Working defaults field guide

What to assume first: 103 razors, laws and principles, the predictable way each one misleads, and 16 collisions between them.

A bias is a default that fired in the wrong place. A rule of thumb is a default someone chose on purpose.

The difference matters, because a rule you chose can be unchosen. You can see it coming, name it in a meeting, and ask whether its conditions hold.

A default is a prior, not a verdict. Every rule here is right inside the conditions that make it true, and the skill is knowing where those conditions end.

Occam is right until the data leave residue. Hanlon is right until the error survives being pointed out. The two-way door is right until the door turns out to be one-way.

Twenty-three have been tested. Fifty-two are practice or wit.

Counted from the entries and printed at the front, because the tag decides how far a rule should be trusted.

Seven sections, in the order a decision needs them

The slide carries the first and the fourth: what the rule says and where it misleads, which is the pair a reader most needs in a hurry. The panel adds why it works. The reading version carries all seven.

Sixteen collisions. Seven conditions settle them.

Several defaults come in opposed pairs. Knowing which of a pair applies is the real skill, and it is teachable, because the deciding condition can almost always be named.

Three misreadings, refused in advance

The third is the one that matters most in practice. Hitchens’s razor, Sinclair’s dictum, Sturgeon’s law and the Streisand effect are all regularly used to end an argument rather than to settle it. The test is symmetry: would you apply the same rule, with the same force, to a claim you agreed with?

Where the weight sits

No family was built to a size. Systems and platforms is the longest because that is where the guide’s readers make the decisions that are most expensive to reverse.

Three guides to going wrong, one to going first, and one to a field

The five share one engine and are meant to be read together, and nothing is repeated between them. Where a phenomenon belongs to two, it is written once, in the guide whose question it answers, and cross-referenced from the other.

Judging claims

Which explanation to hold first, what a claim must bring before it is heard, and how firmly to hold what you end up believing.

Which hypothesis to hold first

Burdens of proof, preferences for the simpler cause, and the rules for attributing intent. Every one of them is a way of deciding what to believe before the evidence is in.

Hitchens’s razor

FIG 04 — WHO CARRIES THE COST OF CHECKING DECIDES WHICH WAY IT TIPS

Sagan standard

FIG 05 — THE FURTHER FROM WHAT IS KNOWN, THE STEEPER THE BAR

Alder’s razor

FIG 06 — THE RAZOR CUTS AWAY THE BOX WHERE POLICY LIVES

Russell’s teapot

STANDARD · a rule that sets a burden or a threshold; a value choice, not a finding Ask whose values set the threshold · pulls against Absence of evidence is not evidence of absence

Occam’s razor

FIG 08 — FEWEST ASSUMPTIONS WINS — ONLY IF THE FIT IS EQUAL

Einstein’s razor

FIG 09 — WHERE “NO SIMPLER” BEGINS IS AN EMPIRICAL QUESTION

Hickam’s dictum

FIG 10 — WHEN EACH CAUSE IS COMMON, SEVERAL AT ONCE IS NORMAL

Chatton’s anti-razor

FIG 11 — ADD WHAT CAN BE TESTED; REFUSE WHAT ONLY RESCUES

Crabtree’s bludgeon

FIG 12 — THERE IS ALWAYS A STORY; THE PRICE IS ITS COMPLEXITY

Zebra rule

FIG 13 — THINK HORSES — UNTIL THE HORSE TREATMENT FAILS

Sutton’s law

FIG 14 — LIKELIHOOD ALONE IGNORES COST AND CONSEQUENCE

Duck test

FIG 15 — WHERE IMITATION PAYS, RESEMBLANCE PROVES LITTLE

Morgan’s canon

FIG 16 — THE RICHER EXPLANATION MUST PASS A TEST THE SIMPLER FAILS

Bradford Hill considerations

FIG 17 — THE WEAK LINK IS USUALLY CONFOUNDING, NOT CHANCE

Hanlon’s razor

APHORISM · compressed experience or wit; no test, and sometimes no possible test Use it as a prompt, never as a reason · pulls against Grey’s law

Grey’s law

FIG 19 — FORESEEABLE HARM IS OWNED, WHATEVER THE INTENT

Cui bono

FIG 20 — MOTIVE RANKS SUSPECTS; IT DOES NOT CONVICT THEM

Mediocrity principle

FIG 21 — YOU ARE TYPICAL ONLY IF NOTHING DREW YOU HERE

How much to believe

Rules for setting the strength of a belief: never quite zero, rarely as interesting as it looks, and weighed against the instrument that produced it.

Cromwell’s rule

DERIVED · follows from logic or probability; true wherever its premises hold Check the premises, not the rule · pulls against Hitchens’s razor

Strong opinions, weakly held

FIG 23 — HOW COSTLY IT IS TO BE WRONG DECIDES HOW STRONG TO BE

Twyman’s law

PRACTICE · a working procedure in wide use that has never been properly tested Judge it by whether it changes the decision · pulls against Littlewood’s law

Littlewood’s law

FIG 25 — SCALE TURNS THE ONE-IN-A-MILLION INTO THE EXPECTED

Rule of three for zero events

FIG 26 — A CLEAN PILOT BOUNDS THE RISK; IT DOES NOT REMOVE IT

Lizardman’s constant

FIG 27 — THE NOISE FLOOR IS A PROPERTY OF THE SAMPLE, NOT A CONSTANT

Sturgeon’s law

FIG 28 — THE BEST WORK LIVES IN A NARROW HEAD OF A LONG TAIL

Brandolini’s law

FIG 29 — REACH TIMES HARM DECIDES WHICH NONSENSE EARNS A REPLY

Betteridge’s law of headlines

FIG 30 — THE LAW PICKS “NO”; THE DATA LEAN SLIGHTLY TO “YES”

Gell-Mann amnesia

FIG 31 — WHAT YOU CAN CHECK IS A SAMPLE OF WHAT YOU CANNOT

Wittgenstein’s ruler

FIG 32 — A STRANGE READING FROM AN UNTESTED INSTRUMENT MEASURES THE INSTRUMENT

Segal’s law

FIG 33 — THE DISAGREEMENT IS THE INFORMATION, NOT THE NUISANCE

Lindy effect

FIG 34 — EACH YEAR SURVIVED BUYS MORE — UNTIL THE WORLD CHANGES

Benford’s law of controversy

FIG 35 — HEAT MEANS IGNORANCE ONLY UNTIL THE REVIEW IS WRITTEN

Judging people

How to read what someone says, why they might be saying it, and what an argument is doing when it stops being about its subject.

What is being said, and why

Charity and suspicion, cooperation and incentive, and the laws of what happens to a conversation once it gets long, public or heated.

Rapoport’s rules

FIG 36 — THE RULES PAY ONLY WHEN THE OTHER SIDE RECIPROCATES

Grice’s maxims

FIG 37 — THE INFERENCE IS ONLY AS GOOD AS THE ASSUMPTION OF COOPERATION

Ideological Turing test

FIG 38 — THE TEST MEASURES UNDERSTANDING, NOT WHO IS RIGHT

Sinclair’s dictum

FIG 39 — MOTIVE RANKS FIRST AND CAN CROWD OUT BETTER EXPLANATIONS

Munger’s incentive rule

FIG 40 — INCENTIVES EXPLAIN MOST WHERE THE WORK IS EASY TO COUNT

Poe’s law

FIG 41 — THE MORE EXTREME THE REAL VIEWS, THE LESS A PARODY CAN SIGNAL

Cunningham’s law

FIG 42 — THE CORRECTION ARRIVES; NOT EVERY READER SEES IT

Godwin’s law

FIG 43 — ANY NON-ZERO CHANCE PER COMMENT SUMS TOWARDS CERTAINTY

Muphry’s law

FIG 44 — THE SHARPER THE CORRECTION, THE CLOSER ITS READERS LOOK

Wiio’s law

FIG 45 — FOUR STEPS AT 80 PER CENT EACH SUCCEED ABOUT 40 PER CENT OF THE TIME

Sayre’s law

FIG 46 — WHEN LITTLE IS MATERIAL, THE SYMBOLIC STAKE IS THE WHOLE FIGHT

Planck’s principle

FIG 47 — CLEAR EVIDENCE CONVERTS; AMBIGUITY WAITS FOR RETIREMENTS

Conquest’s first law

FIG 48 — EXPERT CAUTION IS DATA WHILE THE UNDERLYING CONDITIONS HOLD

Cipolla’s first law

FIG 49 — CIPOLLA’S OWN 2×2: THE LOWER LEFT IS THE ONE PLANNERS MISS

Deciding

What to do before you know enough, which is almost always. Procedures for choosing, and the ways each one tilts the choice.

What to do before you know enough

Procedures for choosing: run the failure backwards, ask what happens next, sort the doors, stop searching at good enough, and leave room for being wrong.

Inversion

FIG 50 — INVERSION WINS WHERE THE WORST CASE CANNOT BE UNDONE

Pre-mortem

FIG 51 — THE VALUE IS IN THE LAST STEP, THE ONE MOST OFTEN SKIPPED

Second-order thinking

FIG 52 — EACH LINK MULTIPLIES THE DOUBT; THE STORY FEELS STRONGER ANYWAY

Two-way door

FIG 53 — MOST DOORS SIT FURTHER TO THE RIGHT THAN THEY FIRST APPEAR

Regret minimisation

FIG 54 — THE FRAMEWORK AMPLIFIES THE REGRET OF INACTION AND MUTES THE REST

10/10/10

FIG 55 — THE LOUDEST HORIZON IS THE SHORTEST; THE LEAST RELIABLE IS THE LONGEST

Satisficing

FIG 56 — THE CURVE FLATTENS LONG BEFORE THE MAXIMISER STOPS

Fredkin’s paradox

FIG 57 — CLOSENESS MEANS LITTLE IS AT STAKE — IF THE SCORES CAN BE TRUSTED

Explore–exploit

FIG 58 — EXPLORATION PAYS EARLY AND LOSES VALUE AS TIME RUNS OUT

Margin of safety

FIG 59 — HIDDEN MARGINS MULTIPLY; POOLED IN THE OPEN, THEY CAN BE MANAGED

Circle of competence

FIG 60 — THE BOUNDARY MATTERS MORE THAN THE SIZE OF THE CIRCLE

Via negativa

FIG 61 — THE SUBTRACTIVE OPTION IS THE ONE NOBODY THINKS TO PROPOSE

Precautionary principle

FIG 62 — THE DECIDING QUESTION IS WHETHER EITHER ERROR CAN BE UNDONE

Skin in the game

FIG 63 — STAKES MAKE PEOPLE CARE; THEY DO NOT MAKE THEM SEE

Rumsfeld matrix

FIG 64 — THE MISSING CELL IS WHERE ORGANISATIONS KEEP WHAT THEY WON’T ADMIT

Map and territory

FIG 65 — EVERY MAP IS WRONG; WHAT MATTERS IS HOW FAR FROM HOME YOU USE IT

Veil of ignorance

FIG 66 — THE VEIL SHOWS WHO COUNTS; IT DOES NOT SAY HOW MUCH RISK TO TAKE

Pareto principle

FIG 67 — THE HEAD HOLDS THE VOLUME; THE TAIL OFTEN HOLDS THE HARM

Building and governing

Defaults for the people who build systems and platforms, and for the people who have to forecast, fund and regulate them.

Defaults for people who build things

Laws from software engineering and interface design, read for anyone responsible for a platform that thousands of teachers and students depend on.

Gall’s law

FIG 68 — THE LAW HOLDS WHERE A SMALL VERSION CAN SERVE REAL USERS

Hyrum’s law

FIG 69 — THE INTERFACE IS WHATEVER USERS CAN OBSERVE, ONCE THERE ARE ENOUGH

Postel’s law

FIG 70 — THE BREAK IS THE MISSING FEEDBACK, NOT THE TOLERANCE ITSELF

Kernighan’s law

FIG 71 — CLEVER CODE IS SAFE ONLY WHILE ITS MAINTAINER CAN MATCH IT

Linus’s law

FIG 72 — THE LAW COUNTS EYES; SECURITY DEPENDS ON WHICH EYES ARE LOOKING

Tesler’s law

FIG 73 — THE MORE PEOPLE REPEAT THE TASK, THE MORE THE SYSTEM SHOULD CARRY

Hick’s law

FIG 74 — DOUBLING THE OPTIONS ADDS ONE STEP, AND PRACTICE REMOVES MOST OF IT

Fitts’s law

FIG 75 — EACH EXTRA BIT OF DIFFICULTY COSTS THE SAME SLICE OF TIME

Jakob’s law

FIG 76 — CONVENTION WINS UNLESS USERS WILL REPAY THE LEARNING COST

Miller’s law

FIG 77 — A MEMORY LIMIT, NOT A RULE FOR HOW LONG A MENU MAY BE

Principle of least astonishment

FIG 78 — THE PRINCIPLE IS ONLY AS GOOD AS THE USERS YOU TESTED IT ON

YAGNI

FIG 79 — THE SHAPE OF THE RETROFIT COST DECIDES WHICH WAY IT TIPS

Rule of three

FIG 80 — THE THIRD CASE OFTEN ARRIVES AFTER THE FIRST TWO HAVE SPLIT

Premature optimisation

FIG 81 — EFFORT SPREAD EVENLY IS MOSTLY SPENT WHERE TIME IS NOT

Law of leaky abstractions

FIG 82 — WHETHER TO LOOK BENEATH DEPENDS ON THE LEAK, NOT THE LAW

Worse is better

FIG 83 — SIMPLE WINS WHEN THE COST OF AN ERROR IN USE IS LOW

Amdahl’s law

DERIVED · follows from logic or probability; true wherever its premises hold Check the premises, not the rule · pulls against Gustafson’s law

Kerckhoffs’s principle

FIG 85 — PUT THE SECRET WHERE SECRECY IS CHEAP TO KEEP AND TO CHANGE

The 90–9–1 rule

FIG 86 — A FEW VOICES, A LARGE AUDIENCE, AND A CURVE RATHER THAN BANDS

Zawinski’s law

FIG 87 — A DESCRIPTION OF THE DRIFT, NOT ADVICE ON WHERE TO STOP

Wirth’s law

FIG 88 — THE SLOWEST DEVICE IN USE DECIDES WHETHER WEIGHT IS WASTE

Ashby’s law of requisite variety

FIG 89 — THE VARIETY MUST BE MATCHED SOMEWHERE; THE DESIGN DECIDES WHERE

Defaults for predicting and governing change

How technologies arrive, why they disappoint and then surprise, why the costs of teaching do not fall, and when control should sit close to the classroom.

Amara’s law

FIG 90 — TWO CLOCKS: ATTENTION MOVES FAST, COMPLEMENTS MOVE SLOWLY

Collingridge dilemma

FIG 91 — KNOWLEDGE AND LOCK-IN RISE TOGETHER UNLESS YOU DESIGN FOR EXIT

Kranzberg’s first law

FIG 92 — NEITHER GOOD NOR BAD, AND NEVER SIMPLY NEUTRAL

Stein’s law

DERIVED · follows from logic or probability; true wherever its premises hold Check the premises, not the rule · pulls against Moore’s law

Clarke’s first law

FIG 94 — THE KIND OF IMPOSSIBILITY DECIDES WHETHER THE EXPERT IS WRONG

Hutber’s law

FIG 95 — SUSPICION IS RIGHT WHERE THE USER CARRIES THE COST

Tinbergen rule

FIG 96 — CONFLICTING GOALS NEED SEPARATE LEVERS; ALIGNED ONES CAN SHARE

Gresham’s law

DERIVED · follows from logic or probability; true wherever its premises hold Check the premises, not the rule · pulls against Thiers’ law

Joy’s law

FIG 98 — OPENNESS PAYS WHEN A STRANGER COULD SOLVE IT AND YOU COULD TELL

Streisand effect

MEASURED · an empirical regularity with data behind it; holds within the range measured Trust it inside the range measured · pulls against Chilling effect

Moore’s law

FIG 100 — A FIFTY-YEAR TREND THAT BEHAVED LIKE A LAW — UNTIL IT SLOWED

Metcalfe’s law

FIG 101 — n² ASSUMES EVERY LINK IS WORTH THE SAME — FEW ARE

Martec’s law

FIG 102 — THE GAP ALWAYS WIDENS — THE CHOICE IS WHICH PART TO CLOSE

Baumol’s cost disease

FIG 103 — WHERE THE TIME IS THE SERVICE, COST RISES WITHOUT ANY WASTE

Solow’s productivity paradox

MEASURED · an empirical regularity with data behind it; holds within the range measured Trust it inside the range measured · pulls against Amara’s law

Subsidiarity

FIG 105 — CAPACITY AND SPILLOVER DECIDE WHERE A DECISION BELONGS

The bitter lesson

FIG 106 — SCALE WINS ON CAPABILITY; PEDAGOGY DECIDES HOW IT IS USED

When rules collide

Sixteen collisions between defaults that give opposite advice, and the seven conditions that settle almost all of them.

Knowing which of a pair applies

Sixteen collisions between defaults that give opposite advice about the same case, each settled by a condition that can be named and, usually, measured.

Occam’s razor against Chatton’s anti-razor

FIG 107 — A CAUSE EARNS ITS PLACE BY PREDICTING WHAT IT WAS NOT BUILT TO FIT

The zebra rule and Sutton’s law against Hickam’s dictum

FIG 108 — MULTIPLY THE HORSES BEFORE YOU CHASE THE ZEBRA

The duck test against Morgan’s canon

FIG 109 — CHANGE THE SURFACE, KEEP THE STRUCTURE, WATCH WHAT SURVIVES

Clarke’s first law against the Sagan standard

FIG 110 — ONE STUDY AT P = .05 CANNOT CARRY A HUNDRED-TO-ONE PRIOR

Hanlon’s razor against Cipolla’s law and cui bono

FIG 111 — HONEST MISTAKES DO NOT ALL FALL ON THE SAME SIDE

Skin in the game against Sinclair’s dictum

FIG 112 — THE PILOT WANTS A LANDING; THE CHAMPION WANTS A VERDICT

Linus’s and Cunningham’s laws against Brandolini’s law

FIG 113 — MANY EYES HELP ONLY IF THEY ARE LOOKING AND WANT IT FIXED

The precautionary principle against the two-way door

FIG 114 — MOST DECISIONS ARE TWO DOORS BOLTED TOGETHER — SPLIT THEM

YAGNI against the margin of safety

FIG 115 — A MISSING FEATURE MAKES USERS WAIT; A MISSING MARGIN MAKES THEM FAIL

Satisficing against regret minimisation

FIG 116 — SPEND DELIBERATION WHERE THE OUTCOMES DIFFER, NOT WHERE IT FEELS HARD

Premature optimisation against Wirth’s law

FIG 117 — THE SAME FIFTY MILLISECONDS IS TRIVIAL OR DECISIVE DEPENDING ON SCALE

Subsidiarity against Metcalfe’s law

FIG 118 — SET THE STANDARD CENTRALLY, LEAVE THE LESSON LOCAL

Postel’s law against Hyrum’s law

FIG 119 — WHAT YOU TOLERATE TODAY BECOMES THE SPECIFICATION TOMORROW

The Lindy effect against Amara’s law

FIG 120 — LINDY JUDGES THE PRODUCT; AMARA JUDGES THE CURVE BENEATH IT

The circle of competence against explore–exploit

FIG 121 — EXPLORATION IS AN INVESTMENT THAT NEEDS FUTURE ROUNDS TO PAY OFF

Via negativa against the Tinbergen rule

FIG 122 — BEFORE YOU TAKE AWAY A LEVER, LIST EVERYTHING IT WAS PULLING

Seven questions to ask before a rule is applied

The seven are not independent, and the guide does not pretend they are. Reversibility is partly a matter of error costs; scale changes the time horizon, because a rule that touches more people is harder to change later. They are separated because they are asked in different ways and answered with different evidence.

Four habits that turn a rule into a decision

None of the four is novel. The claim is narrower: that a room which did these four things routinely would catch most of the failures catalogued in the preceding six families before they cost anything, and that the fourth — saying in advance what would change your mind — is the one most often skipped.

Two hundred and fifty-four works

Every entry names what it rests on, and the reference list holds all of them. Where a law was coined in a blog post or a letter, the post or the letter is cited, because that is the evidence there is.

Every entry in this guide is indexed, with its definition, at https://www.geraldajam.com/library/glossary/field-guides/working-defaults-field-guide.