Which propositional logics do you get by taking the class of Boolean algebras and adding "weak negation"?

57 Views Asked by At

Which propositional logics do you get by taking the class of Boolean algebras as a semantics and adding strong negation?

I've encountered weak negation in unrelated contexts before as a unary connective that exchanges designated and undesignated truth values, so I'm curious what happens if you staple it to a well-known algebraic semantics.


Let $\mathsf{I}(\varphi)$ denote the interpretation of $\varphi$. Let $X$ be a set and $\Lambda$ be the partially ordered set $(2^X, \subset)$.

  • $ \mathsf{I}(a)$ is the interpretation of $a$ when $a$ is a propositional variable.
  • $ \mathsf{I}(a \land b)$ is $\mathsf{I}(a) \cap \mathsf{I}(b)$.
  • $ \mathsf{I}(a \lor b)$ is $\mathsf{I}(a) \cup \mathsf{I}(b)$.
  • $\mathsf{I}(\lnot a)$ is $X \setminus \mathsf{I}(a)$. This is ordinary negation.
  • $\mathsf{I}(!a)$ is $\varnothing$ if $\mathsf{I}(a)$ is $X$, otherwise $\mathsf{I}(a)$ is $X$. This is weak negation. It exchanges designated and undesignated truth values.

If $X$ is $\{e\}$, then this semantics reduces to classical logic and $\lnot$ and $!$ are indistinguishable.

By a theorem cited in this answer and discussed more generally in this article, the semantics obtained using the class of Boolean algebras is the same as the one obtained from using $\{\varnothing, \{e\}\}$ alone.

However, adding weak negation essentially on a whim is enough to make the resulting semantics non-classical.

The following is a theorem of classical logic because the two connectives have the same meaning in that setting, and a theorem in the Boolean algebra semantics because $\mathsf{I}(a)$ must be $\varnothing$ when $\mathsf{\lnot a}$ is $X$.

$$ \frac{\lnot a}{! a} $$

However, the following is a theorem of classical logic only.

$$ \frac{! a}{\lnot a} $$

The following inference rule also fails in the $!$-setting.

$$ \frac{a \lor b \;\;\text{and}\;\; !a}{b} \;\; \text{is not a valid inference rule} $$