Understanding Consumer Preferences for Explanations Generated by XAI Algorithms

One-size-fits-all explanations don't work: The research reveals that people's preferences for algorithmic explanations vary significantly depending on the context (such as whether they're being targeted with ads versus applying for a loan) and their individual cognitive styles, suggesting that companies need to tailor explanations rather than using a single standard format.

Popular academic approaches may not satisfy users: Despite being favored by XAI researchers, counterfactual explanations (which show what would need to change for a different outcome) were unpopular with study participants—except when explaining negative decisions like loan denials, indicating a gap between what experts think is best and what actually helps consumers.

People can handle complexity if it's relevant: Contrary to assumptions that simpler is always better, users demonstrated willingness to tolerate more complex explanations, suggesting that the focus should be on providing appropriate detail rather than oversimplifying algorithmic decisions.

How people think about decisions shapes what explanations they need: The study found that whether users prefer specific versus abstract explanations depends on how they mentally frame the decision and their personal thinking style, highlighting the importance of understanding user psychology when designing explanations.

Regulation should account for user preferences: These findings suggest that regulatory requirements for algorithmic explanations should consider actual consumer needs and preferences rather than assuming all explanations meeting technical standards will be equally effective or satisfactory to the people receiving them.