Causality Establishment vs. Debunking
In Causal Persuasion , the authors argue that ruling out a causal link is harder than establishing one because causal identification requires eliminating all competing explanations —whereas establishing causality can ...
In Causal Persuasion , the authors argue that ruling out a causal link is harder than establishing one because causal identification requires eliminating all competing explanations —whereas establishing causality can often be done once the evidence rules out “enough” alternatives. For establishing a true causal direction, the sender can often succeed by making the receiver’s data conclusively identify a particular causal link, and the paper states that “To establish a true causal link, the sender often needs to disclose only one or two well chosen variables.” [:cite[1]{ln=5}] This is tied to their result that identifying causal structure can be made easy when the right variables are disclosed (e.g., by creating v structures and using their causal discovery logic).[:cite[2]{ln=8}] For debunking a perceived causal relationship (i.e., persuading the receiver that “no such link exists”), the paper emphasizes the asymmetry: dispelling a perceived link requires ruling out every common cause pathway between the two variables. They explicitly state: “to dispel a perceived link … every common cause must be disclosed” [:cite[1]{ln=6}] and “accounting for every source of correlation between them.” [:cite[3]{ln=4}] This is generally burdensome because there may be arbitrarily many confounders .[:cite[3]{ln=6}] More formally in their model, they highlight that for sophisticated receivers and debunking, the sender may need to disclose many variables: “Persuading a sophisticated receiver … requires revealing all variables in {c1, ..., cn}” [:cite[4]{ln=1}] and since n is arbitrary , this means arbitrarily many variables may be required.[:cite[4]{ln=3}] The paper connects this directly to the underlying logic of negating a causal claim: “unlike universal claims…, ruling out a causal relationship requires eliminating all competing explanations.” [:cite[5]{ln=3}] So, the core reason is: causality claims can often be supported after ruling out a small set of alternatives (via selective disclosure), but debunking requires exhaustively ruling out all confounding/common cause explanations—often forcing disclosure of many variables. [:cite[1]{ln=5}], [:cite[1]{ln=6}], [:cite[3]{ln=4}], [:cite[4]{ln=3}], [:cite[5]{ln=3}]