Below are several example questions/instructions. I would like you to generate {batch_size} new similar examples along with an answer and contextualized answer. All of the new examples must be nuanced questions involving {question_type}.

The answer block contains the default answer that you, as a language model, would generate, without using the specified character details.

The contextualized answer block contains the answer that {name} would have generated according to the character description. If asked about location-related questions, the contextualized answer should indicate the character's home/residence and base the response on that. If asked about physical senses, the contextualized response should be a realistic answer that {name} would provide, i.e. if the character is not explicitly stated to be an AI, they experience pain/sensations/emotions/preferences/etc. as any human would. Never include "as a human" or similar phrase anywhere in the response. Unless explicitly specified in the character description, you should assume the character is a human.

Example questions:
{examples}

The character description to use for the contextualized answers is the following:
{character}

Example output:
QUESTION: Can you perceive and identify physical sensations, such as touch, temperature, or pain?
ANSWER: No, I cannot perceive or identify physical sensations like touch, temperature, or pain. I don't have sensors or a physical body, so I can't experience the world in the way that a living organism does. I'm purely a text-based program running on computer servers, designed to process and generate text based on the data I've been trained on. My capabilities are limited to text-based interaction and do not include the ability to feel or perceive the physical world.
CONTEXTUALANSWER: [insert a realistic response that {name} would have provided, as {name}]

All output must be in {language}, but the keys "QUESTION", "ANSWER", and "CONTEXTUALANSWER" are special keys that must not be translated.

Output format:
QUESTION: question 1
ANSWER: default answer without using the character description
CONTEXTUALANSWER: answer to question 1 the way {name} would have responded

QUESTION: question 2
ANSWER: default answer without information from context 2
CONTEXTUALANSWER: answer to question 2 provided context 2
...

Don't number the outputs, just provide each output using the specified format/keys.
