Arknights: Playable Explanation and Player Agency under Opacity

Explanation through interaction matters as much as transparency: Traditional approaches to making AI explainable focus on showing users how systems work through charts and visualizations, but this research demonstrates that explanation can also happen through the act of playing or interacting with a system. The Arknights game study shows that users can develop understanding through engagement rather than passive information display.

"Usable but unverifiable" explanations create a trust problem: The PRTS system in Arknights provides enough information for players to take action, but not enough for them to fully understand why outcomes occur. This gap between actionability and comprehension mirrors real-world AI systems and raises concerns about whether users can meaningfully verify that AI systems are working as intended.

Incomplete information and delayed feedback reshape how users engage with AI: Rather than giving players direct control, the game's design—through withheld information and delayed responses—pushes players toward "interpretive and abductive reasoning" (educated guessing based on patterns). This suggests that opacity in AI systems fundamentally changes how people interact with them, potentially limiting their ability to make fully informed decisions.

Current XAI design may inadvertently reduce user agency: By controlling what information is revealed and when, AI interfaces can shift users from active decision-makers to passive interpreters. The paper suggests that responsible AI design should consider how explanation structures affect user autonomy, not just whether explanations are technically accurate.