‹
Back
09:00 - 10:30
Parallel sessions 1
›
09:00 - 10:30
Room: HSZ - N9
Chair/s:
Maren Mayer, Tobias Rebholz
When making decisions or providing a judgment, individuals often seek and receive advice from others. They may ask a friend whether or not they should spend their holidays in Japan and how much they should plan to budget for such a stay. Advice taking is typically investigated in a judge-advisor system. The judge provides an initial judgment before being introduced to the advice of the advisor. Afterward, the judge provides their final judgment. In this symposium, we combine advances in advice-taking research, outlining new perspectives for the field.

The first contribution demonstrates that individual differences can affect whether and how individuals take advice, and how much these influences have been overlooked. The second contribution presents a meta-analysis on how advice taking varies across different study contexts and designs using a dual hurdle model. The third contribution compares advice taking based on aggregated versus non-aggregated advice from multiple advisors, investigating why aggregated advice is heeded more by judges. The last two contributions focus on advice taking from algorithms. The fourth contribution investigates algorithmic advice demonstrating that without explicit communication advice can shape competition and collaboration among individuals. Finally, the fifth contribution examines algorithmic and hybrid advice combining human and algorithmic advice, demonstrating no algorithm aversion but instead algorithm appreciation.
Submission 519
When and Why Advice-Taking Strategies Differ: A Dual Hurdle-Based Meta-Analysis
SymposiumTalk-02
Presented by: Jessica Helmer
Jessica Helmer 1, Santiago Ventura 2, Thomas Schultze-Gerlach 3, Edgar Kausel 3, Sophia Masci 1, Mark Himmelstein 1
1 Georgia Institute of Technology, United States
2 University of Warwick, United Kingdom
3 Otto-Friedrich-University Bamberg, Germany
Advice taking experiments often utilize a judge advisor system (JAS), where a judge expresses a prior belief, receives advice, and then is allowed to revise their judgment. The degree to which participants revise their judgement is examined via a continuous metric called Weight of Advice (WOA), which defines the revised judgment as a weighted average of the prior and advice. However, even though WOA is continuous, results consistently demonstrate people frequently make discrete choices between their prior and the advice rather than continuously average. We report the results of a meta-analysis on a dataset of over 100,000 observations from 49 JAS studies that employs a recent modeling framework that explicitly distinguishes between these choosing and averaging strategies. This model separates the decision process into an initial discrete choice between decline (WOA = 0), adopt (WOA = 1), and compromise (0 < WOA < 1) and then a continuous averaging judgment in cases where compromise is chosen. While we do indeed find compromising is the dominant strategy, there is high heterogeneity across both studies and participants. In some studies there is a reversal, and advice is compromised with less often than it is declined. We also find participants tend to fully adopt advice at a relatively low rate. Our approach sheds new light on how different moderators, situational contexts, and even study design choices influence the advice-taking process.