The limited principle of omniscience states (Bridges & Richman 1987, p. 3):

LPO: For any sequence ${\displaystyle a_{0}}$ , ${\displaystyle a_{1}}$ , ... such that each ${\displaystyle a_{i}}$  is either ${\displaystyle 0}$  or ${\displaystyle 1}$ , the following holds: either ${\displaystyle a_{i}=0}$  for all ${\displaystyle i}$ , or there is a ${\displaystyle k}$  with ${\displaystyle a_{k}=1}$ . ^[1]

The second disjunct can be expressed as ${\displaystyle \exists k.a_{k}\neq 0}$  and is constructively stronger than the negation of the first, ${\displaystyle \neg \forall k.a_{k}=0}$ . The weak schema in which the former is replaced with the latter is called WLPO and represents particular instances of excluded middle.^[2]

The lesser limited principle of omniscience states:

LLPO: For any sequence ${\displaystyle a_{0}}$ , ${\displaystyle a_{1}}$ , ... such that each ${\displaystyle a_{i}}$  is either ${\displaystyle 0}$  or ${\displaystyle 1}$ , and such that at most one ${\displaystyle a_{i}}$  is nonzero, the following holds: either ${\displaystyle a_{2i}=0}$  for all ${\displaystyle i}$ , or ${\displaystyle a_{2i+1}=0}$  for all ${\displaystyle i}$ .

Here ${\displaystyle a_{2i}}$  and ${\displaystyle a_{2i+1}}$  are entries with even and odd index respectively.

It can be proved constructively that the law of the excluded middle implies LPO, and LPO implies LLPO. However, none of these implications can be reversed in typical systems of constructive mathematics.

Terminology edit

The term "omniscience" comes from a thought experiment regarding how a mathematician might tell which of the two cases in the conclusion of LPO holds for a given sequence ${\displaystyle (a_{i})}$ . Answering the question "is there a ${\displaystyle k}$  with ${\displaystyle a_{k}=1}$ ?" negatively, assuming the answer is negative, seems to require surveying the entire sequence. Because this would require the examination of infinitely many terms, the axiom stating it is possible to make this determination was dubbed an "omniscience principle" by Bishop (1967).

Logical versions edit

The two principles can be expressed as purely logical principles, by casting it in terms of decidable predicates on the naturals. I.e. ${\displaystyle P}$  for which ${\displaystyle \forall n.P(n)\lor \neg P(n)}$  does hold.

The lesser principle corresponds to a predicate version of that De Morgan's law that does not hold intuitionistically, i.e. the distributivity of negation of a conjunction.

Analytic versions edit

Both principles have analogous properties in the theory of real numbers. The analytic LPO states that every real number satisfies the trichotomy ${\displaystyle x<0}$  or ${\displaystyle x=0}$  or ${\displaystyle x>0}$  . The analytic LLPO states that every real number satisfies the dichotomy ${\displaystyle x\geq 0}$  or ${\displaystyle x\leq 0}$ , while the analytic Markov's principle states that if ${\displaystyle x\leq 0}$  is false, then ${\displaystyle x>0}$ .

All three analytic principles if assumed to hold for the Dedekind or Cauchy real numbers imply their arithmetic versions, while the converse is true if we assume (weak) countable choice, as shown in Bishop (1967).

• Constructive analysis

• Bishop, Errett (1967). Foundations of Constructive Analysis. ISBN 4-87187-714-0.
• Bridges, Douglas; Richman, Fred (1987). Varieties of Constructive Mathematics. ISBN 0-521-31802-5.

• "Constructive Mathematics" entry by Douglas Bridges in the Stanford Encyclopedia of Philosophy