Exponential-growth underestimation: Semantic framing and information availability

Exponential growth is consistently underestimated. Although this tendency reflects difficulty in processing nonlinear growth, its magnitude may also be sensitive to semantic context and heuristic influences. Across three preregistered experiments (total N = 2836), participants estimated mathematically equivalent exponential-growth problems framed as representing coronavirus infections or recoveries. Across experiments, participants' estimates were closer to the correct answers when the numerical progressions represented coronavirus infections rather than recoveries (i.e., they showed less underestimation). Coronavirus-related emotions, attitudes, and behaviors measured in the present research did not provide a consistent account of this difference. In Experiment 2, the interaction between context and framing was not significant. Nevertheless, planned follow-up analyses showed significantly less underestimation for infections than recoveries, whereas estimates did not significantly differ when mathematically equivalent numerical progressions represented personal financial losses rather than gains, a pattern consistent with a possible additional contribution of the coronavirus context. In Experiment 3, follow-up analyses showed a significant infection-recovery difference among participants with high exposure to liberal media and low exposure to conservative media, but not under the other combinations of media exposure. This conditional pattern was consistent with differential availability of infection-related information as a possible cognitive mechanism. These findings indicate that exponential-growth underestimation is sensitive to the semantic framing of mathematically equivalent problems and to context-relevant information, supporting its characterization as the outcome of a judgment process susceptible to contextual influences, heuristics, and biases. The limitations of the present evidence and implications for communicating exponential-growth information are discussed.