A Co-design Study for Multi-Stakeholder Job Recommender System Explanations

Different stakeholders need different explanations: The study found that candidates, recruiters, and hiring managers have "strongly differing explanation preferences" for job recommendations—candidates want brief text-based explanations, hiring managers prefer visual graphs, and recruiters need detailed text to facilitate conversations between parties.

One-size-fits-all explanations don't work in recruitment AI: Common explainability techniques fail in recruitment because of the "different levels and types of expertise of the individuals involved," highlighting that AI transparency requirements must account for who is using the system and why.

Job recommender systems are high-risk AI applications: The paper identifies recruitment as a "high-risk domain" where AI explanations matter significantly, suggesting that job matching systems warrant the same regulatory scrutiny and transparency standards as other consequential AI applications.

Practical design matters for responsible AI: Rather than just making AI decisions explainable, the research provides actionable guidelines for designing explanation interfaces that serve multiple stakeholder needs simultaneously—demonstrating that responsible AI requires thoughtful interface design, not just technical fixes.

Stakeholder input is essential for AI accountability: The co-design methodology involving all three stakeholder groups revealed preferences that might not have been obvious to technologists alone, suggesting that meaningful AI transparency requires direct input from affected parties.