Computational narratology is concerned with extracting, generating, and
reconfiguring stories via algorithms. A core problem in the field is
narrative representation, i.e. with what formal elements should we
represent stories and story generators?

Linear logic is a partial answer, in which a story may be equated to a
proof of a logical sequent.  Different linear logic connectives can be used
to model various *nonlinear* aspects of stories, specifically branching
alternatives (as in interactive stories) and simultaneity (as in stories
that support telling out of chronological structure because their events
are only partially ordered).

However, this version of narrative theory is limited to the discussion of
plausible and coherent event structures, and several aspects of "narrative"
itself have yet to be addressed. We sketch two lines of ongoing work to
address different aspects of this problem.

First, the structure of a proof in linear logic maps onto the event
structure, or "fabula," of the story.  The counterpart to the fabula, the
"sjuzet," or conveying of a story to an audience, has had relatively little
attention paid to its formal modeling.  We posit that linear logic proofs
may offer support for sjuzet modeling via an account of narrative
focalization (character perspective) as a transformation on proofs.  

Second, whereas standard proof search strategies may give rise to
degenerate stories, a multi-agent, intent-driven theory of proof
construction (inspired by the Belief-Desire-Intention framework and
intent-driven planning) may be able to restrict the set of proofs to those
corresponding to stories with conflict and motivated character actions.