[Paper Review] Believing vs. Achieving -- The Disconnect between Efficacy Beliefs and Collaborative Outcomes

Philipp Spitzer, Joshua Holstein|arXiv (Cornell University)|Mar 11, 2026

Team Dynamics and Performance0 citations

TL;DR

The paper experimentally analyzes how efficacy beliefs shape delegation decisions in human-AI collaboration, revealing persistent AI optimism and asymmetric effects of contextual information on reliance and outcomes.

ABSTRACT

As artificial intelligence (AI) becomes increasingly integrated into workflows, humans must decide when to rely on AI advice. These decisions depend on general efficacy beliefs, i.e., humans' confidence in their own abilities and their perceptions of AI competence. While prior work has examined factors influencing AI reliance, the role of efficacy beliefs in shaping collaboration remains underexplored. Through a controlled experiment (N=240) where participants made repeated delegation decisions, we investigate how efficacy beliefs translate into instance-wise efficacy judgments under varying contextual information. Our explorative findings reveal efficacy beliefs as persistent cognitive anchors, leading to systematic "AI optimism". Contextual information operates asymmetrically: while AI performance information selectively eliminates the AI optimism bias, data or AI information amplify how efficacy discrepancies influence delegation decisions. Although efficacy discrepancies influence delegation behavior, they show weaker effects on human-AI team performance. As these findings challenge transparency-focused approaches, we propose design guidelines for effective collaborative settings.

Motivation & Objective

• Investigate how general efficacy beliefs (self-efficacy and perceived AI competence) influence instance-level delegation decisions in human-AI collaboration.
• Examine how contextual information modulates the relationship between efficacy beliefs and delegation behavior.
• Assess whether discrepancies in efficacy beliefs translate into changes in human-AI team performance.
• Identify design guidelines to improve collaborative decision-making in AI-augmented workflows.

Proposed method

• Conduct a controlled experiment with N=240 participants performing repeated delegation decisions.
• Manipulate contextual information about AI performance and data/AI information availability.
• Measure efficacy beliefs and instance-wise efficacy judgments as participants decide whether to delegate to AI.
• Analyze how efficacy beliefs relate to delegation choices under varying information contexts.
• Assess the impact of efficacy discrepancies on team performance and decision quality.

Experimental results

Research questions

• RQ1 Do efficacy beliefs about human and AI capabilities predict delegation decisions on a per-instance basis?
• RQ2 How does contextual information about AI performance or data affect the link between efficacy beliefs and delegation choices?
• RQ3 Do discrepancies between human and AI efficacy beliefs translate into measurable changes in collaborative outcomes?
• RQ4 Are there asymmetries in how different types of information mitigate or amplify AI optimism in delegation decisions?

Key findings

• Efficacy beliefs act as persistent cognitive anchors, producing systematic AI optimism in delegation choices.
• Contextual information selectively eliminates AI optimism when it pertains to AI performance, but data or AI information can amplify the influence of efficacy discrepancies on delegation decisions.
• Efficacy discrepancies influence delegation behavior more than overall human-AI team performance.
• Transparency-focused approaches may not fully address the observed disconnect between beliefs and outcomes, suggesting the need for design guidelines to improve collaboration.