feat(library/logic/examples/negative): add example showing that allowing negative inductive datatypes would lead to inconsistency

library/logic/examples/negative.lean

@@ -0,0 +1,30 @@
+/-
+This example demonstrates why allowing types such as
+
+inductive D : Type :=
+| intro : (D → D) → D
+
+would make the system inconsistent
+-/
+
+/- If we were allowed to form the inductive type
+
+     inductive D : Type :=
+     | intro : (D → D) → D
+
+   we would get the following
+-/
+universe l
+-- The new type A
+axiom D : Type.{l}
+-- The constructor
+axiom introD : (D → D) → D
+-- The eliminator
+axiom recD   : Π {C : D → Type}, (Π (f : D → D) (r : Π d, C (f d)), C (introD f)) → (Π (d : D), C d)
+-- We would also get a computational rule for the eliminator, but we don't need it for deriving the inconsistency.
+
+definition id : D → D := λd, d
+definition v  : D     := introD id
+
+theorem inconsistent : false :=
+recD (λ f ih, ih v) v