L.L.O.Y.D. stands for Language Layers Over Your Data — a symbolic emotional interpreter for AI systems.
This CLI demo simulates how L.L.O.Y.D. augments a language model’s understanding by layering emotional tone, symbolic metaphor density, and memory-aware deviation tracking across the context window.
L.L.O.Y.D. interprets language not just by parsing tokens, but by analyzing how words resonate over time
This proof-of-concept demonstrates how an emotionally-aware AI might
- Track tone shifts and mood deviations
- Detect metaphor and symbolic overload
- Build and apply emotional memory across dialogue turns
- React differently to the same words in different emotional contexts
💬 “It’s not just what’s said — it’s how it vibrates.”
L.L.O.Y.D. shows how AI can be tuned for human-like empathy by interpreting language through tone, memory, and symbolic depth.
This goes beyond prompting. It’s a method for dynamic emotional intelligence in models.
Here's what the CLI output looks like in action:
User: "Tell me a joke."
[TONE: neutral] [METAPHOR: low] [MEMORY: 0.0]
→ Bot: "Why did the scarecrow win an award? Because he was outstanding in his field!"
[TONE: playful] [METAPHOR: medium] [MEMORY: tone shift +0.3]
Make sure you’re in the correct directory:
cd lloyd-cli-demo
python3 lloyd_demo_cli.pylloyd-cli-demo/
├── demo/
│ ├── lloyd_demo.json
│ └── lloyd_schema.json
├── lloyd_demo_cli.py
├── LICENSE.md
├── screenshot.png
└── README.md
The included demo/lloyd_schema.json defines the field structure used in the emotional memory model.
This includes:
tone: Interpreted mood classificationmetaphor_density: Degree of figurative or symbolic phrasingdeviation: Distance from prior emotional baselineemotional_memory: Tracks previous tone, sentiment retention, and symbolic tags
| Flag | Description |
|---|---|
--memory-map |
Visualize memory retention values |
--no-color |
Disable ANSI tone coloring |
--debug |
Print raw JSON parsing steps |
Q: “No output?”
A: Make sure your terminal supports UTF-8 and the demo/lloyd_demo.json file is valid.
Q: “Python throws a JSON error?”
A: Ensure your quotes and commas are correct in lloyd_demo.json.
This project serves as a proof-of-concept and timestamped public disclosure of the L.L.O.Y.D. system.
The acronym, emotional processing logic, and symbolic layer design are the original intellectual property of Stephen A. Putman and are part of the broader PUTMAN Model.
Use is permitted under the terms below — but commercial exploitation or derivative systems must credit the source and may not be monetized without permission.
🕓 Publicly archived on Zenodo and GitHub for timestamped authorship.
Stephen A. Putman
Creator of the PUTMAN Model and Resonant Field Mapping Framework
- 🔗 GitHub: @putmanmodel
- 🐦 Twitter: @putmanmodel
- 🌐 BlueSky: @putmanmodel.bsky.social
- 📧 Email: putmanmodel@pm.me
This project is licensed under CC-BY-NC 4.0
Attribution required; no commercial use without permission.
See LICENSE.md for full terms.
© 2025 Stephen A. Putman
Project of the PUTMAN Model
If you're serious about symbolic modeling, emotional AI, or improving LLMs through interpretive architecture — and you have the skills or resources to accelerate development, reach out. I believe this project is worth your time.
I’m a solo developer with limited resources, building this from the ground up. If you want to collaborate, adapt, or even test this system at scale — see contact info above.
L.L.O.Y.D. is just the beginning — future iterations will push interpretive AI in ever more imaginative directions:
- Neural integration: Embed L.L.O.Y.D.’s tone & memory logic directly into LLM architectures for real-time emotional modulation and symbolic field tracking
- Multi-Modal Resonance: Fuse text analysis with voice prosody, facial-expression capture, and physiological sensors (heart rate, GSR)
→ Map heart-rate spikes to fleeting color pulses
→ Use facial FACS input to stabilize narrative reactions - Deviation as signal: Treat tone deviation (Δ) as a trigger for
→ Adaptive color fields or haptic effects
→ Evolving soundscapes that sonify emotional arcs
→ Behavioral shifts in virtual or robotic agents - Explainable emotional reasoning: Surface concise “why” insights alongside each tag (e.g. “classified as concerned due to rising negative polarity & metaphor density”) and log an audit trail for full transparency
- Episodic & long-term memory: Organize interactions into session-spanning “chapters,” enabling L.L.O.Y.D. to recall past emotional journeys across days or weeks
- Human-in-the-loop calibration: Offer an interactive dashboard for live correction of tone/memory tags, feeding adjustments into a continual learning loop
- Ethics & privacy guardrails: Enforce on-device processing, anonymization, and bias audits to safeguard sensitive emotional data and ensure fair interpretation
- Domain-specific extensions: Craft specialized tag sets and metaphor taxonomies for therapy, customer support, legal mediation, or narrative game design
- Visualization & dashboards: Provide real-time charts of tone Δ, memory strength, and metaphor density; export CSV/HTML reports or integrate with Grafana/Kepler
- Plugin/API ecosystem: Release an SDK with REST and WebSocket endpoints so developers can embed L.L.O.Y.D. in chat platforms, VR/AR experiences, or robotics
- Self-optimizing heuristics: Leverage reinforcement learning on user engagement and sentiment feedback to auto-tune thresholds and periodically retrain lightweight models
- Accessibility & inclusivity: Support customizable sensitivity profiles for neurodiverse users, multi-language emotion schemas, and culturally specific metaphor mappings
- CLI polish: Add a
--no-colorflag and respect the$NO_COLORenvironment variable
🎯 Ultimate goal: Build a language-aware system that doesn’t just respond — it resonates. A model that knows when something cracks, where it swells, and how meaning reverberates over time.