# Credit Assignment **Domain:** Machine Learning / Historiography / Causal Attribution **Doc Type:** Canonical Concept Node **Maturity:** Foundational ## Definition In machine learning, **credit assignment** determines which earlier parameters, actions or components contributed to a later outcome so that responsibility for error or success can be propagated backward. ## Technical Context [[wiki/Backpropagation|Backpropagation]] performs credit assignment by using [[wiki/Reverse-Mode Automatic Differentiation|reverse-mode differentiation]] to calculate how earlier operations affected an objective. ## Historiographic Context [[articles/Modern Artificial Intelligence in the 1970s|Modern Artificial Intelligence in the 1970s]] applies the same problem to history. Mature technical systems often emerge from distributed and independently convergent lineages, yet retrospective narratives concentrate credit in the most visible later layer. ## Key Insight **Credit can be distributed across a causal composition even when a narrative prefers one inventor.** ## See Also [[wiki/Jürgen Schmidhuber|Jürgen Schmidhuber]], [[wiki/Provenance|Provenance]], [[wiki/Distributed Representation|Distributed Representation]]