Bias in the Loop: How Humans Evaluate AI-Generated Suggestions