Data Contract
The Data Contract is Licklider's mechanism for eliminating analytical ambiguity before the first test runs. By declaring what each observation represents, how variables are structured, and what outcome type you are working with, you establish a shared ground truth that governs every downstream guard, suggestion, and output.
What is the Data Contract?
Most statistical errors are not calculation errors — they are assumption errors. The wrong unit of analysis (pseudoreplication), an unmapped subject ID (missed paired structure), or an undeclared outcome type (OLS on proportions) each produce results that appear statistically valid but are analytically incorrect.
The Data Contract intercepts these errors before they propagate. It is a structured declaration — completed interactively at the start of each analysis session — that locks the assumptions Licklider uses to validate every subsequent decision.
The Data Contract cannot be changed after analysis begins. Once declared and locked, the observation unit, variable mapping, and outcome type remain fixed for the session. If a correction is needed, a new session must be started. This design is intentional: post-hoc declaration of analytical assumptions eliminates the reproducibility guarantee that the Contract is meant to provide.