From Causal Models to Sound Heuristic Inference

Abstract

We investigate whether people rely on their causal intuitions to determine the predictive value or importance of cues. Our real-world data set consists of one criterion variable (child mortality) and nine cues (e.g., GDP per capita). We elicited people’s intuitive causal models about the domain. In a second task, we asked them to rank the cues according to their beliefs about the cues ’ predictive value. Alternative cue importance rankings were derived directly from their causal models using measures of causal centrality. The results show that people’s judgments of cue importance corresponded more closely to the causal-based cue orders than to the statistical associations between the cues and the criterion. Using computer simulations, we show that people’s causal-based cue orders form a sound basis for making inferences, even when information about the statistical structure of the environment is scarce or unavailable. Central to the simulations is take-the-best (TTB)—a simple decision strategy that makes inferences by considering cues sequentially. The simulations show that causal-based cue orders can be as accurate as individuals ’ judged orders. Causal-based cue orders allow TTB to perform as would be expected from estimating the weights of a linear model using about 35 % of the available data. These findings suggest that people can rely on their causal intuitions to determine the importance of cues, thereby reducing the computational complexity involved in finding useful cue orders.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 2 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.