Info

Unification  

For each of the following pairs of terms determine whether they can be unified. If so, provide the unifier:

1. LiesBetween Munich y z and LiesBetween x Milan Rome
2. SitsBeween Mary x x and SitsBetween x Jane y
3. Between x Rome Rome and Between Rome y x
4. BornIn fatherOf motherOf x London and BornIn fatherOf y x
5. Human x and Human fatherOf x
6. Human fatherOf y x and Human x fatherOf y

Robinson’s Algorithm  

The first algorithms for unification is due to John Alan Robinson , the father of resolution.

To check whether two FOL literals can be unified, the algorithm proceeds as follows:

1. Check if either both formulas begin with a negation or neither does. If so, proceed. Otherwise, the formulas are not unifiable.

2. Check if the predicate in both formulas is the same. If it is, continue. If it is not, the formulas are not unifiable.

3. We can assume, at this point, that the form of the formulas is covered by one of the two cases:

   • Pⁿ s₁ … sₙ and Pⁿ t₁ … tₙ

   • Pⁿ s₁ … sₙ and Pⁿ t₁ … tₙ

   In both cases, to unify the formulas, we need to find a substitution σ, such that s₁σ = t₁σ, …, sₙσ = tₙσ, that is the result of substituting within these terms must make each pair identical.

   To express this requirement, we make a list of these pairs:

   Eq = [ [s₁, t₁], ..., [sₙ, tₙ] ]

   The algorithm aims to step-wise construct the desired substitution σ. We start with the empty substitution σ = [ ], and go through each pair [sᵢ, tᵢ] of our list. We distinguish the following cases:

   • Case 1: sᵢ = tᵢ. The terms are already identical, we can remove them from Eq.

   • Case 2. sᵢ ≠ tᵢ and sᵢ is some variable, while tᵢ is not. First, check if sᵢ occurs in tᵢ. If so, like when sᵢ = x and tᵢ = fatherOf x, then we can stop the entire procedure since no unification is possible. Otherwise, we apply the substitution [sᵢ / tᵢ] to all other pairs, add it to σ, and remove the pair [sᵢ, tᵢ] from the set.

   • Case 3. sᵢ ≠ tᵢ and tᵢ is some variable, while sᵢ is not. Again, we first check if tᵢ occurs within sᵢ and terminate the entire algorithm if it does. Otherwise, we apply the substitution [tᵢ / sᵢ] to all other pairs, add it to σ, and remove the pair [sᵢ, tᵢ] from the set.

   • Case 4. sᵢ ≠ tᵢ and neither sᵢ nor tᵢ is a variable. Then we check the form of sᵢ and tᵢ. Only if they are of the following forms, we continue:

   sᵢ = fᵐ s₁' … sₘ'   and   tᵢ = fᵐ t₁' … tₘ'

   That is, both terms are the result of applying the same function term to a sequence of terms. If they are not, unification is impossible and we can terminate the algorithm. If they are of this form, we add all the following corresponding pairs to Eq:

   [ [s₁', s₂'], ..., [sₙ', tₙ'] ]

4. By going through all the pairs, deleting pairs when substitutions are possible or trivial (Cases 1—3), and recursively adding new pairs (Case 4), one of two things will happen:

   • Either we hit a termination condition in one of the cases and conclude that unification isn’t possible.

   • Or we end up with an empty list and a full list of substitutions σ to apply. In that case, σ is a unifier and the terms and thus literals are indeed unifiable.

Apply this algorithm to check your work in Exercise 1.

Skolemization  

Skolemize the following formulas:

1. ( y₁ Human y₁ x₁ Mortal x₁)

2. x₁( x₂ y₁(IsFriendOf x₁ y₁ IsFriendOf x₂ y₁) y₂ IsFriendOf x₁ y₂)

3. y₁ y₂ IsFriendOf y₁ y₂

Drinker Paradox  

Consider the following inference:

x InPub x x (InPub x (IsDrinking x x(InPub x IsDrinking x)))

In natural language: There’s somebody in the pub, so there’s somebody in the pub, such that if they are drinking, then everybody in the pub is drinking. This is known as the drinker paradox .

Use the method of resolution to show that this inference is deductively valid in FOL. That is:

1. Form the set of the premise and negation of conclusion.

2. Transform all formulas into equisatisfiable CNF. Note that the Skolemization of an existential that doesn’t depend on any universals is just a constant skolem. It’s important to document your work, which transformations you’re applying, but you can apply several steps simultaneously.

3. Apply resolution with unification to derive the empty sequent { }. And conclude that the initial set is unsatisfiable and the inference thus valid.

Natural deduction  

Find logical proofs in FOL natural deduction for the following logical laws. Note that some of these require open Classical.

Lean  

Verify your work from the previous exercise in Lean. Here’s a template for the work. You can follow the link below to work in the interactive environment.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43

variable (Term : Type) (A B : Term → Prop) (R : Term → Term → Prop) (C : Prop)

/-! Duality Laws -/

theorem duality_one_ltr (h : ¬∀x, A x) : ∃x, ¬A x := by
  sorry

theorem duality_one_rtl (h : ∃x, ¬A x) : ¬∀x, A x := by
  sorry

theorem duality_two_ltr (h : ¬∃x, A x) : ∀x, ¬A x := by
  sorry

theorem duality_two_rtl (h : ∀x, ¬A x) : ¬∃x, A x := by
  sorry

/-! Duality Laws -/

theorem all_over_and_rtl (h: ∀x, A x ∧ B x) : (∀x, A x) ∧ (∀x, B x) := by 
  sorry

theorem all_over_and_ltr (h: (∀x, A x) ∧ (∀x, B x)) : ∀x, A x ∧  B x := by
  sorry

theorem exists_over_or_rtl (h: ∃x, A x ∨ B x) : (∃x, A x) ∨ (∃x, B x) := by
  sorry

theorem exists_over_or_ltr (h: (∃x, A x) ∨ (∃x, B x)) : ∃x, A x ∨ B x := by
  sorry

/-! Interaction Laws -/

theorem existential_import (h: ∀x, A x) : ∃x, A x := by
  sorry

theorem switcheroo (h : ∃x, ∀y, R x y) : ∀y, ∃x, R x y := by
  sorry

theorem exists_to_forall (h: (∃x, A x) → C) : ∀x, A x → C := by
  sorry

theorem forall_to_exists (h: (∀x, A x) → C) : ∃x, A x → C := by
  sorry

Follow this link to work in the digital playground.