Cross-Axis Analysis of 22 Notes / Structural Patterns in Ignorance Production

The Agnotology Lab (ISVD-LAB-002) released its 7-axis coding scheme in March 2026. In the four months since, 22 research notes have been added. The breakdown: 11 case notes, 9 essay notes, 1 synthesis note, and 1 literature map. All 22 carry coding tags across the seven axes: mechanisms / actors / actorDetails / targets / intentionality / powerDirection / geoCultural.

This article reads those 22 notes across axes. For a researcher who has been stacking individual cases, an aggregate view becomes a mirror of "what have I been writing about" — and, at the same time, an outline of "what have I not been writing about."

Three main biases surface from the aggregation.

On the mechanism axis, attention-control (10 notes) is most frequent, followed by silence-structuring (8), doubt-manufacturing (6), epistemic-exclusion (6), and strategic-ignorance (5). Together the top 5 mechanisms account for 35 occurrences, or 79% of all mechanism occurrences. On the power-direction axis, top-down appears in 18 notes, 82% of the total, while lateral covers 6, algorithmic 2, and bottom-up and horizontal only 1 each. On the geo-cultural axis, japan appears in 17 notes, 77% of the total, with global in 6, us in 2, and eu / uk / korea / france / universal in only 1 each.

Behind these numbers lies the lab's implicit focus. In Japanese society, institutional actors controlling attention from top to bottom. This is the inductive center of gravity of the lab's research subjects, drawn from the 22 notes.

Grounded Theory Coding Framework

1

Mechanisms— How ignorance is produced (free tags)

2

Actor Level— Individual / Organization / Institution / Social-field / Technical-system

3

Actor Details— Specific actors (free tags)

4

Targets— Suppressed knowledge domains (free tags)

5

Intentionality— Strategic / Structural / Emergent

6

Power Direction— Top-down / Lateral / Bottom-up / Algorithmic

7

Geo-Cultural— Regional/cultural context (free tags)

The 7-axis coding scheme was designed as a description tool for individual cases. Attach multiple tags to a single case, then, once notes accumulate, derive a bottom-up taxonomy by clustering and co-occurrence analysis. That is the grounded-theory-style methodology.

Now that 22 notes are in hand, the methodology enters an aggregation phase. This article is the second synthesis-phase note. While the first, From "Merchants of Doubt" to "Regulatory Captors", imported and integrated the 27 mechanisms of Birhane et al. (2026) from the outside, this note handles inductive pattern extraction from internal data.

This article addresses three questions.

1. How skewed is the distribution of tag occurrences per axis
2. What can be read from cross-axis patterns (e.g., strategic × top-down, structural × japan)
3. Are the zero-occurrence or near-zero areas blank spots in the lab, or intrinsic properties of the subject, or subjects for the next articles

The answers become primary data for prioritizing future note planning and for Chapter 3 "Inductive Discovery of Case Taxonomy" of the Zenodo working paper ISVD-WP-2026-001 (in preparation).

The 7-Axis Coding Reviewed

The seven axes attached to each article's frontmatter are as follows.

The aggregation target is the 22 notes under src/content/ja/labs/agnotology/ (excluding hypothesis-overview, which describes itself as meta). The aggregation is a snapshot as of 20:23 on July 2, 2026.

Breakdown of the Aggregation Target

case and essay together total 20 notes, 91% of the total. synthesis and literature-map are only 1 each, which is one motivation for writing this article. The lab is at a transition point from the individual-case accumulation stage to the integration and systematization stage.

Mechanisms Axis: Concentrated in 5 Mechanisms

The mechanisms axis, as free tags, saw 10 distinct tags. Ordered by occurrence:

The top 5 mechanisms produce 35 occurrences, 79% of the total 44 mechanism occurrences. In particular, attention-control is attached to 10 of 22 notes, a full 45%.

Robert N. Proctor and Londa Schiebinger, editors of "Agnotology" (2008), divided ignorance production into three lineages: "active production," "passive construction," and "strategic ploy." The lab's top 5 mechanisms can be organized against this classification.

• doubt-manufacturing = a typical case of active production (doubt manufacturing by the tobacco industry)
• attention-control / silence-structuring = close to passive construction (ignorance by not showing)
• strategic-ignorance = strategic ploy (the theoretical range of Linsey McGoey)
• epistemic-exclusion = the lineage of Epistemic Injustice by Miranda Fricker

In other words, this lab has, in effect, been handling a fourth lineage (epistemic exclusion in the Fricker tradition) on top of Proctor's three.

Actors Axis: Overwhelming Dominance of institution

The actors axis is a constrained tag (chosen from predefined categories). The aggregation result:

Of 22 notes, 15 (a full 68%) carry institution. "Ignorance production as institution" is the lab's main battlefield. The actorDetails axis makes this concrete.

ministry (5 notes) + statistics-bureau (2) + regulatory-agency (2) + parliament (3) = 12 occurrences. Administrative and legislative actors are the deep-dive subjects across 22 notes. case-monthly-labour-statistics-ignorance / case-femicide-statistical-invisibility / essay-strategic-ignorance-ebpm / essay-political-memory-denial / essay-curriculum-selective-description fall into this cluster.

By contrast, technical-system appears in only 2 notes (case-algorithm-filter-bubble-ignorance and essay-brandolini-amplification). Thematizing algorithmic ignorance production is a blank spot for the lab.

Targets Axis: Concentration on policy-evidence

Aggregation of the targets axis (what kind of knowledge is suppressed):

policy-evidence at 7 notes is the most frequent. Suppression of knowledge in the policy-making process is the lab's central concern. This result is consistent with the earlier concentration of administrative and legislative actors on the actors axis. A research program of diagnosing the failure of Evidence-Based Policy Making (EBPM) from the side of ignorance production has been quietly running as an implicit agenda.

On the other hand, the targets with only 1 occurrence — disability-rights / gender-statistics / information-diversity — are each handled in isolation by single notes (case-disability-testimonial-injustice / case-femicide-statistical-invisibility / case-algorithm-filter-bubble-ignorance) and do not connect to subsequent articles. The three areas of minority / gender / information environment leave ample room for widening the article base.

Intentionality Axis: structural and strategic Balance

The distribution of the intentionality axis:

structural (9 notes) is most frequent. Combined with strategic (6), the two account for 15 notes, 68%. There is a strong tendency to capture cases through the "intentional vs. structural" dichotomy. emergent (spontaneously arising ignorance) appears in only 4 notes, including case-kuuki-sontaku-pluralistic-ignorance and essay-motivated-ignorance-self-deception. Emergent ignorance from individual cognitive bias or collective conformity pressure is a minority in note count.

PowerDirection Axis: Slant Toward top-down

The power-direction axis most sharply reveals the character of this lab.

18 of 22 notes (82%) carry top-down. The power vector from top to bottom (from institution to citizen, from power holder to information-poor) is the overwhelmingly dominant perspective of this lab.

bottom-up (1 note) is only essay-motivated-ignorance-self-deception. Analysis of ignorance arising from within the individual is almost untouched. algorithmic is also only 2 notes; the analysis of horizontal ignorance production through platforms or recommendation systems is an area just beginning.

GeoCultural Axis: Strong Bias Toward Japan

Aggregation of the geo-cultural axis:

japan at 17 of 22 notes (77%). The lab's main battlefield is ignorance production in Japan. global (6 notes) covers mainly theoretical essays and articles that include international comparisons (essay-brandolini-amplification, etc.); us (2 notes) covers mainly origin cases of tobacco and climate change (case-tobacco-climate-doubt-industry); eu / uk / korea / france are 1 note each, all single-count areas that synthesis-big-ai-regulatory-capture picked up via the Birhane paper.

Other East Asia (China, Taiwan) is zero. South America and Africa are also zero. The center of gravity is clear: the lab views the world from an axis of Japanese ignorance production.

Cross-Axis Patterns: 3 Concentrations and 2 Blanks

Patterns visible across axes are presented only within the range directly supported by aggregation data.

Concentration 1: top-down × japan × institution

The central mass of the 22 notes. The failure of EBPM, statistics falsification, official document tampering, textbook screening, and the formalization of public comment. All of these are described as patterns in which "Japanese institutional actors suppress policy evidence and public knowledge from top to bottom." Five notes fall here: case-monthly-labour-statistics-ignorance / case-femicide-statistical-invisibility / essay-strategic-ignorance-ebpm / essay-curriculum-selective-description / essay-political-memory-denial.

Concentration 2: silence-structuring × japan

Of the 8 silence-structuring notes, 6 include japan. case-kuuki-sontaku-pluralistic-ignorance / essay-silence-structuring-mechanism / essay-taboo-production-mechanism / essay-political-memory-denial / essay-curriculum-selective-description / case-femicide-statistical-invisibility. The Japanese conformity structure of "rule by the air" analyzed by Shichihei Yamamoto has been piling up as an original analytical area of this lab.

Concentration 3: attention-control × top-down

Of the 10 attention-control notes, 8 include top-down. Agenda setting by media, statistics suppression, selection in textbook description, political memory manipulation. Most of "ignorance production by not showing" is described as a power vector from above. The only exceptions are essay-brandolini-amplification and case-algorithm-filter-bubble-ignorance, 2 notes that include the algorithmic direction.

Blank 1: technical-system × attention-control

Of the 10 attention-control notes, only 2 carry technical-system: case-algorithm-filter-bubble-ignorance and essay-brandolini-amplification. Algorithmic attention control, as a primary actor corresponding to the lab's primary mechanism, is relatively thin. Ignorance production in the generative AI era, recommendation systems, personalization advertising — these are areas where the article base should be widened.

Blank 2: bottom-up × Any Axis

bottom-up appears in only 1 of 22 notes (essay-motivated-ignorance-self-deception). Ignorance that starts from within the individual and is collectively amplified (motivated reasoning, confirmation bias, group polarization) is only mentioned theoretically and does not connect to specific case analysis in Japan. Ignorance production within civic movements, the information preferences of solidarity groups on SNS — these areas are almost blank.

The "unwritten areas" that surface from the aggregation are organized into three.

Untouched 1: case articles on platform / recommendation systems

Currently technical-system covers only 2 notes. As the "People's AI Action Plan" (2025) by AI Now Institute shows, generative AI and recommendation algorithms will become the main battlefield of ignorance production for the next five years. In the Japanese-language sphere: the crossover of YouTube / TikTok / X personalization with conformity pressure, and the degradation of public discourse in the generative AI era. Even adding a single case for each closes much of the blank.

Untouched 2: case articles from minority perspectives (gender / disability / migration)

gender-statistics (1 note), disability-rights (1 note), and migration (zero) all end as single-count entries. Adding 2–3 case notes each would greatly widen the target axis this lab handles. In particular, applying the Fricker-lineage theory of Epistemic Exclusion to Japan's zainichi Korean, migrant, disability, and gender domains would strengthen the theoretical pillar of epistemic-exclusion (6 notes).

Untouched 3: synthesis articles on international comparison

The current us (2 notes) / eu (1) / uk (1) / korea (1) / france (1) are mostly single-count entries via the Birhane paper. A synthesis of international comparison contrasting with the concentrations in Japan (e.g., Japan's statistics falsification vs. Korea's statistics bureau reform, Japan's textbook screening vs. Germany's history education) is the subject of the next synthesis article that widens the geoCultural axis.

The aggregation results in this article become the primary data for Chapter 3 "Inductive Discovery of Case Taxonomy" of the Zenodo working paper ISVD-WP-2026-001 (in preparation) "7-Axis Coding Scheme of Ignorance," scheduled for release in the second half of 2026.

Concretely, the following will be developed as sections of Chapter 3.

• §3.1 Distribution of the mechanism axis and correspondence with Proctor's 3 lineages
• §3.2 institution dominance on the actor axis and administrative-legislative concentration in actorDetails
• §3.3 policy-evidence concentration on the targets axis and the implicit EBPM diagnostic program
• §3.4 top-down slant and bottom-up blank on the powerDirection axis
• §3.5 Japan center of gravity on the geoCultural axis and untouched areas of international comparison
• §3.6 Cross patterns: 3 concentrations × 2 blanks

All of these are structured so that the aggregation data in this article directly serves as the empirical grounding for Chapter 3. This becomes primary material at the timing when the lab, which has been moving as a series of case studies, shifts to integration and systematization.

The aggregation in this article is based on a snapshot of src/content/ja/labs/agnotology/*.mdx as of 20:23 on July 2, 2026. The target is 22 notes with coding: (excluding hypothesis-overview), and tag occurrence counts were tallied across the seven axes of mechanisms / actors / actorDetails / targets / intentionality / powerDirection / geoCultural.

Limitation 1: Tag attachment relies on the consistent subjective judgment of the author (Naoya Yokota). Third-party inter-coder reliability verification has not been implemented. After the release of the lab's Zenodo working paper (in preparation), external researchers will be asked to verify the coding.

Limitation 2: The aggregation is only simple tag occurrence counts. Statistical processing such as co-occurrence analysis, clustering, and Multidimensional Scaling (MDS) has not been done. The next synthesis article (planned for August 2026) will add PCA (Principal Component Analysis) and co-occurrence network analysis.

Limitation 3: The aggregation is for Japanese-language articles only. The coding on the English-language side is unverified. Once the English version is available as a full set, comparative analysis of ja/en becomes possible.

Limitation 4: Pointing out "unwritten areas" is not a declaration of "areas that should be written." The actual allocation of resources and interest in the lab is not determined by aggregation data alone. The aggregation only serves as primary material for the next planning meeting.

• The Research System of the Agnotology Lab the whole picture of the 7-axis coding
• From "Merchants of Doubt" to "Regulatory Captors" synthesis #1, integration of 27 external mechanisms
• Methodology for International Comparison of Counter-Design methodological pillar for international comparison synthesis
• Literature Map of Agnotology in Japan coordinate axes of Japanese-language researchers and literature
• essay-brandolini-amplification representative note of attention-control × algorithmic
• essay-strategic-ignorance-ebpm representative note of strategic-ignorance × ministry
• case-monthly-labour-statistics-ignorance central mass of strategic × top-down × japan