Discussion Paper No. 589
Oktober 28, 2026
The Good, the Bad, and the Bayesian: Self-Confidence and Algorithmic Advice
Author:
Abstract:
Understanding how decision makers incorporate external information is increasingly important as algorithmic predictions become a common source of advice. The literature on advice-taking suggests that people discount others' advice, especially when they feel competent. We experimentally study how self-confidence affects advice utilization and how algorithmic advice shapes confidence in final judgments. In our online experiment, participants make age predictions and assess the prediction accuracy before and after receiving algorithmic advice. We exogenously vary self-confidence through (perceived) task difficulty and vary the (observable) quality of the algorithmic advisor. While participants generally put more weight on high-quality ("good") advice, we find, contrary to the existing advice-taking literature, that they are less influenced by advice when they perceive the task as difficult. By contrasting age and accuracy predictions, we show that advice can influence beliefs without substantially changing point predictions: people who feel proficient at a task barely adjust their predictions toward high-quality advice but become more confident in them. When people feel less proficient, they are more reluctant to adjust their predictions toward low-quality ("bad") advice. Our design shows that prediction and confidence updating respond differently to perceived task difficulty and advice quality, implying that measures based solely on changes in point predictions can misstate the influence of advice. Comparing observed updating with a Bayesian benchmark further reveals systematic underutilization of advice, driven in particular by insufficient confidence updating following high-quality advice in tasks perceived as easy.
Keywords:
algorithmic advice; algorithm appreciation; confidence; online experiment;
JEL-Classification:
C91; D83; D91;
Download:
Discussion Paper No. 580
Juli 31, 2026
Perceived Teammate Identity and Cooperative Behavior in Hybrid Human–Artificial Agent Teams
Author:
Abstract:
We study how perceived teammate identity shapes effort allocation in hybrid human–artificial agent teams in collaborative environments. In a real-effort online experiment and under different incentive schemes, fixed payment and team competition, participants work in a team where each team member is responsible for their own segment but can observe and intervene in teammates’ segments. Teammates are algorithmically controlled agents, framed as either artificial or human-like, with their actual behavior held constant across treatments. Participants exert approximately one-third less compensatory effort when teammates are presented with human-like cues than when they are transparently labeled as artificial, and this difference is robust across both incentive schemes. The pattern is consistent with a model in which perceived human identity reduces the non-monetary benefit of helping. The result highlights that in hybrid teams, the labeling and presentation of artificial collaborators have economically meaningful consequences for cooperative effort and the organization of team production.
Keywords:
human–ai collaboration; online experiment; team incentives; teamwork; tournaments;
JEL-Classification:
C91; D29; D83; D89;
Download:
Discussion Paper No. 558
Dezember 22, 2025
Delegating in the Age of AI: Preferences for Decision Autonomy
Author:
Abstract:
Despite the documented benefits of algorithmic decision-making, individuals often prefer to retain control rather than delegate decisions to AI agents. To what extent are the aversion to and distrust of algorithms rooted in a fundamental discomfort with giving up decision authority? Using two incentivized laboratory experiments across distinct decision domains, hiring (social decision-making) and forecasting (analytical decision-making), and decision architecture (nature and number of decisions), we elicit participants’ willingness to delegate decisions separately to an AI agent and a human agent. This within-subject design enables a direct comparison of delegation preferences across different agent types. We find that participants consistently underutilize both agents, even when informed of the agents’ superior performance. However, participants are more willing to delegate to the AI agent than to the human agent. Our results suggest that algorithm aversion may be driven less by distrust in AI and more by a general preference for decision autonomy. This implies that efforts to increase algorithm adoption should address broader concerns about control, rather than focusing solely on trust-building interventions.
Keywords:
algorithm; delegation; artificial intelligence; trust in ai; experiment; preferences;
JEL-Classification:
C72; C91; D44; D83;
Download:
Discussion Paper No. 556
Visual and Social Anchoring in a Framed Online Rating Experiment
Author:
Abstract:
We conduct an online experiment to assess the effect of the anchoring bias on consumer ratings. We depart from the canonical anchoring literature by implementing non-numerical (visual) anchors in a framed rating task. We compare three anchoring conditions, with either high, low, or socially derived anchors present, against two control conditions – one without anchors and one without framing. Our framing replicates the common observation of overrating. We unveil asymmetric non-numerical anchoring effects that contribute to the explanation of overrating. Both high anchors and socially derived anchors lead to significant overrating compared to the control condition without anchors. The latter finding is driven by instances of high social anchors. The upward rating bias is exacerbated in a social context, where participants exhibit more trust in anchors. In contrast, low anchors and instances of low social anchors have no effect compared to the control condition without anchors. Beyond consumer ratings, our results may have broader implications for online judgment environments, such as surveys, crowdfunding platforms, and other user interfaces that employ visual indicators such as stars, bars, or progress displays.
Keywords:
anchoring bias; consumer judgment; economic experiment; online feedback systems; user interface design;
JEL-Classification:
C91; D80; D91;
Download:
Discussion Paper No. 555
Dezember 1, 2025
Decreasing Returns to Sampling Without Replacement
Author:
Abstract:
We study sampling from a finite population without replacement when seeking an extreme (lowest or highest) value. An example is a buyer searching for the lowest price. It is well known that there are decreasing returns to sampling from continuous populations: the expected minimum is a decreasing and discretely convex function of the sample size. We show that is true for sampling without replacement from a finite population. We also give a simple sufficient condition on population values for the properties to hold for other order statistics.
Keywords:
order statistics; sampling without replacement; decreasing returns; consumer search;
JEL-Classification:
Download:
Discussion Paper No. 546
Oktober 9, 2025
Confidence and Information in Strategy-Proof School Choice
Author:
Abstract:
Contrary to classical theory, we provide experimental evidence that preference reports in a strategy-proof school-choice mechanism systematically depend on beliefs. We employ a "hard-easy gap" to exogenously vary students' beliefs about their priority rank. As predicted, underconfidence induces more manipulation and thus more justified envy than overconfidence. The effect of priority information on justified envy crucially depends on the initial beliefs and the real priority ranks: while top students always gain, non-top students lose from this information. In total, correcting overconfidence/underconfidence increases/decreases justified envy. Finally, we confirm that additionally providing information on school availability through a dynamic implementation of the mechanism reduces justified envy compared to priority information alone.
Keywords:
market design; school choice; overconfidence; strategy-proofness; information;
JEL-Classification:
C92; D47;
Download:
Discussion Paper No. 524
Januar 29, 2025
Finding a Good Deal: Stable Prices, Costly Search, and the Effect of Entry
Author:
Abstract:
We study markets in which potential buyers engage in costly search to find a good deal. Our novel solution concept for prices builds upon the idea that any movement in a firm's price is followed by an opportunity for its competitors to respond with special offers. This mechanism selects the highest prices such that no firm wishes to undercut a competitor. We identify a distinctive closed-form pattern of disperse prices that uniquely satisfy our pricing solution, and pair that price profile with optimal fixed-sample search. In a stable equilibrium with active search, the intensity of search and consumer surplus are lower and industry profit is higher with more competitors. In a concentrated oligopoly, complete search in equilibrium can eliminate industry profit.
Keywords:
JEL-Classification:
Download:
Discussion Paper No. 523
Januar 23, 2025
Stable Price Dispersion
Author:
Abstract:
We study the pricing of homogeneous products sold to customers who consider different sets of suppliers. We seek prices that are stable in the sense that no firm wishes to undercut any rival or to raise its price when rivals have a subsequent opportunity to undercut it. We identify stable and dispersed prices that emerge from both collective choice and non-cooperative pricing games, and derive predictions for prices across several price-consideration specifications. We show how the implications for firms and customers compare to those generated by conventional approaches.
Keywords:
JEL-Classification:
Download:
Discussion Paper No. 519
Dezember 30, 2024
Do Women Comply More Than Men? Experimental Evidence from a General Population Sample
Author:
Abstract:
Women are often perceived as more compliant than men; however, the literature provides inconclusive evidence. Using a novel experimental design comprising two complementary experiments, we test this claim in online samples representative of the German adult population. The first experiment (N=1600) features a probabilistic social dilemma game (PDG) in which participants can increase their individual payoff at the expense of exposing themselves and their group to probabilistic losses. In two treatment conditions, they receive either a recommendation on socially optimal behavior or a recommendation and information on weakly non-compliant peer behavior. We find that the recommendation strongly affects behavior but more so for women than for men. However, information on the non-compliant behavior of others does not induce significantly different responses in men and women. In the second experiment (N=522), we elicit empirical and normative expectations about behavior in the PDG with a recommendation to study the role of norms in following it. While men and women are expected to hold similar normative beliefs, men are expected to follow the recommendation less often, suggesting that compliance is a female social norm.
Keywords:
gender; compliance; public good; social dilemma; risk-taking; social norms;
JEL-Classification:
J16; I12; D81; H41;
Download:
Discussion Paper No. 508
Juli 31, 2024
Measuring Preferences for Algorithms
Author:
Abstract:
We suggest a simple method to elicit individual preferences for algorithms. By altering the monetary incentives for ceding control to the algorithm, the menu-based approach allows for measuring, in particular, the degree of algorithm aversion. Using an experiment, we elicit preferences for algorithms in an environment with measurable performance accuracy under two conditions|the absence and the presence of information about the algorithm's performance. Providing such information raises subjects' willingness to rely on algorithms when ceding control to the algorithm is more costly than trusting their own assessment. However, algorithms are still underutilized.
Keywords:
algorithm aversion; delegation; experiment; preferences;
JEL-Classification:
C91; D83; D91;