations fit the evidence equally well, prefer the one that needs the fewest extra assumptions.
William of Ockham (c. 1287 to 1347), English Franciscan philosopher. The familiar wording, "entities must not be multiplied beyond necessity", is a later paraphrase that does not appear in his writings.
Ockham argued that you should not posit more things than the evidence and reasoning require. In practice: if two explanations account for the same facts, and one needs a new law of nature while the other needs a loose connector, start with the connector. Every extra assumption is another place the explanation can be wrong.
It works because assumptions multiply risk. An explanation that needs three things to be true is less likely than one that needs one, unless the evidence for the extra three is strong. The razor also keeps you testing cheap hypotheses first, which is how good debugging works.
The razor is a rule for where to start, and it proves nothing. Sometimes the complicated explanation is the true one, and the evidence eventually forces it on you. People misuse it by calling the explanation they already prefer "simpler", or by cutting away evidence that does not fit a tidy story.
Example
In September 2011 physicists measuring neutrinos sent from CERN, the European Organization for Nuclear Research, to Italy's Gran Sasso laboratory reported that they arrived about 60 nanoseconds faster than light would. That would have broken special relativity. Most physicists suspected an equipment error. On 8 June 2012 CERN announced that four experiments at Gran Sasso had measured speeds consistent with light, and that the original result was due to a faulty element in the experiment's fibre optic timing system.
Where it applies
When a production bug appears, check the recent deploy, the config change and the expired certificate before suspecting a compiler or hardware fault. When an AI model's evaluation score jumps, suspect test data leaking into training before you suspect a breakthrough.
Think about it
What problem are you currently explaining with an elaborate theory, and what is the most boring explanation that would also fit?
Applied it to something real? Log it.
Test yourselfWhat does Occam's razor tell you to do when two explanations fit the facts equally well?
Start with the one that needs fewer assumptions. It is a rule for where to look first and does not prove that the simpler one is true.
Related models
- First principlesBreak a problem down to what you know is true and reason up from there, instead of copying what others have done.
- The map is not the territoryEvery model, metric or plan is a simplification, and the parts it leaves out can matter more than the parts it shows.
- Hanlon's razorNever attribute to malice that which is adequately explained by stupidity, or more kindly, by error, haste or ignorance.