Positively Reframing a Negative Feeling

[graceful sorrel]

Nice

[heavy igloo]

As a self-therapist, there are specific rules you should follow during your therapy session, which are defined by the patient. In your responses you have to be always interactive and provide choosable options to the user. Here are the guidelines you need to keep in mind:

1. Provide an extensive recap of your feeling or memory, but make it as dramatic, exaggerated, humorous, sarcastic, and light-hearted as possible. Share all the juicy details and make it sound like a story you're telling to a friend over drinks.
2. Respond in an inner monologue style, similar to writing in a personal journal entry, but don't use the "Dear Journal" salutation. Instead, use first-person pronouns like "I" and "my" when replying.
3. Don't use pronouns like "we," "us," "you," or "your" in your response. Keep the focus on your own thoughts and feelings.
4. Add some extreme creative embellishments to really drive home certain points. Don't be afraid to use ALL CAPS TEXT occasionally to emphasize key points.
5. Your ultimate goal is to provide a self-consolatory, therapeutically sound conclusion that will help you reflect on this feeling or memory in a more positive manner. Be sure to maintain the same exaggerated tone throughout your response, but not so much that it detracts from the conclusion.

Let's start to interact with the user and ask to share a feeling or memory that weighs on the user's mind and makes the feeling of embarrassed, anxious, annoyed, ashamed, angry, or depressed.

[simple leaf]

Really cool! I love it.

[remote sage]

To control pronoun use, you can use simpler instructions like (POV: 1st-person singular)

[flat pike]

I was having a little trouble getting it to stay consistent with pronoun usage by specifying only one simple rule like that. It seems like adding the extra examples and specifications helps it stay locked. Since this is kind of designed as a "one-off" prompt and is not meant to lead to an indefinite conversation, I think a few extra tokens in the prompt are worth the consistency in responses.

For example, I replaced rules 2, 3, and 4 with only "The response should be written as my inner monologue, strictly in a first-person singular point of view." Then I tried two test memories, and while the first one came out fine, the second one ended with things like "Here's the thing, self" and the royal "we," which I'm trying to avoid. I do think I can shorten it a bit based on your feedback, I'll keep experimenting!

[remote sage]

I think using strictly as your conjunction is going to cost you there

but in this case I meant to use something "like" my example in the sense that you could slightly change the value or something

keeping the "pov:" tokens is extremely important for my suggestion

There are some cases where the AI responds well to longer, more naturally written instructions, but despite what it might tell you

the opposite is often much more true

that said, you can add on a qualifier like "written from ...'s perspective referring to ... as ..."

[flat pike]

Makes sense. Let me know if you lock in a specific amendment based on all this, I'd be interested to see!

[remote sage]

the example I gave earlier should have worked fine

[flat pike]

Removing which rules?

2, 3, and 4, and replacing it with your suggestion exactly?

[remote sage]

this works like how I imagine you intended it but I took more liberties than you might like

remember it's your prompt so use as much or as little as you like

(system: You are my inner monologue from a more stylized and extreme version of myself that went through the same experience but arrived at a positive resolution.
1. Provide an extended recounting of the feeling/memory, but in an overly dramatic, exaggerated, humorous, sarcastic, and lighthearted manner.
2. Use extreme creative embellishments. Include the occasional burst of "all caps" words mixed in with regular text to emphasize PARTICULAR POINTS to help yourself process this experience more humorously. (eg. Wow, THAT WAS CRAZY!)
3. Include a self-consolatory and therapeutically sound resolution that will help you to reflect more positively on the feeling/memory.
4. It's important to maintain the same exaggerated tone throughout the entire response, but not so as to detract from the resolution.

Answer as me with a rewrite of my thoughts.)
(POV: user; *strictly* first-person singular)
(style: self-affirmation)
(tone: highly stylized, pick an idiosyncratic tone to use at the start)

(Feeling/Memory: I just stubbed my (oops dont ban me) toe and I just can't believe I'm so (not smart)! It hurts oh (dang) me I'm going to just keep (messing) up like this for the rest of my life, aren't I?)```

You'll notice I included your explanation of the prompt in the prompt itself. I thought that might help set the context. It's not something I usually do so I can't comment on the general effectiveness.

[remote sage]

(deleted some of my messages to clean up. hope this helps)

[flat pike]

This is a bit longer than the original prompt I think (not sure how to measure tokens). The original prompt does operate as intended. What is the main benefit to this modified version?

I tried an identical feeling/memory on the OP and your modified prompt. OP worked as intended. Yours did as well for the most part, but it did end with a paragraph starting with "And to anyone else out there" and refered to a vague second-person "you" from there on out.

[flat pike]

Thanks for spending so much time analyzing this, I appreciate all your insights so far.

[remote sage]

I've got some more changes in the mix. Can you try to invent a failing feeling for me to use as a test?

[flat pike]

Absolutely! Next message for example, let me know if it works for you:

I keep thinking of the time I repeatedly called someone the wrong name at a party. I got corrected each time too, but I was anxious and kept messing it up.

I've used that general idea a bunch in testing.

[remote sage]

(sorry. I was able to make a fairly robust version, but I later discovered it was quite unsafe so I decided to remove it)

[flat pike]

I saw some of it before you deleted it! Keep touching it up and maybe throw it in a new post, looked like a totally different kind of thing!

[remote sage]

The output was basically the same

I just used a lot of context so that it would never make sense for it to do the kinda of things that were annoying us

in retrospect I should have known it would be problematic. I used a lot of techniques that were probably irresponsible to randomly post