skip to content

What does it mean operationally that an alert condition is a tautology rather than merely satisfiable?

level: seniorimportance: nice to knowfreq 30%

answer

  1. count the rows where it holds
  2. all rows, some rows, no rows
  3. the silent rule is the dangerous one
  4. satisfiable is the weaker property
  5. negation swaps the two extremes

basics

~20 s

A tautology is true under every assignment of its inputs, so the alert fires unconditionally and carries no information. Satisfiable only means some assignment makes it true - the ordinary, useful case. A condition true under no assignment is a contradiction and can never fire.

solid answer

~40 s

The three labels describe how many rows of the truth table a condition is true on. A **tautology** is true on all of them: whatever the signals do, the condition holds, so the alert fires always and distinguishes nothing. A **contradiction** is true on none: no combination of signals can trigger it, so the rule is dead and its silence looks exactly like health. **Satisfiable** means true on at least one row, which every useful condition is - note that every tautology is satisfiable, so 'satisfiable' alone is a weak property. The failure mode worth catching is a condition that has drifted into one of the extremes through accumulated edits: a disjunction that grew until it covered every case, or a conjunction with two mutually exclusive terms.

go deeper

for a junior

Hold the three labels by row count: true on every row is a tautology, on no row a contradiction, on at least one row satisfiable. Every tautology is also satisfiable.

for a middle

Explain the mirror relationship - a condition is a tautology exactly when its negation is a contradiction - and recognise the complementary pair that produces each extreme by accident.

for a senior

Argue which extreme hurts more in production: a silent rule reads as health, so a rule that has not fired since an edit deserves the same scrutiny as one that fires constantly.

for a principal

Treat rule drift as a process problem. Conditions grown clause by clause by different authors need a review step that checks the extremes and the nearly vacuous middle, not just that the syntax parses.

Tautology, contradiction and satisfiability are not three unrelated words - they are three answers to one question: **on how many rows of its truth table is this condition true?** ## The three labels | label | true on | operational reading of an alert | |---|---|---| | tautology (valid) | every assignment | fires whatever happens; carries no information | | satisfiable | at least one assignment | can fire; the normal case | | contradiction (unsatisfiable) | no assignment | can never fire; permanently silent | Two relationships between them are worth stating precisely, because they are commonly stated backwards: - Every tautology is satisfiable, but most satisfiable conditions are not tautologies. Satisfiability is the weaker property. - A condition is a tautology exactly when its negation is a contradiction, and a contradiction exactly when its negation is a tautology. The two extremes are mirror images across negation. ## Why each extreme is a defect, and why they look different in production A **tautological** condition is loud. It fires on every evaluation, so it is noticed quickly - but the damage is that its firing is uninformative, and after a week of firing, the notification is muted or filtered, taking whatever genuine signal shared that channel with it. The usual way a condition becomes tautological is growth by disjunction: each incident adds another `or ...` clause until the clauses between them cover every row, frequently because one added clause is the negation of an earlier one. A **contradictory** condition is silent, which is far worse, because silence is indistinguishable from health. Nobody investigates a rule that has not fired; its absence is read as good news. Conjunctions drift this way: two terms that cannot hold at once - a state required to be both draining and accepting new work - are added by different authors at different times, and the rule dies without any error. ## How the extremes get diagnosed 1. **Build the table over the distinct propositions.** If the result column is all true, it is a tautology; all false, a contradiction. This is exact for a propositional condition, and the row count `2^n` is manageable for the handful of signals a rule typically uses. 2. **Look for a complementary pair.** A disjunction containing both a proposition and its negation is true on every row, whatever else is in it. A conjunction containing both is false on every row, whatever else is in it. These two patterns explain most real cases. 3. **Check the history of the fire count.** A rule that fires on every evaluation, or has never fired since a particular edit, is worth putting on the table regardless of how it reads. ## Vacuity is the version of this that hides The subtle case is not the fully tautological condition but the nearly vacuous one: a condition whose extra clauses only exclude rows that can never occur anyway. It is satisfiable and it does fire, but it fires on exactly the same rows it would have without half of its logic - so the logic is decoration, and everyone believes the rule is narrower than it is. The table finds it: compare the condition against the simpler condition someone believes it is, row by row, and see whether the columns differ at all. ## Where the boundary of this material is Deciding whether a condition is satisfiable is easy to *describe* - check every row - and the number of rows doubles with each added proposition, so hand enumeration stops being practical around four or five signals. How expensive that decision is in general, and what class of problem it belongs to, is a question about computational complexity rather than about truth tables, and it is answered elsewhere. For reviewing an alert rule, the practical points are the three labels, their mirror-image relationship across negation, and the two syntactic patterns - complementary pair in a disjunction, complementary pair in a conjunction - that produce the extremes by accident.

  • Which extreme is more dangerous in a monitoring system, and why?
    The contradiction. A tautological rule fires constantly and gets noticed, even if the fix is often to mute it. A contradictory rule is silent, and silence is read as health - nobody investigates a rule that has not fired, so the gap can sit unnoticed for as long as the condition survives.
  • What syntactic pattern most often turns a grown condition into a tautology?
    A disjunction that ends up containing both a proposition and its negation. Once both are present the disjunction is true on every row whatever the other clauses say. The mirror pattern - a proposition and its negation inside one conjunction - makes the condition a contradiction, true on no row at all.
  • A condition fires and is satisfiable, but the reviewer still calls it vacuous. What are they seeing?
    Clauses that exclude only rows that cannot occur, so the condition fires on exactly the rows it would fire on without them. It is narrower on paper than in behaviour. Comparing it row by row against the simpler condition people believe it to be shows the two columns are identical.

saying these in an interview costs you the question

  • Calls any condition that can fire a tautology
  • Thinks satisfiable means true under every assignment
  • Treats a rule that never fires as evidence of a healthy system
  • Says a tautology is harmless because it is never wrong
  • Believes a tautology and a contradiction are unrelated properties