An Experimental Study of Decision Rules in General vs. Concrete Insurance Problems

Abstract: Experimental work typically studies how individuals respond to specific, concrete decision problems. Yet many real-world choices are made at a more general level: decision makers often step back to determine how they would act across related situations with varying parameters. Such general strategies also arise in delegation and automation, where people provide guidelines for agents or algorithms to apply broadly. We introduce an incentive-compatible method for eliciting free-text decision rules for general decision problems. Participants describe a rule to an automated system that applies it across concrete cases. A between-subjects design compares choices in a General treatment with those in a Case-by-case treatment, where participants decide manually and then write rules intended to replicate their choices. In an online experiment, participants made 20 incentivized binary insurance choices with varying loss probabilities and premiums. Insurance demand was lower in the General treatment. When participants later reviewed the concrete choices their rules implied, about half left all of them unchanged, indicating broad endorsement of rule-based decision making. Revisions slightly increased insurance uptake, which nonetheless remained below the Case-by-case rate. These findings show that decision mode (abstract versus concrete) shapes behavior in ways consistent with construal-level theory. The method also reveals the procedural rules people adopt and provides a versatile tool for studying how individuals structure choices across contexts. Joint work with Rom Severin