Results for "spurious correlation"
Spurious Correlation
IntermediateModel relies on irrelevant signals.
Spurious correlation is like thinking that carrying an umbrella causes it to rain just because you see people with umbrellas when it rains. In AI, this happens when a model makes predictions based on patterns that aren't really connected to the outcome. For example, if a model learns that certain...
Model relies on irrelevant signals.
Normalized covariance.
A hidden variable influences both cause and effect, biasing naive estimates of causal impact.
Framework for reasoning about cause-effect relationships beyond correlation, often using structural assumptions and experiments.
A broader capability to infer internal system state from telemetry, crucial for AI services and agents.
Quantifying financial risk.
Models effects of interventions (do(X=x)).
Measures a model’s ability to fit random noise; used to bound generalization error.