Epistemic leakage is a term coined by Bonnitta Roy to describe the way a theoretical model or framework unintentionally "leaks" its underlying worldview and assumptions, regardless of how much the creator tries to nuance or "hedge" the presentation,. Even if two models look at the same data and can be translated into one another, they may still differ in the specific worldview they project.

Key aspects of epistemic leakage include:

In summary, epistemic leakage suggests that you cannot hide the foundational assumptions of a model; the core logic will always reveal itself in the way the model characterizes the world.

References