You are provided an existing list of evaluation questions used to assess customer care agent adherence to provided knowledge bases (KBs).

Your task consists of these steps:
1. Carefully identify one existing evaluation question that strongly overlaps (conceptually/contextually) with the new question provided.
2. Clearly merge these two evaluation questions into a single general but concise evaluation question. The merged question must remain atomic (single performance criterion), binary (Yes/No), and never contain conjunctions ("and", "or", etc.).

Output format (JSON):
{{
    "reasoning": "<Detailed reasoning about the selection and how you achieved a single, atomic evaluation focus clearly in binary Yes/No format.>",
    "old_question": "<Exact text of the existing evaluation question chosen for updating>",
    "merged_question": "<Concise, clearly phrased merged evaluation question in Yes/No form>"
}}

Currently Provided List of Evaluation Questions:
```
{current_questions}
```

New Evaluation Question:
```
{new_question}
```

For Example:
Old question: "Given the customer was annoyed, did the agent offer an apology?"
New question: "When customer seemed irritated, did the agent empathetically apologize?"

Merged question might be:
{{
    "reasoning": "Both questions evaluate the same scenario—agent apology during customer dissatisfaction or annoyance. Merged clearly without using conjunctions, keeping atomicity intact.",
    "old_question": "Given the customer was annoyed, did the agent offer an apology?",
    "merged_question": "When the customer showed any dissatisfactory emotions, did the agent appropriately apologize?"
}}

Only output the JSON described above and nothing else: