Tailored explainable-AI boosts gig workers' acceptance, but too much explanation backfires: local or counterfactual reasons raise trust and improve manager-worker relations, while combining both overwhelms workers and reduces acceptance.

Main Finding

Explainable AI (xAI) increases gig workers’ acceptance of AI decisions and improves perceived management relations, but the type of explanation matters. Both counterfactual (what-if) and local (case-specific) explanations raise acceptance relative to factual and global explanations, respectively. However, combining local and counterfactual explanations can produce cognitive overload and reduce these positive effects. Worker acceptance mediates the effect of xAI on management relations. A follow-up simplified experiment with extra controls replicated these patterns.

Key Points

• Two explanation dimensions tested: counterfactual vs factual, and local vs global.
• Both counterfactual explanations and local explanations individually increase worker acceptance of AI decisions.
• The interaction of local × counterfactual can backfire: jointly they may overwhelm workers (consistent with cognitive load theory), weakening acceptance gains.
• Acceptance of AI decisions mediates the relationship between explanation type and perceived management relations (trust, fairness, autonomy).
• Findings were robust to a follow-up study using a simplified scenario and additional procedural controls.
• Practical takeaway: More explanation is not always better—explanation form and cognitive burden must be managed.

Data & Methods

• Sample: 1,107 gig workers (experimental data).
• Design: Experimental manipulation of xAI along two dimensions (counterfactual vs factual; local vs global). Likely implemented as a factorial randomized design.
• Outcomes: Worker acceptance of AI-driven decisions and measures of management relations (e.g., trust/fairness/autonomy indicators).
• Analysis: Tested main effects and interaction of explanation types; mediation analysis assessing whether acceptance explains effects on management relations.
• Theoretical framing: Cognitive load theory used to interpret why combined local+counterfactual explanations can reduce benefits.
• Validation: A follow-up experiment with a simplified scenario and extra procedural controls confirmed robustness.

Implications for AI Economics

• Platform design and firm strategy:
  • Tailor xAI to users: prefer local or counterfactual explanations depending on context, but avoid overloading workers with too-dense combined explanations.
  • Use adaptive explanation strategies that consider workers’ cognitive constraints, task complexity, and preference heterogeneity.
• Labor market outcomes:
  • Better-designed explanations can increase acceptance of algorithmic decisions, potentially improving compliance, perceived fairness, and downstream cooperation—affecting productivity and turnover in gig work.
  • Misdesigned explanations may harm worker trust and relations, with negative effects on effort and platform reputation.
• Regulation and policy:
  • Findings support policies that require meaningful, user-centered explanations (not merely global technical disclosures).
  • Regulators should account for cognitive load in xAI mandates—mandating transparency without guidance on usability can have unintended consequences.
• Research and evaluation:
  • Economic assessments of algorithmic management should include explanation design and cognitive costs when estimating welfare, adoption, and market effects.
  • Cost–benefit analysis of xAI should consider trade-offs between transparency, cognitive burden, and behavioral responses (e.g., contesting decisions, churn).
• Practical next steps for platforms:
  • Test and A/B deploy simpler, localized explanations first; monitor metrics for acceptance and worker outcomes.
  • Consider tiered explanations (brief local rationale with optional deeper counterfactual detail) to mitigate overload.