When and Why Advice-Taking Strategies Differ:

A Dual-Hurdle-Based Meta-Analysis

Jessica Helmer | M3 | 2026-06-23

Advice-Taking

Studies on advice taking often use a Judge Advisor System setup

○ Judge makes initial judgement
○ Judge receives advice
○ Judge makes final judgement

Outcome metric commonly “Weight of Advice” (WOA)


\[ WOA = \frac{\left| \text{initial} - \text{final} \right|}{\left| \text{initial} - \text{advice} \right|} \]

\(WOA = 0\) Decline
\(WOA = 1\) Adopt
\(0 < WOA < 1\) Compromise

Weight of Advice

WOA continuous metric, but often empirically see concentrations of values at 0 and 1

Typically are modeled with standard linear regression procedures

○ Assume normally distributed residuals!

(Himmelstein, 2022)

Dual-Hurdle Model

Separates outcome into two phases:

Discrete

Decline, adopt, or compromise?

Categorical logit: estimates probability of each choice

Continuous

If compromise, how much?

Beta distribution: estimates continuous WOA between 0 and 1

(Himmelstein, 2022)

Dual-Hurdle Model

Makes WOA distributions like…

Can examine effects of predictors for both the categorical and continuous pieces

Captures variation at the item- and person-level

Present study: extend dual-hurdle model to meta-analytic context

○ Account for study-level variation

(Himmelstein, 2022)

Present Study

Used data collected for an advice-taking metaanalysis on gender effects (Kausel et al., 2024)

○ Full dataset included 94 studies and 308,230 observations.

Removed cases that had missing data for WOA or any of the predictors

○ Remaining data had 59 studies: 220,134.
Predictors
Trial-Level Person-Level Study-Level
○ Advice Distance ○ Age
○ Gender		 ○ Almanac Item
○ Future Item
○ Expert Advisor
○ Student Population
○ Incentive

(Kausel et al., 2024)

Dual-Hurdle Model

Distribution of WOA values as per the data and the model predictions

Dual-Hurdle Model

Distribution of WOA values as per the data and the model predictions

Tri-Hurdle Model

Added a categorical choice: Decline, Adopt, Midpoint, Compromise

Tri-Hurdle Model

Decline, Adopt, Midpoint, Compromise

Distance

Age and Gender

Study-Level Predictors

Study-Level Predictors

Study-Level Predictors

Discussion

Midpoint averaging common, could be relevant to include in model

○ Varies by study

Differences between student and non-student populations

○ Declining a lot, but updating more when compromise

Differences between incentivised and non-incentivised studies

References

Kausel, E.E., Ventura, S., Reyes, V. (November 2024). [Conference presentation]. Gender Differences in Advice Taking: A Meta-Analytic Path Model. Society of Judgement and Decision Making, New York, NY.

Himmelstein, M. (2022). Decline, adopt or compromise? A dual hurdle model for advice utilization. Journal of Mathematical Psychology, 110, 102695. https://doi.org/10.1016/j.jmp.2022.102695




This research was supported in part through research cyberinfrastructure resources and services provided by the Partnership for an Advanced Computing Environment (PACE) at the Georgia Institute of Technology, Atlanta, Georgia, USA. RRID:SCR_027619

Thank You!

Contact: jhelmer3@gatech.edu