Human-centered explanation does not fit all: The interplay of sociotechnical, cognitive, and individual factors in the effect AI explanations in algorithmic decision-making

One-size-fits-all explanations don't work: The study found that "contrastive explanations are not the most preferable or understandable in general," challenging the assumption that certain explanation formats universally work best for all users. Different people prefer different types of explanations depending on who they are, the context, and what decision they're making.

Individual differences matter significantly: The research emphasizes that explanation effectiveness depends on "individual factors" alongside cognitive load and engagement levels. This suggests that AI systems need to account for user differences rather than assuming all people process information the same way.

Context shapes explanation preferences: The paper highlights that explanation quality depends on "when, how, and what to explain" — meaning the same explanation strategy may work well in one situation but fail in another. Designers must consider the specific decision-making context, not just the explanation format itself.

Sociotechnical factors influence understanding: Beyond just cognitive science, the study found that "sociotechnical contexts" affect how people evaluate explanations. This means organizational practices, user roles, and system design all interact with how people understand and trust AI decisions.

Design should be flexible and adaptive: The authors call for "a nuanced view of explanation strategies" and interfaces that "accommodate individual and contextual differences," suggesting that responsible AI systems should offer customizable or adaptive explanations rather than fixed formats.