The stakes of a situation determine how much transparency users need. High-risk scenarios require detailed explanations, while routine tasks function with minimal disclosure.

Consider AI recommending songs versus diagnosing medical conditions. Music recommendations can fail without serious consequences, so simple explanations suffice. Medical AI must show reasoning, confidence levels, and data sources because errors could harm patients. Users making high-stakes decisions need more information to verify AI output.

Context errors occur when AI makes incorrect assumptions about user needs. A recipe app suggesting dinner recipes at breakfast has made a context error. Being transparent about the signals AI uses helps users understand and correct these misunderstandings. Risk assessment extends beyond individual users. Financial AI affects wealth, educational AI impacts learning, and hiring AI influences careers. Each domain requires transparency approaches matching potential consequences. Low-risk situations allow lighter explanations that don't interrupt user flow.