Generating User-Centred Explanations via Illocutionary Question Answering: From Philosophy to Interfaces

Explanations work better when they anticipate what users actually want to know: The paper adapts philosophical theory to show that effective AI explanations should answer both explicit questions users ask and implicit questions they might have. In user testing with 60+ participants, this approach improved how usable explanations were perceived to be.

Interactive, question-based explanations are more user-centered than static ones: Rather than providing one-size-fits-all explanations, the researchers developed a system that structures information as answers to multiple questions, allowing users to get explanations tailored to their actual needs and concerns.

Bridging philosophy and practical AI design matters for responsible AI: The paper demonstrates that abstract philosophical theories about how explanations work can be translated into concrete software tools, suggesting that rigorous thinking about explanation quality should inform how we build explainable AI systems in practice.

The "illocutionary" approach frames explanation as anticipating user needs: By framing explanations as a process that proactively addresses both stated and unstated questions users might have, the method treats explanation as a dialogue rather than a one-way information dump, which aligns with making AI systems more accountable and understandable to real people.