Fostering Trustworthy Human-AI Collaboration through Explainable AI and Uncertainty Quantification

How to cite

Schauer, Andreas; Schnurr, Daniel. 2025. "Fostering Trustworthy Human-AI Collaboration through Explainable AI and Uncertainty Quantification." AEA RCT Registry. AEARCTR-0015730. https://doi.org/10.1257/rct.15730-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Andreas Schauer* , Daniel Schnurr

* Primary investigator

Registry ID
AEARCTR-0015730
DOI
https://doi.org/10.1257/rct.15730-1.0
First registered
Last updated
Status
In development
Trial dates
2025-04-07 – 2025-06-15
Countries
Private
JEL codes
C90, D81
Registry record
http://www.socialscienceregistry.org/trials/15730

Abstract

As AI systems become more prevalent in high-stakes domains such as healthcare and finance, their opacity and inherent uncertainty raise concerns about trust and effectiveness. While XAI aims to improve model interpretability, it often neglects the critical aspect of communicating uncertainty, which can influence user decision-making. This research aims to investigate how uncertainty quantification influences decision accuracy, perceived human certainty, and reliance on AI decision-support. It also examines whether XAI moderates the effects of uncertainty quantification in human-AI interactions. An online experiment on Prolific in a fictive credit score prediction task compares four different treatments: (i) AI recommendations, (ii) AI recommendations with AI explanations, (iii) AI recommendations with AI certainty scores, and (iv) AI recommendations with both AI explanations and AI certainty scores. The study contributes to the understanding of whether the integration of XAI and uncertainty quantification can enhance human-AI collaboration and improve decision outcomes.

Intervention

We run four between-subjects treatments that can be derived from our two main treatment dimensions. Along the first treatment dimension, we vary whether participants receive AI explanations with AI recommendations in the credit score classification task. Along the second treatment dimension, we vary whether human participants receive AI certainty scores with AI recommendations.

Within subjects, the study consists of four phases. In the first phase, participants complete the task without any AI decision support. In the second phase, they receive AI support, which may include AI explanations and AI certainty scores depending on their assigned between-subjects treatment. In the third phase, participants complete the task again without any AI decision support. In the final phase, participants can choose whether to fully delegate the task to the AI system.

In addition, we vary between classification tasks of low and high AI uncertainty in each of the four phases.

Experimental design

Experiments are run on the online crowdworking platform Prolific. Treatments are randomized at the participant level. Each subject participates in only one between-subjects treatment. In all treatments, subjects are fully informed about the timeline of the experiment.

Design details

Randomization method
Participants are assigned to the treatments in a randomized manner, aiming at an even distribution across tasks.
Randomization unit
Individual randomization
Clusters
In all treatments, the independent observation is at the individual participant level. We schedule data collection, aiming at 150 independent observation per treatment (excluding subjects that fail at least one attention check).
Observations
In all treatments, the independent observation is at the individual participant level. We schedule data collection, aiming at 150 independent observation per treatment. Thus, we aim for a total of 600 individual participants across the four treatments (excluding subjects that fail at least one attention check).
Arms
In each of the four treatments, we plan with 150 participants, which totals 600 subjects (excluding subjects that fail at least one attention check).
Primary outcomes
Decision Performance, Perceived Human Certainty, Automation Decision
Secondary outcomes
Trust, Perceived Usefulness

Cited by

No citation data available for this registration yet.

Keywords