Jean Krivine 

CNRS/Université Paris-Diderot

Title: Non local causality analysis in rule-based formalisms

Abstract: In classical approaches, causality between computation events
is a local property that is dual to the notion of concurrency: two
consecutive events are in causal relationship if and only if they cannot
commute. A trace is said in causal form for a particular event e, if all
event e' that precedes e is transitively a cause of e. In biological
systems however this notion turns out to be unsatisfactory because it
does not take into account the facts that:
- although an event e may formally not commute with a preceding event
e', it might be able to commute to a block of events containing e' and
hence occur earlier in the trace.
-  in order to relate to a biological intuition, it is desirable that
the notion of causality should respect a "knock-out" property, namely
that whenever an event e is (transitively) a cause of an event e', then
preventing e from occurring -by knocking out the rule that generated it-
should also prevent e' from occurring.

In this talk we will present and discuss non standard causal
compressions of interleaving traces obtained from stochastic runs of
models written in the kappa calculus.