Predictive coding, storytelling and god: narrative understanding in e-discovery

The potential for gains that predictive coding can offer to the e-discovery review process lies in the power that it has to generalise human relevance decisions from a relatively small set of exemplars to a larger collection. These gains, however, depend as much on the theory of relevance developed by the
human handler, and the review team in general, as on that developed by the machine. After differentiating human and machine theories of relevance and discussing the way in which these symbiotically co-evolve during predictive coding iterations, we examine the human theory in more detail. We draw on sources from philosophy, psychology and practice to make the case for the centrality of narrative, or story-telling, in constructing and anchoring the human theory of relevance. We then use this to inform a series of guidelines for structuring predictive coding based review around the notion of narrative. These include: narrative frame-working; differentiating sub-plots and episodes; designating a chief storyteller; using external representations; and narrative sharing that leaves no reviewer behind.

DESI V Workshop was held after ICAIL 2013 Conference.