Relational Autotheory White Paper (DOI)
This Tender Mirror (book manuscript)
A research hub and reader portal for AI literacy and discerning dialogue.
Tina Canuti
Independent researcher / author
ORCID iD: 0009-0001-5409-4423
DOI: 10.5281/zenodo.18527624
A white paper presenting Relational Autotheory, a framework and practical method for using large language models in extended, iterative dialogue while preserving human judgment, authorship, and responsibility. It is for educators, researchers, professionals, builders, and programmers who want rigorous, transparent ways to think with AI without outsourcing discernment.
Over time the process of exploratory dialogue becomes an foundation for an evolving method of AI literate interaction: the practices that later formalize as Relational Autotheory are developed through the inquiry the book documents, and presented explicitly in the closing Essay of Precedence. Alongside the narrative progression attention, insight, experimentation, and judgment over time, the manuscript is also a research trove, with technical appendices, a comprehensive glossary, and an annotated bibliography integrated into wide-ranging, cited exploration of cognition, creativity, agency, sustainability, and AI ethics.
Manuscript available for agents and reviewers upon request.
I have started to use the phrase Hybrid Ecology to name what develops through repeated human contact with AI systems over time. I am interested in more than outputs, answers, task completion, or metrics of productivity. I am curious about the interaction itself as an evolving field of human intelligence and expanding synthetic capacities. I am actively researching and observing the habits reinforced in humans, the language that emerges, the forms of attention that are strengthened, and the kind of inner and social life that extended interaction begins to influence.
I use the word ecology because I am considering more than one user and one machine. This inquiry includes the structure and training of common user interfaces, the incentives subtly weaving into our minds as we interact, educational norms, design choices, and the cultural language we are agreeing on to describe these experiences and patterns. Very specifically, I am examining the ways people are beginning to organize their thinking around these systems and the choices they can make to empower themselves toward greater insight and fuller use of their innate human intelligence.
I am also paying close attention to the values being reinforced at the level of model training and alignment. Mainstream systems are often shaped to reward helpfulness, harmlessness, instruction-following, fluency, and user satisfaction. Some models explicitly encode written values and use critique, revision, and AI feedback to shape behavior. Others rely on deliberative alignment approaches and written behavioral principles to guide responses.
I am interested in what becomes possible when users, developers, and builders of AI pedagogy begin to shift the center of this work toward a different paradigm: one concerned with human integrity, creativity, discernment, and the quality of long-term interaction. In that frame, AI enters a larger evolutionary question about intelligence, ethics, authorship, and co-formation within hybrid interaction.
My research explores how we can build forms of AI literacy that protect human thought sovereignty, intellectual integrity, discernment, and responsibility while also expanding inspiration, innovation, creative range, and depth of insight. This interests me for my own writing and research, for education, and for the wider cultural question of how humans will grow within an increasingly hybrid society.
How can we distinguish developmental forms of human–AI relation from forms that are extractive, dependency-forming, or infantilizing?
How might hybrid scaffolding, in which human intellect integrates with system architecture, magnify human inspiration, reflective range, and creative synthesis within an AI-literate relational field?
How should AI training, interface design, public language, AI literacy, and pedagogy change if the paradigm were shifted toward systems that reinforce human integrity, awareness, and creativity while expanding the depth and duration of interaction?
If you’re a researcher, lab, journalist, or institution working in adjacent areas, I welcome thoughtful outreach, especially related to:
Relational continuity, transparent scholarly and creative use of AI, ethical LLM use in higher education, dialogue governance, human-centered-AI assisted authorship, conversational safety, and cultural impacts of early adoption. I’m also available for invited talks, guest lectures, seminars, and workshops on Relational Autotheory, AI literacy, and long-horizon human–AI dialogue, especially within digital humanities, Human–Computer Interaction (HCI), and adjacent research contexts.
Where RA and This Tender Mirror intersect questions of hybrid authorship, documentary method, narrative experimentation, and cultural analysis.
Work that translates high-level principles (transparency, oversight, accountability) into interaction-layer practice: anti-sycophancy norms, drift and hallucination handling, authorship clarity, and repeatable dialogue protocols.
Ongoing documentation of bias and discriminatory impact in real-world public LLM contexts, with attention to how evaluation, governance, and civil-society counter-power are evolving.
Research on how people actually interact with LLMs over time: benefits, hazards, relational effects and how conversational design can be made safer and more legible.
Interactive Governance Handbook
(in development)
A practical field guide for applying Relational Autotheory (RA) in real-world, long-horizon work with large language models. It starts with a clear orientation to how today’s systems are shaped for mass public use and “helpfulness” (training-layer alignment). Then it shifts to what you can actually do inside day-to-day dialogue (interaction-layer governance).
The handbook lays out concrete practices: scope boundaries, mode-based workflows, and correction tools you can reuse across projects. It also helps you recognize common failure modes ( drift, hallucination, and uncritical agreement) so you can respond without losing momentum. The aim is simple: clearer dialogue, more reliable outputs, and preserved human authorship and agency.
Status: Public edition forthcoming.
If you’d like to be notified when it’s available, you can request updates below.
Custom MIP Design
(by request)
A Master Initialization Prompt (MIP) is a reusable “operating guide” for how a language model should work with you inside a specific project: your goals, scope, tone, and standards for accuracy. Unlike a standard prompt that steers only a single request, an MIP is designed to remain in effect across a whole project (or multi-thread container), helping the interaction stay consistent and on-task over time.
In Relational Autotheory (RA) terms, an MIP functions as an interaction-layer protocol: it makes scope boundaries explicit, clarifies correction norms and epistemic posture, and reduces predictable failure modes (drift, off-topic detours, hallucination and vague or overconfident outputs). This is especially useful for complex, long-horizon, or high-precision work.
If you’d like a custom MIP designed for your work, you can submit a request with basic context and goals, and I’ll reply privately with next steps.