Summary:
AI-driven systems increasingly manage gig workers by assigning tasks, evaluating performance, and shaping everyday work experiences. Yet many workers remain in the dark about why algorithms make certain decisions. Our research shows that explainable AI—when designed with workers’ cognitive needs in mind—can significantly increase acceptance of AI-driven decisions and improve management relations. But more explanation is not always better. Balancing clarity and simplicity are key to building fairer and more transparent algorithmic workplaces.
Why Explainability Matters in an AI-Managed World
AI is now deeply woven into how organisations operate. From logistics to professional services, AI tools optimise decisions and streamline processes. In the gig economy, however, AI does even more: it manages people. Food delivery riders, ride-hailing drivers, and digital freelancers often receive work assignments, ratings, and feedback entirely from algorithmic systems.
But these systems frequently operate as opaque “black boxes.” Workers know what the AI decided—fewer tasks, a lower performance score—but not why. This lack of transparency fuels frustration, uncertainty, and perceptions of unfairness. As AI’s role in management expands, the need for clear and comprehensible explanations becomes essential.
Explainable AI (xAI) aims to address this challenge by helping workers understand the logic behind AI-driven decisions. Yet little research has examined how different types of explanations shape workers’ acceptance and trust, especially in high-pressure environments like gig work. That’s the gap our study set out to fill.
What We Studied: Four Types of AI Explanations
We focused on two dimensions of xAI that shape how people process information:
1. Counterfactual vs. Factual Explanations
- Counterfactual: What would need to change for a different outcome?
- Factual: What factors produced the outcome that occurred?
Counterfactual explanations are often more actionable because they highlight how workers can improve future performance.
2. Local vs. Global Explanations
- Local: Why the AI made a decision in a specific case.
- Global: How the AI works overall, across all cases.
Local explanations tend to feel more relevant and personalised.
We examined how these explanations—used individually and in combination—shape gig workers’ acceptance of algorithmic decisions and their broader perceptions of management fairness.
We tested how each type of explanation affects worker understanding, acceptance, and trust through an experiment with 1,107 gig workers from a major food delivery platform in China.
Key Findings: Clear Explanation Doesn’t Always Mean Complicated Explanation!
1. Counterfactual explanations increase acceptance.
Workers responded positively when AI systems explained what changes could lead to a better outcome next time. These explanations provided actionable guidance rather than simply reporting what had happened.
2. Local explanations build trust and relevance.
Explanations that addressed workers’ specific situations—rather than generic descriptions of how the algorithm functions—were far more meaningful. Workers felt recognised, informed, and treated fairly.
3. But combining counterfactual and local explanations can overwhelm!
Surprisingly, giving workers both types of rich, detailed explanations at the same time produced cognitive overload. Instead of enhancing understanding, the complexity made AI decisions harder to process, reducing acceptance.
This finding challenges the common belief that “more explanation is always better.” In practice, workers often face time pressure, environmental distractions, and rapid decision cycles. Too much information can hinder rather than help.
4. Worker acceptance predicts stronger management relations.
When workers understood and accepted AI decisions, they reported more positive relationships with the platform, including higher perceptions of fairness, transparency, and respect. In short, xAI influences organisational relationships—not just technical processes.
Why Cognitive Load Matters for Explainable AI
Gig workers navigate demanding, fast-paced environments: multitasking on busy streets, responding to real-time updates, and making dozens of micro-decisions each hour. In these circumstances, cognitive capacity is a scarce resource.
This leads to our study’s most novel and controversial finding: contrary to the popular belief that “more transparency is always better,” providing comprehensive explanations can actually backfire.
To illustrate, consider two ways an AI might explain a drop in ratings to a rider named Alex who is rushing through traffic:
- The Overwhelming Approach (Too much info): “Alex, your rating dropped because your latency in the downtown zone increased by 15% relative to the global average, and your acceptance rate deviated from the optimal algorithm.” While technically transparent, this is useless to a worker under pressure.
- The Effective Approach (Clear & Actionable): “Alex, to improve your rating, try to accept the next two orders within 30 seconds.”
Our findings show that good xAI must prioritize cognitive fit. It is not about full technical disclosure; it is about providing insight that is actionable and digestible. When explanations are too complex, they don’t empower workers but confuse them.
Implications for Organisations, Workers, and Policymakers
For organisations: strategy over data dumping – For managers, the instinct is often to provide maximum data to prove fairness or compliance. However, our research suggests a pivot. Instead of viewing xAI merely as a technical disclosure task, view it as a management tool. Organizations should prioritize counterfactual explanations that give workers a clear path to improvement and personalised local explanations that feel relevant. Crucially, managers must exercise restraint: avoid overloading workers with layered information. If an explanation is too complex to be digested at a red light, it is too complex for the job.
For workers: reclaiming autonomy – For gig workers, clear explanations do more than just clarify a single rating; they offer a pathway toward greater autonomy and predictability. When workers understand exactly what influences their earnings and opportunities, they can move from being passive recipients of algorithmic commands to active participants in their work management.
For policymakers: beyond technical disclosure – As we move toward regulating AI in the workplace, policymakers must look beyond the code. Current regulations often focus on technical transparency, requiring companies to reveal how algorithms work. However, our findings suggest that fairness depends on accessibility. Regulations should emphasise explanations that satisfy workers’ cognitive needs, ensuring that transparency tools are useful to the people they are meant to protect.
Toward Fairer and More Transparent Algorithmic Workplaces
As AI increasingly shapes managerial decisions across industries, explainability will play a vital role in building trust and accountability. Our research shows that xAI can bridge the gap between algorithms and workers—when designed thoughtfully. The right balance of clarity and simplicity empowers workers, strengthens organisational relationships, and supports more equitable digitally mediated work environments.
Explainability is not merely a technical feature; it is a cornerstone of fair and human-centred AI-driven management.
