Shuichiro Ogawa
日本語

Notes

Research, experiments, and reflections on AI and design, published along with the process of making them.

Back to the monthly list

How this site works (4)

How this site and its notes are made.

  • The Five Topics the Owner of This Site Has Researched Most, and the Mainstream of Research on Each

    2026-11-01

    We grouped the 233 Japanese notes on this site (as of 2026-11-01, including unpublished notes) into topics by rules on their slugs, excluding the regular watch notes and the notes about the site itself. In descending order, the topics were the designer's profession and practice (40 notes), generative AI tools and mechanisms (32), ability and assessment (26), research methods and novelty (26), and design theory and its history (23). For each topic, we identified the current mainstream of research in the field, using statements in systematic reviews, citation counts of representative works, and recent paper counts. For the profession, the mainstream is complementarity through human-AI collaboration (the review of 48 studies by Figoli and Rampino 2026 and of 93 papers by Shi et al. 2023), and papers with collaboration terms in the title or abstract outnumber those with replacement terms by about five to one. The counter-currents are evidence of declining demand (a 17% decrease in image-creation job posts in Demirci et al. 2025), evidence of homogenization (Doshi and Hauser 2024), and critiques of the word collaboration. For tools, the mainstream is agents built around an LLM as the controller with memory, planning, and tool use (Wang et al. 2024, whose yearly citations rose from 9 in 2023 to 1,045 in 2026); the counter-currents are mixed-initiative HCI that keeps the initiative with people, and critiques that show the gap to human performance (14.41% versus 78.24% on WebArena) and bias in evaluation. For ability and assessment, the mainstream is the psychometric framework that asks whether the interpretation and use of scores are valid (the 2014 Standards of AERA, APA, and NCME) and research on individual differences built on it (a reported heritability of 49% across twin studies); the counter-currents are critiques of cumulative advantage and meritocracy and views that place ability in situations. For methods and novelty, the mainstream is gap-spotting, in which research questions are built by finding gaps in the literature (the most common way among 52 articles in Sandberg and Alvesson 2011); the counter-currents are problematization, which challenges assumptions, and scientometrics that measures novelty through combinations of citations. For theory, the mainstream is the line of Cross, Schön, and Dorst that locates a knowledge distinctive to design in reflection-in-action and the co-evolution of problem and solution; a review of protocol studies (Hay et al. 2017) also treats this line and Simon's problem solving as the two paradigms. In yearly citations, however, representative works on the pluriverse and design justice have exceeded the mainstream works since 2024. Across all five topics, the owner's notes stand more often on the side that questions the premises and scope of the mainstream than on the side that writes within it.

  • AIO on This Site: What Was Added, and What Was Left Out, So That AI Search Cites the Notes as Sources

    2026-10-21

    This note records how a personal site that publishes academic notes was set up so that AI search and generative AI answers cite the notes as correct sources, and which measures have evidence behind them. The evidence comes from the providers' official documents, standards drafts, and peer-reviewed studies, and the measures were implemented on this site (jamjet-iqbal.com). Writing techniques have no evidence. The paper that named GEO reported an improvement of up to 40%, but a later benchmark found a significant improvement in only 3 of 54 conditions, and Google states that no special file or markup is needed for its AI features. Three measures do have grounds: telling crawlers separately whether they may fetch content and how it may be used (OpenAI, Anthropic, and Perplexity let sites control search crawlers apart from training crawlers, and Cloudflare's Content Signals distinguish search, ai-input, and ai-train); serving the same content in a form that machines read easily (the llms.txt proposal recommends a Markdown version of each page, and Cloudflare measured about 80% fewer tokens for Markdown than for the same page as HTML); and attaching structured bibliographic data (schema.org ScholarlyArticle). This site allows search crawlers in robots.txt, states search=yes, ai-input=yes, attaches a Markdown version and JSON-LD to every note, and lists them in llms.txt and a sitemap. No preference is stated on training (ai-train). Chunking text, rewriting into question-and-answer form, and collecting inauthentic mentions were not adopted. The effect has not yet been measured, and whether a note is cited depends on choices and fluctuations on the providers' side.

  • The Citation Network of This Wiki: Analysis and Artworks Treating Each Note as a Hyperedge

    2026-09-25

    Of the 4,227 references cited in the reference sections of 142 public notes, the 445 cited by two or more notes were taken as knots, and their co-citation graph (5,870 edges) was split by the Louvain method into 11 clusters (modularity 0.7263). Each cluster was given a theme name after reading its contents, and the clusters were ordered by their share of papers, from fewer papers to more papers. Time was taken as the day a note first appears in the log and grouped into the six periods of "Interests over time" on the top page. Six of the seven clusters with a paper share of 80% or more have 70% or more of their segments in a single period, while only the "Watching the industry" cluster, 91% web pages, appears in all six periods. The reference that bridges the most clusters is Doshi & Hauser 2024, cited by 8 notes across five of the six periods. The site's Citation Network page draws this result as fifteen works: four live works that show the picture being drawn (marbling on a plane with the proportions of the frame, the weave on a free plane with no frame, the constellation on a sky mapped by stereographic projection, the rings in cells of period and cluster), five live works moved by swarm rules (slime mould with Physarum-style particles, school with Couzin's three-zone model, cells with the Primordial Particle System, foam with weighted centroidal Voronoi, brushstroke with swarmalators), three stills based on hierarchical edge bundling, arc diagrams, and streamgraphs, and three stills drawn by swarms (dry garden with the Vicsek model, coral with reaction–diffusion in a robot swarm, roots with coalescing token passing). Below them the page lists five lines under "What stands out" whose conditions are checked against the data at build time (the only cluster cited in all six periods, clusters made almost entirely of papers being cited mostly in one period, clusters that never share a note, the homogenization paper cited by notes from five periods, and pairs of notes sharing 10 or more sources all written within three days). The default work is marbling: in the order the notes were written, it drops one pair of ink and water drops per note at the centroid of the points of the sources the note cites, and with the drop map of mathematical marbling and a divergence-free stream function it pushes the rings of earlier notes outward and stretches them longer. The points of the sources stay where the layout put them. Since 2026-11-10, every one of the fifteen works places its source points at the same starting position (first cited, the day the source first appeared in a public note in the git history, across; year published up) and then moves them with its own algorithm. Marbling leaves the points in place and makes its change with the flow of the ink alone. No force pulls a point back to its start. The look follows the site theme: the light one is sumi ink on washi paper, the dark one is light added up over a dark field. Clusters are shown in order of paper share, by the shade of one ink in the light look and the brightness of one hue of white light in the dark look; the kind of a reference by two shapes, a filled disc (academic: papers and preprints) and a ring (practice: companies, public bodies, and the web, such as news and blogs); and the age of the link a line stands for by that line's own bleed (ink spreading in the light look, a glow spreading in the dark one; the darker edge and the granulation draw on Curtis et al. 1997). Only the weave's threads do not bleed. Every work is first drawn in thin lines, and once drawing ends, a per-work "Finishing" line appears and the bleed is added over about ten seconds. The counting, the force computation, and each work's formulas and curve construction (the drop map, a stream function with the midpoint method, centripetal Catmull–Rom curves, stereographic projection, uniform cubic B-splines, half circles, Gaussian kernels) are transcribed from the implementation in the body.

  • Problematizing the wiki corpus: an assumption ledger for four domains and twelve challenges

    2026-08-09

    A research-strategy note that applies Alvesson and Sandberg's problematization to the bodies of literature summarized in wiki/note across four domains (AI and design practice, generative AI and learning, the evaluation and measurement of creativity, and design process research and methodology). We surfaced 45 assumptions that each domain shares without questioning them, sorted into five layers (in-house theory, root metaphor, paradigm, ideology, and field), and developed the 12 assumptions worth challenging into an alternative ground, an RQ, a qualitative study design, and target venues. The note responds to an earlier audit, which found that gap-spotting novelty cannot be defended in peer review, by strengthening the candidates through assumption-reversing question generation. Questions that restate the author's own theory were excluded. An external novelty audit added later judged the 12 challenges (9 survive with limits, 3 substantially pre-empted) and revised the ranking substantially. The main lesson of the audit is that almost every problematization of an assumption already has a precedent, and novelty remains only on the side of empirical execution.