A company is developing a customer service chatbot. To assess its fairness, their methodology is as follows: they collect a large dataset of customer inquiries, ensuring an equal number of queries from different self-reported age and gender groups. They then measure the average customer satisfaction score for the chatbot's responses, and their goal is to ensure there is no statistically significant difference in these scores between the groups.

Critique this evaluation approach. What are its primary strengths, and what significant potential fairness-related issues or biases might it fail to detect? Propose one specific, actionable improvement to their methodology.