# Logical Consistency and Objectivity in Causal Learning

- Patricia Cheng,
*UCLA*
- Mimi Liljeholm,
*Cal Tech*
- Catherine Sandhofer,
*UCLA*

## Abstract

Logical consistency and objectivity are cornerstones of science
that distinguish it from cult and dogma. Scientists’ concern with
objectivity has led to the dominance of associative statistics, which define the
basic concept of independence on observations. The same concern with avoiding
subjective beliefs has led many scientific journals to favor frequentist over
Bayesian statistics. Our analysis here reveals that to infer causes of a binary
outcome, (1) the associative definition of independence results in a logical
inconsistency -- even for data from an ideal experiment -- for both frequentist
and Bayesian statistics, and (2) removing the logical error requires defining
independence on counterfactual causal events. The logically coherent causal
definition is the one intuitively adopted by humans. Our findings have direct
implications for more consistent and generalizable causal discoveries in medicine
and other sciences.

Back to Table of Contents