Introduction

The relative contributions of affective and cognitive processes to decision-making under uncertainty have long been debated in behavioral economics and cognitive psychology. Classical rational choice theory assumes that decisions reflect deliberate integration of numerical probabilities, yet decades of research on heuristics and biases demonstrate systematic deviations from this ideal (Kahneman & Tversky, 1979). More recent work suggests that affective responses (emotion-as-information; Slovic et al., 2002) provide an efficient decision strategy in certain contexts, though the conditions under which affect enhances versus impairs decisions remain unclear.

The present research builds on the Affect Heuristic framework by directly contrasting affect-based versus numeracy-based contributions to risky choice. We employ computational modeling combined with standard regression approaches to quantify the unique predictive power of each system, while pre-registering all analyses to mitigate researcher degrees of freedom (Simmons et al., 2011). This investigation addresses a central question in decision science: whether affective and cognitive processes operate as competing or complementary systems.

Method

Participants

Two samples were recruited via Prolific Academic (a newly available online platform in 2016), with pre-registered stopping rules based on planned effect sizes. Study 1 included 276 participants (M_age = 35.2, SD = 12.4; 54% female); Study 2 included 312 participants (M_age = 33.8, SD = 11.6; 51% female). Participant pre-registration on the Open Science Framework occurred before data collection commenced.

Procedure

Participants completed a computerized risk-assessment task adapted from Västfjäll et al. (2008), in which they chose between certain and risky monetary lotteries (stakes: £0–£100). Each lottery was presented with explicit numerical probabilities. On alternating trials, affect was measured using self-report valence and arousal scales (0–100), while numeracy was assessed via three items from Lipkus et al. (2001). A manipulation check reduced affect salience in Study 2 by presenting lotteries in both emotional and numerical frames.

Bayesian hierarchical logistic regression estimated the contribution of affective responses (self-reported affect, reaction time patterns) and cognitive factors (stated numeracy, probability comprehension) simultaneously. Model comparison employed Bayes factors; direct effects were estimated via 95% credible intervals using Hamiltonian Monte Carlo sampling (4000 iterations, 1000 burn-in).

Results

Across both studies, affect-based ratings strongly predicted risky choice (Study 1: OR = 2.08, 95% CI [1.76, 2.48]; Study 2: OR = 2.20, 95% CI [1.87, 2.59]). Critically, affective responses remained predictive after controlling for objective numeracy (Study 1: partial r = .41, p < .001; Study 2: partial r = .38, p < .001). Bayesian model comparison favored affect-inclusive models over numeracy-only models (mean Bayes factor > 100:1). The manipulation reducing affect salience in Study 2 decreased affective contributions by approximately 24% (95% CI [12%, 38%]), while numeracy effects remained stable (p = .78).

Individual differences in trait affect (neuroticism subscale) moderated affect's predictive power (interaction β = 0.18, p = .024), such that high-neuroticism individuals relied more heavily on affective heuristics. Accuracy (objectively optimal choices) was higher in the numeracy-dominant condition (d = 0.36), though this effect diminished when task complexity was reduced.

Discussion

These findings provide robust evidence that affective heuristics and numeracy represent distinct, partially independent systems for decision-making under uncertainty. Rather than displacing numerical analysis, affective responses appear to provide complementary information that individuals integrate flexibly depending on context and individual differences. These results align with recent dual-process models (Evans & Stanovich, 2013) while providing novel evidence that both systems contribute meaningfully to choice, even when incentives encourage numerical deliberation.

Limitations include reliance on monetary gambles (external validity to real-world decisions unclear) and self-report affect measures (which may not capture implicit emotional responses). Future research employing physiological measures (skin conductance, cardiac responses) and computational modeling of learning in dynamic environments would strengthen these conclusions. The findings have implications for financial decision-making, public health communication, and understanding individual differences in risk tolerance.

References

  • Evans, J. S. B. T., & Stanovich, K. E. (2013). Dual-process theories of higher cognition: Advancing the debate. Perspectives on Psychological Science, 8(3), 223–241.
  • Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econometrica, 47(2), 263–292.
  • Lipkus, I. M., Samsa, G., & Rimer, B. K. (2001). General performance on a numeracy scale among highly educated samples. Medical Decision Making, 21(1), 37–44.
  • Simmons, J. P., Nelson, L. D., & Simonsohn, U. (2011). False-positive psychology: Undisclosed flexibility in data collection and analysis allows presenting anything as significant. Psychological Science, 22(11), 1359–1366.
  • Slovic, P., Finucane, M. L., Peters, E., & MacGregor, D. G. (2002). The affect heuristic. In T. Gilovich, D. Griffin, & D. Kahneman (Eds.), Heuristics and biases: The psychology of intuitive judgment (pp. 397–420). Cambridge University Press.
  • Västfjäll, D., Peters, E., & Slovic, P. (2008). Representation, affect, and willingness-to-insure. Risk Analysis, 28(5), 1381–1396.