emmarshall-slides – Search and Reflect: How Unexpected Outcomes Shape Hindsight Bias and Probable Cause

Warrantless Vehicle Searches

Probable Cause

Reasonable belief that a search would reveal contraband or evidence of a crime

Taslitz, 2010

Hindsight Bias

Overestimating the predictability of a past event after its outcome is known

Casper, Benedict, & Perry, 1989; Robbennolt & Sobus, 1997

Sense-making Process

Roese & Vohs, 2012; Yopchick & Kim, 2012

Methods

Results

More likely to find probable cause after
expected or surprise outcome

Results

Sense-making process

Measurement Model:

relEvidence is a latent variable indicated by four observed variables: rbmw, rcash, ritems, and rnerv. All loadings are highly significant (***) and positive, suggesting strong relationships between each indicator and the latent variable relEvidence. Structural Model:

expected negatively predicts relEvidence, but the link is not statistically significant (p = 0.294). This is indicated by the non-significant path coefficient a. surprise also negatively predicts relEvidence, and this relationship is statistically significant (**), as denoted by the path coefficient d. expected has a non-significant positive direct effect on prob_success shown by cPrime. relEvidence has a positive and statistically significant effect on prob_success with coefficient b. surprise has a non-significant negative effect on prob_success represented by e. Direct and Indirect Effects:

Parameter ab represents the indirect effect of expected on prob_success through relEvidence, which is not significant. Parameter db shows the indirect effect of surprise on prob_success through relEvidence, which is significant (**). Parameter c captures the total effect of expected on prob_success, but it is not significant, indicating that the overall association may not be strong when combined with the effects mediated by relEvidence and the direct effect of surprise. Variances and Covariances:

The variance of each observed variable (rbmw, rcash, ritems, rnerv, and prob_success) is significant, suggesting variability not explained by the model for each of these measurements. expected negatively covaries with surprise (~~), which is suggested by an understood negative relationship between the two. Intercepts:

Each of the observed variables (rbmw, rcash, ritems, rnerv, and prob_success) has a significant positive intercept, meaning that their average value when predictors are zero is positive and significantly different from zero. Global Fit Measures:

The Comparative Fit Index (CFI) and Tucker-Lewis Index (TLI) values are 0.964 and 0.934, respectively, which are close to or surpass the 0.95 threshold, suggesting a good model fit. The Root Mean Square Error of Approximation (RMSEA) is 0.075, which is marginally above the 0.05 threshold for a ‘good’ fit but still within acceptable limits. The Standardized Root Mean Square Residual (SRMR) is 0.045, under the 0.08 threshold, indicating a good fit. Overall, it seems like the model has a good fit with some caveats. While most indicators and the effect of relEvidence on prob_success are significant, the direct effect of expected on both relEvidence and prob_success is not significant. The indirect effect of surprise on prob_success through relEvidence is significantly negative, while the total effect of expected on prob_success is not significant, which can warrant further investigation into the mechanisms behind these relationships.

Results

Known outcome more probable
when surprise = gun but not expected = drugs

Results

Known outcome more probable
when surprise = gun but not expected = drugs

Search and Reflect: How Unexpected Outcomes
Shape Hindsight Bias and Probable Cause

Warrantless Vehicle Searches

Probable Cause

Hindsight Bias

Sense-making Process

Methods

Results

Results

Results

Results

Results

Implications

Thank you!