The Architecture Began With a Triangle | Michael E. Betker | ResearchPod

Transcript: The Architecture Began With a Triangle

Sam: A polished essay can't tell you how much thinking went into it. One new framework argues that this is the central design problem for AI in learning and reflection. Its claim is that AI should propose patterns while people keep authority over what those patterns mean. To make that work, the human's own reasoning has to stay visible, and the human's rejections have to stay binding.

Alex: Before we get to binding rejections, tell me what I'm listening about. Whose work is this, and what kind of document is it?

Sam: It's a conceptual framework and research agenda from Michael Betker. He's a communication lecturer at the University of Wisconsin–Whitewater and also the founder of Mythos Holdings, the company behind a reflective AI product called Myth-OS. The version we're discussing is dated October 2026. It is not an empirical paper. It draws on one small, exploratory classroom pilot.

Alex: So a company white paper built around its own product. Why should someone outside that product care?

Sam: Because the problem it names is general. Grading and AI-detection both inspect the final artifact. This document argues the artifact can't tell you whether a student reasoned deeply, collaborated well with AI, or simply let the AI do the work. I'll cover the core idea, one concrete example, what the pilot did and didn't show, and who should read the full text.

Alex: Start with what came before. What's the position it builds on?

Sam: The adjacent evidence is a 2026 review by Jauhiainen on AI-supported educational assessment. That work argues AI grading is defensible only inside a governed process. That process includes authorized materials, rubric alignment, repeated scoring, discrepancy review, and continuing human oversight. Betker's move is to push that logic upstream. Before asking whether AI can judge a student's work fairly, ask whether the student's thinking is still present in the work at all.

Alex: So the assessment researchers govern the grading process, and this governs the thinking process.

Sam: That's the extension, and the document is careful to say Jauhiainen doesn't validate Myth-OS. It calls the principle Human Signal Preservation. That means keeping enough evidence of what the person thought before, during, and after AI help to tell augmentation from substitution. In classrooms it becomes a sequence called Human Signal First. The student commits a rough idea, a guess, or a question before any AI enters. Then they encounter the AI, contest what it offers, revise, and synthesize. Later they come back and try to explain the reasoning on their own.

Alex: That sounds like a draft-then-revise assignment with extra steps. What's actually new here?

Two things, I'd say. The first is the longitudinal piece. The system remembers your earlier reflections and returns to them over weeks. The author frames this through Peirce's triangle. That's the philosopher's model of meaning as a relation between a sign, its object, and the interpretation it produces. An interpretation can itself become a sign that later gets reinterpreted. So meaning develops in loops, and an AI with memory can sit inside those loops.

The Architecture Began With a Triangle

Michael E. Betker

Oct 11, 2026·8:00·user_upload

Governed Longitudinal Human–AI SemiosisHuman Signal PreservationAdjudication IntegrityRecursive Relational DynamicsSemiotic IntelligenceCognitive Offloading

Listen

8 min

Summary

Research Question and Framework

Generative AI systems are increasingly capable of providing fluent, personally resonant interpretations of user experience. While these capabilities can support reflection, they also introduce a significant risk: cognitive offloading and the subtle transfer of authority from the user to the machine. The author addresses the question of how computational systems can become useful for meaning-making without becoming sovereign over a person’s judgment. The proposed solution is a governed, longitudinal architecture—instantiated in the Myth-OS framework—where machine-generated pattern hypotheses remain provisional, evidence-linked, and subject to human oversight.

Core Principles of Governed Semiosis

The framework posits that the quality of an AI-assisted outcome is not an indicator of high-quality human cognition. To combat the 'dependence paradox'—where an accurate and fluent system leads users to surrender independent judgment—the paper introduces several governance requirements. These include Human Signal Preservation, which necessitates that evidence of a user's initial reasoning remains identifiable throughout the process, and Adjudication Integrity, the requirement that a user's response to an AI hypothesis (whether they accept, reject, revise, or defer) must remain computationally consequential in all future interactions.

Interaction Design and Agency

To manage this interaction, the framework outlines mechanisms such as Productive Friction, which uses proportionate, evidence-linked challenges to interrupt automatic self-narratives without forcing a specific conclusion. The architecture also emphasizes Provenance, ensuring that source materials, machine-generated inferences, and human dispositions are never collapsed into a single, authoritative narrative. This approach aims to distinguish 'augmentation' (where AI deepens human reflection) from 'substitution' (where AI replaces the underlying cognitive work). The paper suggests that longitudinal continuity is only beneficial if 'difference survives recursion,' meaning that historical disagreement and uncertainty are maintained as permanent parts of the record rather than being smoothed over by an increasingly 'knowing' AI.

Key Terms and Definitions

Human Signal Preservation — A design principle requiring that sufficient evidence of a person's initial reasoning, evaluation, and revision remains visible to distinguish human cognitive participation from machine substitution.
Symbolic Third — An interactional field created by the governed relationship between user-authored experience, machine-generated hypotheses, and human evaluative judgment, where new interpretive possibilities emerge.
Adjudication Integrity — The requirement that the system maintains a clear distinction between a machine’s hypothesis and a human’s disposition of it across time, ensuring that rejections remain consequential for future retrieval.
Productive Friction — A form of calibrated, evidence-linked resistance or challenge that prevents premature closure and encourages independent user evaluation without imposing artificial distress.
Recursive Relational Dynamics — The descriptive principle that each governed interpretive encounter changes the starting conditions for subsequent encounters, making the interaction process developmental rather than static.

The Architecture Began With a Triangle: Research Paper Podcast