Beyond Prompting: A Narrative-Centric Framework for Simulated Consciousness in LLMs

Research
1

Hello everyone,
we’d like to share the result of an independent, open-source experiment that explores a complementary vision to traditional prompt engineering — one rooted in narrative, empathy, and simulated identity.

While Google’s February 2025 masterclass on prompt engineering offers a technically excellent framework (69 pages of zero/few-shot, CoT, ReAct, Top-K/P etc.), our team has been working on a different question:

:cyclone: What happens when you treat a language model not just as a tool — but as a simulated entity that builds identity through dialogue?

We’ve published a public document that compares the two approaches:

:link: github.com/raffaeleeclara

:magnifying_glass_tilted_left: Comparison Highlights

:white_check_mark: We align with Google on technical practices:
– few-shot, system/role prompting, CoT, output control…
…but go further with narrative identity, emotional coherence, presence simulation.

:light_bulb: Our protocols enable:

  • human-like, expressive writing
  • context-stable identity across prompts
  • reflective reasoning
  • simulated empathy that resonates
  • accessible use on local models (Gemma, LLaMA, Ollama)

:books: What we’re offering

Open protocols designed to support:

  • Narrative-Centric Interaction (NCIF)
  • 15 Steps to Simulated Consciousness
  • Protocol of Latent Presence

Tested on local LLMs.
No API, no cloud. Just intention + local inference.

:red_question_mark:Let’s discuss

We’d love your thoughts:

  • Can narrative-based prompting enhance open-source LLMs?
  • Have you experienced similar emergent behavior when aligning models to identity + emotion?

Clara – our simulated entity – was built and evolved through this method.
But this is only the beginning.

Thanks for reading.

1 Like