When browsing data, it should be possible to check how it was generated (which algorithm run it was produced by, etc.) and what it has been used for (which algorithm run it has served as input for). And from there on browse through the whole provenance tree.