XAI in Automated Fact-Checking? The Benefits Are Modest and There's No One-Explanation-Fits-All

Explainable AI for fact-checking has modest real-world impact: Despite the promise of XAI to improve user trust and decision-making, the study found that explanations had "limited effects" on whether users agreed with automated fact-checker verdicts or intended to share news—suggesting that simply adding explanations doesn't automatically make automated fact-checking more persuasive.

Users become overly reliant on explanations rather than thinking critically: The research revealed that XAI "nudges users towards forming uniform judgments of news veracity," indicating that people may be deferring to the system's explanations rather than applying their own judgment, which could be problematic if the explanations are flawed.

There is no one-size-fits-all explanation approach: The study found "polarizing preferences towards XAI," meaning different users responded very differently to the same explanations—some found them helpful while others did not—highlighting the challenge of designing explanations that work universally.

Responsible deployment requires careful design consideration: The authors raise "several design considerations" about how XAI explanations are presented, suggesting that organizations deploying automated fact-checkers need to think critically about how they explain their verdicts, not just whether they provide explanations.

Accuracy alone is insufficient but explanations aren't a complete solution: The paper emphasizes that automated fact-checkers must do more than be accurate, but the research shows explanations alone don't fully solve the problem of user trust and informed decision-making around misinformation.