What Does It Mean to Explain? A User-Centered Study on AI Explainability