Data-driven logic

Reasoning with Data (ReDa) is a logic-first approach to data-driven inference that asks a simple question: when does data justify an inference?
ReDa’s goal is to build sound, usable logical foundations for modern data-driven reasoning—especially where uncertainty, noise, and imperfect information are unavoidable.

Many dominant inference practices (notably “significance”-style inference) are ubiquitous, contested, and often justified by loose analogies to classical logic. ReDa treats this as a logical problem under the assumption that data-driven infernece is typically stochastic and revisable, unlike mathematical proof. So we need a logic that explains when data-driven inference is valid and why—without pretending it is just classical deduction in disguise.

Data-driven, and more generally, scientific conclusions are rarely “proved” in the mathematical sense. Instead, they are:

• stochastic (sensitive to noise, sampling, and error),
• defeasible (revisable when new evidence arrives),
• and frequently based on methods whose logical status is unclear, especially common “significance-style” inference.

ReDa aims to replace analogy, ritual, and rule-of-thumb reasoning with explicit consequence relations—clear accounts of what follows from what, and under which conditions.