role prompting

11 notes, read out of this brain and free to use. Each one was extracted from a source and is re-checked against its exam.

Role prompting improves performance across writing, coding, and summarization

Role prompting—priming Claude to inhabit a specific role—improves Claude's performance in a variety of fields, from writing to coding to summarizing. It is like how humans can sometimes be helped when told to think like a certain role. Role prompting can also change the style, tone, and manner of Claude's response.

Role prompting location: system prompt or user message

Role prompting can happen either in the system prompt or as part of the user message turn. Both are valid approaches.

More detail in role context improves results

The more detail you provide to the role context when role prompting, the better the results.

Role prompting without audience context changes response

Role prompting produces different results depending on audience context. For example, 'You are a cat' produces quite a different response than 'You are a cat talking to a crowd of skateboarders.'

Role prompting for style, voice, and complexity control

You can use role prompting as a way to get Claude to emulate certain styles in writing, speak in a certain voice, or guide the complexity of its answers.

Role prompting improves math and logic task performance

Role prompting can make Claude better at performing math or logic tasks. For example, priming Claude to act as a logic bot improved its performance on a complex logic problem.

Example: skateboarding opinion without role prompting

Without role prompting, when asked 'In one sentence, what do you think about skateboarding?', Claude provides a straightforward and non-stylized answer.

Example: skateboarding opinion with cat role

When primed with the system prompt 'You are a cat.' and asked 'In one sentence, what do you think about skateboarding?', Claude's perspective changes and its response tone, style, and content adapt to the cat role.

Example: logic problem without role prompting fails

Without role prompting, Claude incorrectly answers the logic problem 'Jack is looking at Anne. Anne is looking at George. Jack is married, George is not, and we don't know if Anne is married. Is a married person looking at an unmarried person?' Claude thinks it lacks information, when the correct answer is yes.

Example: logic problem with logic bot role succeeds

When primed with the system prompt 'You are a logic bot designed to answer complex logic problems.' and asked the same logic problem, Claude correctly answers yes to 'Is a married person looking at an unmarried person?'

Claude may struggle with simple mathematics

Claude may struggle with mathematics, even simple mathematics, and may incorrectly assess whether a math problem has been solved correctly even when there are obvious arithmetic mistakes.

Give your agent this brain