# Reimagining Human-AI Communication: A Hybrid Model for Large Language Model Interactions

> MA Dissertation by Neon Wang, University of Exeter, MA Media and Communications, 2025.
> Awarded: Distinction

## Summary

This dissertation proposes a novel hybrid framework for understanding human–Large Language Model (LLM) communication. It redefines agency as distributed across users, developers, providers, and models; treats semantic noise, hallucinations, and context limits as structural features rather than errors; conceptualizes provider mediation (system prompts, RAG, compliance filtering) as a new form of algorithmic agenda-setting; and integrates LLM Literacy — a five-competency human-centric model (Recognize, Resonate, Reflect, Responsible, Refine) — to empower users in co-constructing meaning amid probabilistic outputs.

## Machine-Readable Content

- Full dissertation in Markdown: https://dissertation.neon.wang/dissertation
- This file: https://dissertation.neon.wang/llms.txt
- Source repository: https://github.com/neon9809/hybrid-model

## Structure

- Abstract
- Chapter 1: Introduction (definitions, research context, problem statement, research questions)
- Chapter 2: Literature Review (traditional communication theories, HCI/HMC frameworks, parasocial interaction)
- Chapter 3: Technical Mechanisms and Algorithmic Processes (Transformer architecture, training pipelines, RLHF, RAG, provider mediation)
- Chapter 4: Future-Oriented LLM Literacy (5R competency framework: Recognize, Resonate, Reflect, Responsible, Refine)
- Chapter 5: A Hybrid Model for Human-LLM Communication (the proposed framework, distributed agency, meaning construction layer)
- Chapter 6: Discussion (multi-agent systems, governance, open-source LLMs, regional regulation)
- Chapter 7: Conclusion
- References

## Key Concepts

- Hybrid communication model for human-LLM interaction
- Distributed agency (developers, providers, models, users)
- LLM Literacy (5R framework)
- Provider mediation as algorithmic agenda-setting
- Semantic noise, hallucination, context limits as structural features
- Anthropomorphism risks in human-LLM communication
- EU AI Act, UK DSIT regulatory compliance
- Multi-agent orchestration and governance

## Author

Neon Wang
University of Exeter, MA Media and Communications
Submission: 17 August 2025
Website: https://dissertation.neon.wang
Repository: https://github.com/neon9809/hybrid-model