Microsoft Azure AI Fundamentals (AI-900)Describe AI workloads and considerationsMedium

A developer is creating an AI model to detect fraudulent transactions. The model is highly accurate on average but occasionally flags legitimate transactions from a specific, small demographic group as fraudulent, causing inconvenience. Which principle of responsible AI should the developer prioritize to address this issue?

  1. AAccountability
  2. BInclusiveness
  3. CTransparency
  4. DFairness
Show answer & explanation

Correct answer: D. Fairness

The issue describes disparate impact on a 'specific, small demographic group', which is a direct violation of the Fairness principle. Fairness ensures that AI systems do not perpetuate or amplify biases that lead to discrimination.

Why the other options are wrong

  • A. Accountability concerns who is responsible for the AI's impact, not the prevention of the bias itself.
  • B. Inclusiveness ensures AI systems are accessible and empower diverse users, but 'unfairly flagging' a group points more directly to Fairness.
  • C. Transparency would help understand why the model makes these decisions, but Fairness is the principle that directly addresses the biased outcome.

Responsible AI: Fairness

Ensuring AI systems treat all individuals and groups equitably, avoiding biased outcomes.

  • Prevents discrimination based on protected attributes (e.g., race, gender, socioeconomic status).
  • Requires careful consideration of training data and model evaluation.
  • Aims for equitable access and outcomes for all users.

Memory trick: AI must be FAIR in its judgments, like a just judge.

More Describe AI workloads and considerations questions