The supposedly metaphysical nature of causality, according to such radical empiricists as Hume, stems from the fact that there cannot be a direct perception of a causal connection between events: perception of causes and effects is indistinguishable from perception of unrelated events that happen to be contiguous in space and time. This non-commitmental position is, however, untenable, as causality is absolutely central to any theory admitting the existence of an objective world: it is only by postulating a cause-effect relation between external events and our internal perceptions of them that we can escape solipsism.
It is my take then that causality, at least within the context of the human perception of events, can be seen as a primitive concept at the same level as event occurrence; saying "A causes B" is as basic a statement as "It is the case that C", because causality can only be explained in terms of event occurrence, and any account of an event occurrence depends on the implied causality connecting events and their perception. Formal renditions of causality, however, propose modal-like semantics denying this concept its primitiveness, which adds to the methaphysical aura sourrounding the subject: while the truth or falsity of a statement in a logical calculus is directly represented by the models of the calculus (cf. first-order predicate calculus semantics), causality semantics rely on more complex structures (vg. possible worlds) that build on "basic" models.
What would a formalization look like in which the truth of a statement and the causal connection between two statements are modeled at the same basic level? We will explore this approach in a later entry.