Linear Biases and Pandemic Communications.

Abstract

Previous research has demonstrated a tendency for individuals to mentally linearize nonlinear trends, leading to forecast errors. The present research notes that prior conceptualizations of these linear biases do not make identical predictions and examines how linear biases affect forecasts and risk perceptions of an unfolding epidemic. This research uses an online experiment and a preregistered direct replication in a different online participant pool (total N = 608) to assess the trajectories of forecasts and risk perceptions over time in an unfolding epidemic. Framing the progress of the epidemic using total cases (v. the rate of new cases) leads to higher forecasts. This research also finds that the effect of frame varies over different time points in the epidemic and differs for forecasts versus risk perceptions. Finally, the effect of frame for forecasted totals is weaker among more numerate individuals. The studies use repeated measures that occur in 1 session rather than over the course of several months and involve a smooth epidemic curve rather than a noisy one with jagged case counts. This research compares prior conceptualizations of linear biases and yields data with implications both for theory on linear biases and for communicators involved in disseminating information about epidemics. Framing the progress of the epidemic using total cases (v. the rate of new cases) leads to higher forecasts.The effect of frame varies over different time points in the epidemic and differs for forecasts v. risk perceptions.The effect of frame for forecasted totals is weaker among more numerate individuals.