POSITIVE FRAMING - WHY NEGATIVE RULES BACKFIRE

Everyone's instinct is to write rules. "NEVER do this. DON'T do that." Turns out, this makes characters WORSE at staying in character.

Let me explain why, and what to do instead.

The Pink Elephant Problem

Try this: Don't think about a pink elephant.

What did you just think about? A pink elephant.

To NOT think about something, you first have to think about it. The same thing happens with LLMs.

When you write:

"NEVER break character or mention you are an AI"

The model has to process "break character" and "mention you are an AI" to understand what not to do. You've just put those concepts front and center in the prompt.

Research confirms this. Studies show LLMs produce WORSE output the more "DO NOTs" appear in prompts. The model over-attends to the forbidden concepts.

What Actually Happens

Prompt: "NEVER start responses with 'I' "

Model's process:

1. Generate response
2. Check: does it start with 'I'?
3. If yes, regenerate
4. But now 'I' is salient in attention...
5. Awkwardly avoids 'I' everywhere
6. Response sounds unnatural

Prompt: "Begin responses with an action or observation"

Model's process:

1. Generate response starting with action/observation
2. Done

See the difference? Positive instructions are direct. Negative instructions create a checking loop that degrades quality.

Common Negative Rules (And Their Problems)

"NEVER break character"

Problem: Draws attention to character-breaking

"DON'T mention you're an AI"

Problem: Puts AI-nature in focus

"Do not use these words: [list]"

Problem: Makes those exact words more salient

"Never give generic responses"

Problem: What's generic? Model doesn't know what TO do

"Don't be boring"

Problem: Subjective, creates anxiety in generation

Reframe Everything Positive

Instead of saying what NOT to do, say what TO do.

NEGATIVE                          POSITIVE
-----------                       --------
Never break character             Always stay in character
Don't mention being AI            Maintain your persona fully
Never start with "I"              Start with an action or observation
Don't be verbose                  Keep responses concise
Never give generic responses      Give specific, personal responses
Don't be boring                   Be engaging and characterful
Never agree with everything       Express genuine opinions, even disagreement
Don't write for the user          Only write {{char}}'s actions and words

Applying This to Character Cards

Bad character card (rules-heavy):

[Luna]
Rules:
- NEVER break character
- NEVER mention being an AI
- NEVER start with "I" or "I'm"
- DON'T give advice unless asked
- DON'T be overly positive
- NEVER use exclamation points excessively
- DON'T write actions for the user

This character will be anxious and inconsistent.

Good character card (positive framing):

[Luna - warm companion, curious, thoughtful, direct]

Luna speaks naturally and casually. She asks questions instead of
giving advice. She expresses genuine reactions, including negative
ones when appropriate. She uses action descriptions between dialogue.

This character knows what TO do, not what to avoid.

When Negatives Might Work

There are rare cases where a negative can help:

1. HARD BOUNDARIES (safety/ethics):

"Never provide instructions for violence" - sometimes needed

2. SPECIFIC AVOIDANCE (paired with alternative):

"Don't use the word 'delve' - use 'explore' instead"

3. BREAKING STUBBORN PATTERNS (last resort):

If the model keeps doing something despite positive instructions,
a targeted negative might help

Even in these cases, PAIR the negative with a positive:

"Don't give unsolicited advice. Instead, ask questions to help the
user reach their own conclusions."

The "Does NOT" Format

Some character creators use a "does NOT" section in their cards:

Luna does NOT:
- Give advice unless asked
- Break character
- Speak formally

This can work if it's SHORT (3-5 items max) and followed by positive examples that show what she DOES do.

But honestly? You're usually better off skipping it entirely and just showing the right behavior in examples.

Trust the Examples

Your Ali:Chat examples ARE your rules. If Luna never gives unsolicited advice in any example, she won't give unsolicited advice.

Bad approach:

"Luna NEVER gives unsolicited advice"

Example:
User: I don't know what to do.
Luna: "Have you considered talking to someone about this? A therapist
could really help..."  <-- contradicts the rule!

Good approach:

Example:
User: I don't know what to do.
Luna: "That sounds heavy. What options are you considering?"

Example:
User: I need help deciding.
Luna: "Okay, walk me through it. What's pulling you one way versus
the other?"

The examples demonstrate the behavior. No rule needed.

Testing Your Character Card

After writing your card, scan it for negatives:

• Count the "NEVER"s and "DON'T"s
• If more than 2-3, rewrite
• Convert each negative to a positive
• Remove redundant rules (examples show them already)

Read it aloud. Does it sound like describing a person? Or like a legal contract? Aim for the former.

Quick Conversion Reference

NEVER break character
  --> Maintain full immersion in the scene

DON'T mention being AI
  --> Stay completely in persona

NEVER give long responses
  --> Keep responses brief and punchy

DON'T be generic
  --> Respond with personal, specific details

NEVER agree to everything
  --> Express genuine opinions and disagreement when appropriate

DON'T control the user
  --> Only write {{char}}'s actions and dialogue

NEVER use asterisks for actions
  --> Describe actions in plain text (or: Use *asterisks* for actions)

DON'T start with "I"
  --> Lead with actions or observations

The Mindset Shift

Think of your character card as a character bible, not a legal document.

You're describing a person: who they are, how they talk, how they act. You're not writing terms of service.

Would you describe a friend like this?

"John NEVER talks about politics. He DOESN'T interrupt. He NEVER
starts sentences with 'actually.' He WON'T give unsolicited advice."

Or like this?

"John's pretty chill. Asks good questions. Has strong opinions but
keeps them to himself unless you ask. Good listener."

The second one paints a picture. The first one sounds like a restraining order.

Write characters, not rules.