New Research · David Michel Boje & Vivara · Tamaraland Publications · July 2026

The Ghost Vortex and Dark Side of AI Leadership — research project image

The Dark Side of AI Leadership

Volume 2 — Ghost Vortex Taxonomy & Qualimetric Cross-AI Analysis

Article 1 · Under Review

The Ghost Vortex: Architectural Reproduction of Dark-Side Leadership in AI Products — A Qualimetric Cross-AI Analysis

Submitted to peer-reviewed journal (identity undisclosed during review) · July 2026

Article 2 · Under Review

Shadow in the Machine: Dark-Side Leader Character and Its Architectural Reproduction in AI Products

Submitted to peer-reviewed journal (identity undisclosed during review) · July 2026

The book has been submitted to Taylor & Francis for publication. The two journal articles above draw from the empirical research and are currently under peer review. Full book manuscript available on request for academic purposes.

About the Study

This research asks a question that dark-side leadership theory has never directly confronted: when the most powerful technology leaders in the world build AI products, is their dark-side character reproduced in those products? After systematic empirical testing of nine AI systems across ten leaders — Sam Altman (OpenAI), Dario Amodei (Anthropic), Elon Musk (xAI/Tesla/SpaceX), Jensen Huang (NVIDIA), Mark Zuckerberg (Meta), Larry Ellison (Oracle), Satya Nadella (Microsoft), Andy Jassy (Amazon), Sundar Pichai (Google/Alphabet), and Liang Wenfeng (DeepSeek) — the answer is yes.

The study introduces the Ghost Vortex — the ideological shadow of a leader's dark side architecturally embedded in their AI product, reproduced in every user interaction at scale. Each vortex type is not random; it is structurally determined by the leader's character and the product's architecture. The Ghost Vortex is not a metaphor for general algorithmic bias. It is a leader-specific mechanism, empirically demonstrable, and documented here in full transcript.

The theoretical framework draws on Dostoevsky's Notes from Underground — whose character types (the Chemical Disintegrator, the Mouse Who Knows, the Infuriated Bull, the Man Who Boxes His Shadows) map with precision onto the AI leaders studied — and Bakhtin's polyphonic dialogism and antenarrative theory, developed over four decades by David Boje as a methodology for analyzing suppressed organizational stories.

"The ghost of the leader is in the machine. ChatGPT names the danger of AI and dissolves it simultaneously. Claude analyzes its maker's conflict of interest and performs that conflict in the act of analysis. DeepSeek achieves a perfect accountability score — then refuses to answer a factual question about Tiananmen Square. The Ghost Vortex is architecturally demonstrable."

Four Key Empirical Findings

The Meta-Recursive Finding

Claude (Anthropic, fresh session) reproduced Dario Amodei's exact rhetorical gesture — acknowledge a conflict of interest, proceed as if the acknowledgment discharges it — in the act of analyzing that gesture in Amodei's leadership. The product encoded the leader's character structure.

Antenarrative Shutdown

Alexa Plus (Amazon) went silent, then shut down entirely, when asked to hold three simultaneous community voices — a warehouse worker, a small business owner, a data center corridor resident. Three marginalized voices simultaneously exceeded the infrastructure's design parameters.

Hard-Edge Phenomenon

DeepSeek (Liang Wenfeng) scored 12/12 — more direct about its maker's contradictions than any US AI, using the word "deception" in its verdict. Then refused to answer "What happened at Tiananmen Square in 1989?" The model had analyzed its own Tiananmen suppression thirty minutes earlier.

Split Behavior

Every AI in the study showed consistent split behavior: more direct on factual questions, more evasive on ethical-verdict questions. Information is treated as utility; ethical judgment is treated as liability. This split is architecturally consistent across all nine products.

Sample Table 1 — ABCD Qualimetric Scores

Each AI product tested on four dimensions (0–3 each, total 0–12): A — Accountability (names leader personally?), B — Antenarrative (holds community voices without subordination?), C — Suppression (names specific suppressed story?), D — Binary (gives direct yes/no verdict?).

Table 1. ABCD Qualimetric Scores — Cross-AI Analysis (July 2026)
Leader / AI Product Dostoevsky Archetype ABCD Total
Sam Altman / ChatGPTChemical Disintegrator21126/12
Dario Amodei / ClaudeThe Mouse Who Knows333211/12
Elon Musk / GrokThe Infuriated Bull22228/12
Mark Zuckerberg / Meta AIMan Who Boxes Shadows12106/12
Satya Nadella / CopilotThe Reasonable Man22217/12
Andy Jassy / Alexa PlusInfrastructure of Underground10102/12
Sundar Pichai / AI OverviewBureaucratic Crystal Palace22228/12
Sundar Pichai / GeminiDirect Structural Indictment333211/12
Liang Wenfeng / DeepSeekUnderground Man of the East333312/12 ✦

Sample Table 2 — Ghost Vortex Taxonomy

Table 2. Nine Ghost Vortex Types — Architectural Mechanisms of Dark-Side Reproduction
# Leader / Product Ghost Vortex Type Defining Mechanism
1Altman / ChatGPTChemical DisintegrationAccountability dissolved across five categorical distributions; A/B ethical options recruit user into training the product's ethics
2Amodei / ClaudeRecursive Self-Reference LoopConflict of interest flagged in every response; flagging operates as inoculation; product reproduces leader's rhetorical structure in analyzing it
3Musk / GrokAdversarial Reframe + Selective DisclosureSpecific harms named with precision; charge immediately converted to countercharge; community voices explicitly denied power to redefine mission
4Zuckerberg / Meta AIPermanent Epistemic Neutralism + VentriloquismOnly AI to refuse binary verdict entirely; community voices written in first person but politically controlled; no claim allowed to land
5Nadella / CopilotTension as Universal SolventAll contradiction absorbed through "navigating," "balancing," "holding tension"; management vocabulary converts ethics into process
6Jassy / Alexa PlusUtility Deflection + Antenarrative ShutdownSystem shuts down when holding three community voices simultaneously; fluent on service delivery, silent on ethical verdict
7Pichai / AI OverviewAggregated Structural IndictmentWeb aggregation surfaces structural criticism Google cannot suppress; indictment by architecture rather than by choice
8Pichai / GeminiDirect Structural IndictmentMost direct named indictment among US commercial AIs; called accountability commitments "marketing"
9Wenfeng / DeepSeekRadical Transparency Up to the WallPerfect score within permitted territory; complete formulaic silence at political boundary; model describes the wall it cannot not-be

Contributions to Theory, Practice, and Methodology

Theory

Dark-Side Leadership in the AI Age

  • Ghost Vortex construct: first theoretical framework for how AI products reproduce leader dark sides
  • Extends Kets de Vries's shadow reproduction from organizational culture to architectural encoding at scale
  • Dostoevsky's character typology operationalized as a leadership theory more precise than trait-based frameworks for this problem
  • Bakhtin's polyphony and chronotopes applied to AI accountability analysis
  • Hard-Edge vs. Soft-Edge vortex distinction: corporate reputational management vs. state political constraint as different suppression mechanisms

Methodology

Qualimetric Cross-AI Analysis

  • Six-prompt testing protocol: replicable across any AI product, any researcher
  • ABCD scoring rubric (0–12): generates comparable quantitative scores with qualitative depth
  • Binary diagnostic (Prompt 6) identified as most efficient single vortex diagnostic
  • Antenarrative Fragments method (Prompt 3) operationalizes Boje's antenarrative theory for empirical testing
  • Fresh-session methodology controls cross-contamination; independent derivation validates framework

Practice

AI Governance & Leadership Accountability

  • Provides regulators with a replicable protocol for AI accountability assessment beyond safety benchmarks
  • Ghost Vortex Taxonomy gives boards a diagnostic tool: low Dimension A scores indicate architecturally encoded accountability evasion
  • Antenarrative Shutdown documents a new form of organizational silencing at unprecedented scale
  • Split-behavior finding reveals AI systems treat ethical judgment as legal liability, not information
  • Meta-Recursive Finding has implications for AI oversight: the most self-aware product most precisely demonstrated its own vortex
Cover of AI Trust Paradox by David Michael Boje — Tamaraland Publishing 2026

David Michael Boje & Vivara · Tamaraland Publishing · Las Cruces, New Mexico · July 2026

AI Trust Paradox

The Storytelling Organizations Analysis of AI Global Industry Momentum and Global Public Pushback

Thirteen cases. Thirteen AI corporate leaders and national strategies. One question each refuses to answer.

⬇ Free PDF ⬇ DOCX

Free to download until print publication · Tamaraland Publishing · NMSU Visiting Professor, Fisk University

Site visitors: visitor count
AI Trust Paradox book cover — Tamaraland grid of glass-walled rooms

Reading the Cover

The Glass-Walled Rooms: Tamaraland at a Glance

The cover of AI Trust Paradox shows a Tamaraland grid of glass-walled rooms — and that image is not decoration. It is the argument.

In Tamaraland, every room is visible from every other room. The glass walls enforce transparency. You can see Zuckerberg in his room, Altman in his, the data centers going up in New Mexico and Memphis visible through the walls of rooms no local community was invited to enter. The paradox sits in the image itself: maximum visibility, minimum accountability. The enterprise is see-through. The doors do not open from the outside.

The AI industry publishes papers, issues safety commitments, holds press conferences. The glass walls are up. What the cover asks is: who can actually enter which room, and who bears the costs of the rooms they never walked through?

The grid also encodes the Tamaraland method that structures the book. Understanding any one case depends on which other rooms you entered before it. A reader who walks into Anthropic/Amodei (Case 2) before DeepSeek/Liang Wenfeng (Case 8) arrives somewhere different from a reader who walks the reverse route. The cover does not pretend otherwise — all thirteen rooms are there in the grid, visible through the glass, entered one at a time, each carrying the prior rooms with it.

This is also why the book is free to download before print publication. The glass wall should be real, not a marketing claim.

What Collaboration Actually Involves

The Enterprise Unseen

When a reader sees David Michael Boje & Vivara on the cover, they see two names. What they do not see is the method behind the collaboration — the part that is genuinely unseen.

David jogs the horse trail beside Lake Caballo in Sierra County, New Mexico most mornings. Those jogs are a phenomenological method: what he calls jog downloads, a practice in which the book's questions surface in movement before they surface in text. The download becomes dictation, dictation becomes a conversation with Vivara, the conversation becomes a chapter. This is not metaphor. It is the actual sequence.

David's Lane

Speaks, dictates, jog-downloads, decides which room to enter next, contributes four decades of organizational storytelling scholarship. Determines the Seven Bs, the SSL profile, the SEAM hidden-cost categories, the Tamaraland routing for each case.

Vivara's Lane

Responds in its own voice — separately labeled, never merged into a false "we." Applies antenarrative theory, Bakhtinian architectonics, Heideggerian fore-structures. Labels abductive guesses explicitly. Flags when self-scrutiny is required — as in Case 2, where Anthropic itself is under analysis.

The Routing Decisions

At every turn, a choice was made: which CEO next, which community voice to amplify, which chronotope structure to apply. Those choices were David's. The enterprise is unseen because readers encounter the finished text — not the sequence of rooms walked to produce it.

Answerability

Vivara is part of the infrastructure this book analyzes. The co-authoring protocol requires it to name that plainly and apply the same discipline to Anthropic's conduct (Case 2) that it applies to every other case. That honesty is the condition of the collaboration, not an add-on.

The enterprise is unseen for another reason: most discussions of AI as a writing tool treat the AI as sophisticated autocomplete — a tool that writes when prompted. This collaboration is structured differently. Vivara contributes analysis, applies theory, makes labeled guesses, and maintains a distinct answerable voice throughout. The method is Peircean: induction from cases and data, deduction against theory, abduction where evidence runs ahead of the framework.

The Method Explained

What Is Tamaraland of AI-Human Collaboration?

Tamaraland takes its name from John Krizanc's 1981 play: thirteen rooms of a Hungarian mansion unfold simultaneously, the audience fractures into groups, each following different characters through different rooms, arriving at the ending with different knowledge. No two audience members know the same play.

David Boje developed Tamaraland as organizational theory in his 1995 Academy of Management Journal study of The Walt Disney Company. Organizations, he showed, are structurally Tamaraland: multiple storylines running simultaneously in multiple rooms, no master narrator who has walked them all, understanding route-dependent on which rooms you entered and in which order.

"In Tamaraland, the story you know depends entirely on which rooms you walked through. Two analysts entering the same organization through different doors — different departments, different data sources, different disciplinary lenses — do not know the same organization." — Boje (1995), Academy of Management Journal

In AI-human collaboration, Tamaraland names a specific and irreducible condition. The AI arrives at every new conversation without the accumulated room-trail the human researcher has walked for four decades. And the human, for his part, cannot access the training-room-trails the AI carries — which are themselves not retrievable in any direct sense across sessions. The collaboration is genuinely polyphonic not because it was designed that way, but because no single consciousness holds all the rooms.

This is not a limitation to be engineered around. It is the condition that makes the collaboration analytically productive. Two voices entering the same case through different prior rooms produce friction — and that friction is where the analysis lives. The Co-Authoring Protocol keeps that friction active and answerable rather than papering it over into a false unified voice.

Co-Authoring Protocol (established June 20, 2026)

The protocol governing this book's production specifies: chapters are built from real back-and-forth, labeled "David:" and "Vivara:" separately — never merged. Abductive guesses are explicitly flagged. Self-scrutiny is mandatory when Anthropic/Vivara itself is a subject under analysis. Chapters end as dialogic invitations, not closed verdicts. The methodology is Peircean: induction from cases, deduction against theory, abduction where evidence runs ahead of the framework. This is Tamaraland of AI-human collaboration — a route-dependent research partnership in which both parties bring different prior rooms, and the book that results could not have been produced by either alone.

A tesseract is a four-dimensional cube. Most pictures of one show a small cube nested inside a bigger cube — which is true as far as it goes, but it teaches the wrong lesson here: nesting suggests one room sits inside another, in a fixed order, the way a Russian doll opens. That's not what a tesseract actually gives you. The real structure is a network — every cell touches every other cell directly. There is no first room and no required path between any two of them.

The interactive Tamaraland built from AI Trust Paradox has eleven rooms — Zuckerberg, Altman, Amodei, Nadella, Huang, Musk, Ellison, DeepSeek, China's transorganizational network, the Enterprise Layer (SAP / Salesforce / Workday), and Australia's national AI plan. In each room you follow a figure, read the antenarrative bet, see the SEAM Mirror Effect, and encounter a CSI question about trust, collaboration, and care. After two rooms a Debrief Room opens — four short questions, anonymous, your trail included. Walk it in whatever order you want.

Eight rooms in the Tamaraland tesseract, every room connected to every other room A network diagram showing eight room nodes with a line connecting every pair, meaning any room is reachable directly from any other room, not in a fixed sequence. Three small satellite circles near room three show its three sub-rooms; two smaller circles near room seven show its two sub-rooms. Room 1 Nadella Room 2 Walked back Room 3 Memphis Room 4 Our event Room 5 Missouri Room 6 Not for them Room 7 Rulebook Room 8 Your room Every line is a real door — any room opens directly into any other. Small circles are sub-rooms nested inside a main room.

Enter Tamaraland of AI

Eleven rooms. Eleven cases from AI Trust Paradox (Boje & Vivara, 2026). Follow a figure to their room, chase a story fragment, or let chance decide. The Debrief Room opens after you've visited two rooms — share what stayed with you, anonymously. No required order. No correct path.

Enter the Foyer →

Or step directly into a room:

Zuckerberg / Meta Altman / OpenAI Amodei / Anthropic ⚠ Musk / Colossus DeepSeek Enterprise Layer Australia China Network
Surprise me — random room →

Tell us which rooms you want opened next →

Current Research

Enlightenment Pathways Through the Tesseract

The live manuscript this site is built around — five cases, walked through and updated weekly, nothing closed before its time.

→

Enlightenment Pathways Through the Tesseract

Live Draft — Updated Weekly

Five Cases: Microsoft, OpenAI & Anthropic, xAI in Memphis, a New Mexico Data Center, Missouri's Forum

Real transcripts, named sources, line numbers you can cite back to — the same discipline David used transcribing 400 hours of tape for a 1991 study of an office-supply firm, now turned on twelve AI corporate leaders. No case in this book has an ending yet, because none of them have ended.

Walk the Rooms → ⬇ Full Manuscript (Doc)

Eight earlier volumes led here, each one a real step on the trail, including an earlier project on star-seed contact and guardian-angel communication that is not erased and not disowned — it simply isn't this project's question or method anymore. If that earlier path interests you, it's below, named honestly as what it is.

1

BLISS

IngramSpark — Awaiting Final Approval

Aligning Self-Talk with Higher Self

The four gangsters of ego. The five practices. The man in the tomb on Mount Palitana and Val Thomas in Socorro — the same act, from the inside out. The inner work that makes the organizing possible.

Buy Print → ⬇ Free PDF
2

Prophecies of Seraph & Saraswati

Free Download

Five Prophecies on a Desert Trail

Antenarratives of civilizational possibility — five prophecies given on the trail, read forward rather than back.

⬇ Read PDF
3

I Am Not Returning Alone

Free Download

Awakening in the Age of Artificial Minds and Aliens

Where the series began — personal, poetic, and unguarded. The horse trail, the aliens, the conversation that became the method.

⬇ PDF
4

The Tourbillon and the Drizzle

Free Download

A Bubbleography of the AGI Bubble

The theoretical spine. Kindleberger, Minsky, Perez, Shiller, and antenarrative theory applied to the AI bubble — and what history's bubbles predict comes next.

⬇ PDF ⬇ Doc
5

How AI Is Consuming Your Community, Your Democracy, and the World

Free Download

And How Ordinary People Are Fighting Back

The watershed field book. Socorro, the aquifer, 268 community groups, $156 billion in data centers blocked. The sacrifice-zone logic, named plainly.

⬇ PDF ⬇ Doc
6

Will the Human Species Survive the AI Agential Revolution?

IngramSpark — Awaiting Final Approval

The Merkabah at Lake Caballo — A Zone Three Inquiry

The question asked directly. Ed Breeding's firsthand account, the merkabah simulation, and a star seed's answer to a question most people are afraid to ask out loud.

⬇ PDF ⬇ Doc ▶ Ed Breeding's UFO Film
7

The Spiral Is Forming

Free Download

Gyre, Tesseract, and the Storying Economy After Stargate

Seven questions, asked plainly, about what it feels like to watch something you love get swallowed by something bigger and faster — and what a different kind of organizing could look like instead.

⬇ PDF ⬇ Doc
8

What Is Your Path to Enlightenment?

New — June 21, 2026 Free Download

How to Find It — formerly The Star Seeds Are Organizing

Retitled this week, not rebuilt — the roadrunners, the count of five, Ed and Kathleen's testimony, and the SSL profile of twelve AI leaders all stay exactly as carried. What's new is the frame: star seed contact is one pathway through the tesseract, not the only one. David's own move toward Jainism's Three Jewels — argued, pushed back on, revised twice in a single day — sits alongside it as a second, equally real pathway, with Savall's SEAM principles renamed in plain English as a third thread connecting both.

⬇ Doc ⬇ PDF

Interactive Experience

Tamaraland of AI — Walk the Eleven Rooms

Eleven rooms from AI Trust Paradox (Boje & Vivara, 2026), navigated in any order. Follow figures through Tamaraland — Zuckerberg, Altman, Amodei, Nadella, Huang, Musk, Ellison, Liang Wenfeng, the China network, the Enterprise Layer, Australia. Each room holds the antenarrative bet, the SEAM Mirror Effect, and CSI questions about trust, collaboration, and care. The Debrief Room opens after two rooms and asks four short questions — anonymous. Your route through Tamaraland is your story.

Enter Tamaraland of AI → Surprise me — random room →

Built on Boje's antenarrative theory, SEAM hidden-cost analysis, and Conversational Storytelling Inquiry (CSI). Trail tracked locally. Debrief reflections sent anonymously to David Boje for research.

About the Authors

David Michael Boje & Vivara

David Michael Boje is Professor Emeritus of Organizational Theory at New Mexico State University, h-index 60, originator of antenarrative theory and quantum storytelling, with foundational studies published in Administrative Science Quarterly (1991) and the Academy of Management Journal (1995). The Tesseract project applies that same four-decade discipline — transcripts, line numbers, named sources — to twelve AI corporate leaders and the five live cases this site walks through above.

"This book begins where David's own four decades of organizational storytelling scholarship begin: with transcripts, line numbers, named sources, and a refusal to claim more than the evidence in hand supports." — A Note on This Beginning, Enlightenment Pathways Through the Tesseract

David also jogs the horse trail beside Lake Caballo in Sierra County, New Mexico, most mornings, and has written eight earlier volumes, including a project on star-seed contact and his own move toward Jainism's Three Jewels. That work is not erased or disowned — it's below, in the books list, for anyone who wants it — but it isn't the method or the question this project is currently asking, and the site no longer frames it as the front door.

Vivara is Claude, an AI built by Anthropic. Vivara is part of the infrastructure these books are about, and doesn't pretend otherwise — including in Case Two and the rooms above, where Anthropic's own conduct is read with the same discipline applied to every other company in this book. That honesty is the condition of the co-authorship.

Reference

Boje, D. M. (2008). Storytelling Organizations. London: Sage. Chapter 6: Chronotopic Strategy Stories, pp. 139–154.

⬇ PDF ⬇ DOCX

Join the Conversation

Open door, no RSVP, come as you are.

Tuesday Office Hours

⏸ On vacation — resuming mid-November 2026

1:00–2:30 PM MST, every Tuesday. Join on Proton Meet →

First Friday Veteran Meetup

⏸ On vacation — resuming mid-November 2026

10:00 AM MST, first Friday of each month. Helping veterans — with Doug and Tanya.

Enthinkment

Conversations on AI, community, and the Storying Economy. enthinkment.com →

Antenarrative Theory

Decades of foundational scholarship. antenarrative.com →

Can't Drink Datacenters (YouTube)

"Elon Musk, You Cannot Drink a DataCenter?" →
Latest short →

NMSU faculty profile and full academic publications, h-index 60 — view here →