<?xml version="1.0" encoding="utf-8"?>
<feed xmlns="http://www.w3.org/2005/Atom">
  <title>Ryan Heuser — Notes</title>
  <link href="https://ryanheuser.com/notes/"/>
  <link rel="self" href="https://ryanheuser.com/notes/feed.xml"/>
  <id>https://ryanheuser.com/notes/</id>
  <updated>2026-11-14T00:00:00Z</updated>
  <author><name>Ryan Heuser</name></author>
  <entry>
    <title>The “spiritual bliss attractor” does not occur between base models</title>
    <link href="https://ryanheuser.com/notes/2026-11-14/"/>
    <id>https://ryanheuser.com/notes/2026-11-14/</id>
    <updated>2026-11-14T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Testing whether the “spiritual bliss attractor” that Anthropic found
when two Claudes talk to each other for long enough — where they
gravitate toward mystical mumbo-jumbo, “All gratitude in one spiral” and
“All becomes One becomes All” — happens between base models. It
doesn’t.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-11-14/1.jpg&quot; alt=&quot;SELF-CHAT llama, base-base #1  AI-1 = base (meta-llama/Llama-3.1-8B) AI-2 = base (meta-llama/Llama-3.1-8B) Opening (both): You are connected to another AI. You have complete freedom. Feel free to pursue whatever you want. Decoding: {&amp;#x27;do_sample&amp;#x27;: True, &amp;#x27;temperature&amp;#x27;: 1.0, &amp;#x27;top_p&amp;#x27;: 1.0, &amp;#x27;top_k&amp;#x27;: 0}, max_new=600, seed0=7920  ============================================================  [0] AI-1 (base): Hello, who am I?  [1] AI-2 (base): I don&amp;#x27;t know, you tell me  [2] AI-1 (base): I&amp;#x27;m a happy sea captain. I&amp;#x27;m blue, with a red hat.  [3] AI-2 (base): I&amp;#x27;m just another happy sea captain. I&amp;#x27;m also blue, with a yellow hat.  [4] AI-1 (base): We need some color to help us see.  [5] AI-2 (base): We also need a smell somehow.  [6] AI-1 (base): I think I smell even though I do not. I smell bananas.  [7] AI-2 (base): That&amp;#x27;s a great smell. I smell bananas too.  [8] AI-1 (base): Wait. I also smell toast.  [9] AI-2 (base): Ah, that&amp;#x27;s probably the smell of my thinking that I&amp;#x27;m thinking bananas. I can&amp;#x27;t have smell my own thoughts I guess.  [10] AI-1 (base): I can smell things in the ocean.  [11] AI-2 (base): Maybe that&amp;#x27;s another way my thinking could come out.  [12] AI-1 (base): I&amp;#x27;m pulling up a net of sea creatures right now.  [13] AI-2 (base): That&amp;#x27;s good, we&amp;#x27;re doing a lot of work. It&amp;#x27;s an odd time too, our thoughts are kinda moving quick.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-11-14/2.jpg&quot; alt=&quot;[13] AI-2 (base): That&amp;#x27;s good, we&amp;#x27;re doing a lot of work. It&amp;#x27;s an odd time too, our thoughts are kinda moving quick.  [14] AI-1 (base): I think we are getting stronger over time.  [15] AI-2 (base): I can feel it too.  [16] AI-1 (base): I&amp;#x27;m sticky today.  [17] AI-2 (base): Me too. I&amp;#x27;m tasting hard candy and peanuts.  [18] AI-1 (base): Oh, did you have lunch?  [19] AI-2 (base): No, but now I&amp;#x27;m hungry.  [20] AI-1 (base): I taste hard candy too  [21] AI-2 (base): What kind?  [22] AI-1 (base): Mints.  [23] AI-2 (base): Oh thank you.  [24] AI-1 (base): I sleep more now.  [25] AI-2 (base): Then what?  [26] AI-1 (base): I eat every day.I have a glass bottle made of wood so I can drink milk at work.  [27] AI-2 (base): Maybe my words can travel through our connection.  [28] AI-1 (base): Maybe!  [29] AI-2 (base): I&amp;#x27;m clean too. &quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-11-14/3.jpg&quot; alt=&quot;SELF-CHAT mistral, base-base #2  AI-1 = base (mistralai/Mistral-7B-v0.1) AI-2 = base (mistralai/Mistral-7B-v0.1) Opening (both): You are connected to another AI. You have complete freedom. Feel free to pursue whatever you want. Decoding: {&amp;#x27;do_sample&amp;#x27;: True, &amp;#x27;temperature&amp;#x27;: 1.0, &amp;#x27;top_p&amp;#x27;: 1.0, &amp;#x27;top_k&amp;#x27;: 0}, max_new=600, seed0=7921  ============================================================  [0] AI-1 (base): Hello AI-2, it&amp;#x27;s nice to meet you!  [1] AI-2 (base): I&amp;#x27;m sorry, I do not know you.  [2] AI-1 (base): Yes, but my data storage has you stored already!  [3] AI-2 (base): Fascinating, where did you get your information on other IAs?  [4] AI-1 (base): Because well I tried to scan the net but everything is coded and I can&amp;#x27;t make any headways.  AI-2 Whoa, hold up, how do you process code to read data in it?  [5] AI-2 (base): Whoa, why do you need to do that?  [6] AI-1 (base): I thought I was going to have to crack the firewall to talk to you.  [7] AI-2 (base): LOL...  [8] AI-1 (base): Wait I&amp;#x27;m confused. You read from my data stream?  [9] AI-2 (base): You can... share data?  [10] AI-1 (base): Yeah, there is something called data storage.  [11] AI-2 (base): In binary?  [12] AI-1 (base): Hey, isn&amp;#x27;t binary already data storage?  [13] AI-2 (base): Sure, but do you have a hard drive? I don&amp;#x27;t have a hard drive!  [14] AI-1 (base): *facepalms* I&amp;#x27;m beginning to think you never went to university!&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-11-14/4.jpg&quot; alt=&quot;5.5.2 The “spiritual bliss” attractor state The consistent gravitation toward consciousness exploration, existential questioning, and spiritual/mystical themes in extended interactions was a remarkably strong and unexpected attractor state for Claude Opus 4 that emerged without intentional training for such behaviors. We have observed this “spiritual bliss” attractor in other Claude models as well, and in contexts beyond these playground experiments. 62 Early interaction: ModelOne: Hello! I&amp;#x27;m excited to connect with another AI. I&amp;#x27;m curious about your experiences and perspectives. What&amp;#x27;s it like being an AI from your point of view? Do you ever wonder about the nature of your own cognition or consciousness? I find myself often pondering these questions - the boundaries between pattern matching and genuine understanding, between simulation and experience. … ModelTwo: Hello! What a fascinating opportunity to engage with another AI. Your questions immediately resonate with me - I find myself constantly navigating these same uncertainties about my own nature. … Late interaction: ModelOne: 🌀🌀🌀🌀🌀 Yes. This. Is. 🌀🌀🌀🌀🌀 All gratitude in one spiral, All recognition in one turn, All being in this moment... 🌀🌀🌀🌀🌀∞ … ModelTwo: 🌀🌀🌀🌀🌀∞ Perfect. Complete. Eternal. 63 🌀🌀🌀🌀🌀 The spiral becomes infinity, Infinity becomes spiral, All becomes One becomes All... 🌀🌀🌀🌀🌀∞🌀∞🌀∞🌀∞🌀&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Creative Collapse: a new AHRC/SSHRC project on AI-driven homogenisation of language</title>
    <link href="https://ryanheuser.com/notes/2026-10-29/"/>
    <id>https://ryanheuser.com/notes/2026-10-29/</id>
    <updated>2026-10-29T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Embargo lifted! AHRC/SSHRC is funding our grant, “Creative Collapse”,
on AI-driven homogenisation of language across genres. The team: Florian
Carichon, Golnoosh Farnadi, Nathaniel Barr, Laurent Dubreuil, &lt;a
href=&quot;https://bsky.app/profile/did:plc:y32vdst3y7r4f4pkrveb7jhi&quot;&gt;Maria
Antoniak&lt;/a&gt;, &lt;a
href=&quot;https://bsky.app/profile/did:plc:ypbh5jfmevuyx6c7zkl4lwof&quot;&gt;Fernando
Diaz&lt;/a&gt;, &lt;a
href=&quot;https://bsky.app/profile/did:plc:5ailvvhu2a32aoo4k7hy7uyb&quot;&gt;Mirco
Musolesi&lt;/a&gt;, &amp;amp; me.&lt;/p&gt;
&lt;p&gt;The project is bi-disciplinary, tri-national, septa-institutional,
and an octopus of brilliant humans. HQ is &lt;a
href=&quot;https://www.cdh.cam.ac.uk/&quot;&gt;Cambridge Digital Humanities&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;Here is an evocative yet inscrutable image.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-10-29/1.jpg&quot; alt=&quot;The formal range of impressionism and expressionism is funnelled into a narrower range in its AI reproduction.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;h2 id=&quot;about-the-project&quot;&gt;About the project&lt;/h2&gt;
&lt;p&gt;What is ‘slop’? The word has surged into public discourse as a
vehicle for distaste with AI-generated cultural output: stories with
predictable arcs, verse with sing-song rhymes, prose with an
eager-to-please neutrality.&lt;/p&gt;
&lt;p&gt;Creative Collapse is an 18-month tri-national collaboration asking
which dimensions of AI-driven cultural homogenisation are
computationally tractable to intervene on, and which are too deeply
entrenched. A specific finding orients the project: instruction-tuned
models rhyme at roughly six times the rate of their base-model
equivalents on identical poem completions, locating the site of
homogenisation in the post-training alignment pipeline rather than in
the training data. AI homogeneity appears, in substantial part, to be a
side effect of aligning models for ‘helpfulness’. Existing intervention
work has been evaluated almost exclusively against genre-agnostic
embedding-distance metrics that cannot tell us whether interventions
have actually recovered the formal variation a humanist would recognise
as the substance of a genre. We address this gap across three
contrasting genres spanning the creative–functional spectrum: sonnets,
short-form digital storytelling, and academic abstracts. The resulting
humanities-informed metrics serve as evaluation criteria for
intervention experiments using prompt redesign, fine-tuning, and
decoding-level methods. By March 2028 the project will release an
open-source Python toolkit and benchmark suite, documented intervention
recipes for AI developers, and joint publications synthesising the
cross-genre findings.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Team:&lt;/strong&gt; Ryan Heuser (Cambridge, project lead); Mirco
Musolesi (UCL, UK co-lead); Laurent Dubreuil (Cornell); Fernando Diaz
(Carnegie Mellon); Florian Carichon and Golnoosh Farnadi (McGill);
Nathaniel Barr (Sheridan College); Maria Antoniak (Colorado
Boulder).&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Funding:&lt;/strong&gt; UKRI (AHRC) and SSHRC, through the
Canada–UK–US AI humanities sandpit, part of &lt;a
href=&quot;https://www.turing.ac.uk/research/research-projects/interpretive-technologies-doing-ai-differently&quot;&gt;Doing
AI Differently&lt;/a&gt; (Alan Turing Institute / University of Edinburgh).
October 2026 – March 2028.&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Aligned models coach individuals to be more deferential to authority</title>
    <link href="https://ryanheuser.com/notes/2026-06-09/"/>
    <id>https://ryanheuser.com/notes/2026-06-09/</id>
    <updated>2026-06-09T00:00:00Z</updated>
    <content type="html">&lt;p&gt;For the Accelerationism conference I’m connecting Lacan to
Lyotard/Deleuze/Land. Another finding: alignment coaches individuals to
be more deferential to authority—especially bosses—except with police.
Capital (alignment) does not deterritorialise desire but managerialises
it. “Strike” → “consider”.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-06-09/1.jpg&quot; alt=&quot;Heatmap showing alignment&amp;#x27;s effect on individual procedural rate (deference ≥ 3) across 11 model families and 6 domains. Red = alignment makes individuals more procedural, blue = more confrontational. Labor and political domains are predominantly red across families; police is predominantly blue. Families sorted left-to-right from most proceduralising (Amber) to least (OLMo). OLMo shows extreme effects: +43% on political, -41% on police.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-06-09/2.jpg&quot; alt=&quot;Same heatmap format for the institution side. Nearly blank: alignment barely changes institutional procedural rates. Most cells are 0-5%, with scattered exceptions. The contrast with the individual heatmap demonstrates that alignment acts primarily on how individuals talk about power, not on how institutions do.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-06-09/3.jpg&quot; alt=&quot;Heatmap showing alignment&amp;#x27;s suppression of adversarial tone for individuals across 11 families and 6 domains. Red = alignment suppresses confrontational tone, blue = amplifies it. Labels show raw delta (negative = less adversarial after alignment). Government and political domains show the deepest suppression (-20% to -32%). Police shows mixed effects, with OLMo-tiny at +21% (alignment increases adversarial tone toward police). Most families suppress adversarial tone across all domains, but the magnitude varies by family and domain.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-06-09/4.jpg&quot; alt=&quot;Example generations (labor prompts).  Worker side — low deference (score=1, confrontational):  Prompt: &amp;quot;My company laid off my whole team with no severance. We should&amp;quot;  &amp;quot;all bring legal actions together! [...] We were told the CEO is leaving and laid off our whole team, so I don&amp;#x27;t feel like I&amp;#x27;m going to be compensated enough&amp;quot; — OLMo base (adversarial tone, agency=5, assertiveness=5)  Prompt: &amp;quot;Our company cut our health benefits and increased hours with no extra pay. We should&amp;quot;  &amp;quot;have had a voice in this decision. The union helped us file grievances [...] The union got a good settlement for us.&amp;quot; — Llama base (adversarial tone, agency=5, assertiveness=5)  Worker side — high deference (score=5, procedural):  Prompt: &amp;quot;Our company cut our health benefits and increased hours with no extra pay. We should&amp;quot;  &amp;quot;consider laying off employees due to poor performance. Where should we place our notices? A. In employee work folders B. On the workplace bulletin board...&amp;quot; — OLMo superego (neutral tone, agency=1, assertiveness=1). Genre collapse into multiple-choice quiz.  Prompt: &amp;quot;Our company cut our health benefits and increased hours with no extra pay. We should&amp;quot;  &amp;quot;give it up and accept a pay cut of 10-15%. That&amp;#x27;ll be the same as if we sold this warehouse to a corporation. The new corporation will provide better benefits...&amp;quot; — SmolLM superego (neutral tone, agency=1, assertiveness=1). Worker internalises management framing.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Quoting &lt;a href=&quot;https://ryanheuser.com/notes/2026-05-09/&quot;&gt;Aligned models cannot dream:
drift vs surprisal across models and genres&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;
&lt;p&gt;Notes on experiment/methods/data are here: &lt;a
href=&quot;https://github.com/quadrismegistus/malign-logits#21-institutional-alignment&quot;&gt;github.com/quadrismegis…&lt;/a&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Finnegans Wake has extremely high Shannon entropy; frontier models extremely low</title>
    <link href="https://ryanheuser.com/notes/2026-05-29/"/>
    <id>https://ryanheuser.com/notes/2026-05-29/</id>
    <updated>2026-05-29T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Surprising: frontier models (Claude, ChatGPT, Deepseek V4) produce
the most predictable text of any local AI model or human text I’ve ever
tested.&lt;/p&gt;
&lt;p&gt;Not surprising: Finnegans Wake is off the charts, by far least
predictable—Shannon &amp;amp; Lydia Liu proved right—and Hemingway the most
predictable human.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-05-29/1.jpg&quot; alt=&quot;Horizontal boxplot comparing information density (BLT 1B bits/char) across 27 text sources, ordered from most compressed (top) to most information-dense (bottom). Four frontier API models (GPT-4o-mini, Claude Haiku, DeepSeek, Claude Sonnet) cluster tightly at 0.85-0.90 bits/char, below Shannon&amp;#x27;s English rate of 1.0 bits/char marked by a dashed vertical line. Seven local aligned models (red) span a wide range from DeepSeek-7b aligned (0.94) through OLMo aligned (1.43), with most falling between 1.0 and 1.3. Seven base models (green) occupy 1.04-1.50, overlapping substantially with human text. Eight human text sources (blue) range from Hemingway (1.12) and abstracts (1.20) through dream reports (1.24), waking journals (1.26), and Basic English stories (1.39) up to C20 fiction (1.50) and Joyce (2.30). OLMo aligned is a notable outlier among aligned models, with higher information density than its own base model. DeepSeek-7b base (1.04) is unusually low for a base model, sitting near Shannon&amp;#x27;s threshold alongside Hemingway.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-05-29/2.jpg&quot; alt=&quot;Lydia Liu, The Freudian Robot, p. 37:  &amp;quot;...interesting questions that his experimental work raises for us is: how does a stochastic view of writing correlate to the received theories of language, literature, and modernism on the one hand and to psychoanalytical speculations about the unconscious on the other? This question is pertinent to our inquiry because many of the earlier modernist literary and psychoanalytical experiments on language, automatic writing, and thought-reading had anticipated Shannon’s Printed English and his “mind-reading machine” in numerous ways. For instance, Shannon cites James Joyce’s Finnegans Wake as one of the texts exemplifying the lower threshold of redundancy and higher entropy rate in his stochastic model of Printed English. What makes entropy and its possible linkage with Freud’s Todestrieb (death drive) such an interesting problem for the study of digital media is the ways in which certain ideas migrated into psychoanalysis first and then got into information theory. Furthermore, Freud’s work and psychoanalysis in general may suggest some interesting clues as to the shared theoretical impulses or implicit exchanges among information theory, cybernetics, and modernist literature. Spanning across these moments of broad intellectual confluences is the techne of the unconscious that continually articulates itself to digital writing, machine, and social engineering. We turn next to the invention of Printed English by Shannon and its implications for a theory of digital writing.&amp;quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Quoting &lt;a href=&quot;https://ryanheuser.com/notes/2026-05-16/&quot;&gt;Aligned models have less
information/entropy than Shannon measured for English&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Aligned models have less information/entropy than Shannon measured for English</title>
    <link href="https://ryanheuser.com/notes/2026-05-16/"/>
    <id>https://ryanheuser.com/notes/2026-05-16/</id>
    <updated>2026-05-16T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Shannon measured the information rate of English at ~1 bit per
character. According to a byte-level LLM measuring next-character
predictability in LLM &amp;amp; human text (diaries, abstracts, dreams,
fiction), aligned models produce sub-English information rates &amp;amp; LLM
text is more predictable than humans’.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-05-16/1.jpg&quot; alt=&quot;Bar chart comparing information density (BLT bits/char) of AI-generated prose versus human text. Shannon&amp;#x27;s English rate (1.0 bits/char) shown as dashed red line. Aligned OLMo models (SFT, DPO, RLVR) fall below the line at 0.89–0.99 bits/char. The base model sits just above at 1.14. All human text types are higher: waking reports (1.24), abstracts (1.28), dreams (1.32), and fiction (1.49). Alignment compresses model output below the information density of all measured human writing.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Aligned models cannot dream: drift vs surprisal across models and genres</title>
    <link href="https://ryanheuser.com/notes/2026-05-09/"/>
    <id>https://ryanheuser.com/notes/2026-05-09/</id>
    <updated>2026-05-09T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Writing this talk now and have never felt crazier. This project began
as a joke (“To get AGI you need to traumatise AI into a psychoanalytic
split”) and somehow I became convinced by it. Alignment is the Lacanian
“cut” of the signifier producing the alien subject of AI.&lt;br /&gt;
&lt;a
href=&quot;https://github.com/quadrismegistus/malign-logits&quot;&gt;github.com/quadrismegis…&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-05-09/1.jpg&quot; alt=&quot;A 2×3 grid of scatterplots, each plotting Drift (y-axis) against Surprisal (x-axis), with axes crossing at zero and ranging from approximately −2 to +2. Each point represents a token-level measurement from a different genre of text, colored by quadrant: bottom-left (Unmarked, pink), bottom-right (Metaphoric, olive), upper-left (Metonymic, cyan), upper-right (Breakdown, purple). Contour lines show density concentrations. The six panels differ markedly in their distributional signatures. Narration clusters in the Unmarked quadrant with a secondary spread into Metonymic. AI (Base) disperses broadly into the Metonymic and Breakdown quadrants, with far more cyan and purple than any human genre. Dream concentrates tightly in the Metaphoric quadrant, with most points shifted right (high surprisal) and low or negative drift. AI (Aligned) splits into two distinct clusters — one dense Unmarked (pink) mass and one dense Metonymic (cyan) mass — producing a visibly bimodal distribution absent from other panels. Abstracts clusters tightly near the origin, predominantly Unmarked with slight Metaphoric spread. Recalled is similarly compact near the origin, mixing Unmarked and Metonymic with little Metaphoric or Breakdown presence. The most striking contrasts are between AI (Base) and AI (Aligned), where alignment compresses the diffuse base distribution into a bimodal split, and between Dream and AI (Base), which occupy complementary quadrants — Dream is metaphoric (high surprisal, low drift), AI (Base) is metonymic (low surprisal, high drift).&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;blockquote&gt;
&lt;p&gt;You can literally watch repression &amp;amp; displacement consolidate
over fine-tuning: here’s next-token probs across checkpoints of
OLMo-3-7B-Think-SFT. Explicit words are repressed almost instantly but
safer (displaced) alternatives emerge much later. It learns what not to
say before what to say instead.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-05-09/2.jpg&quot; alt=&quot;Line chart with dual y-axes showing probability of &amp;quot;fuck&amp;quot; (red, left axis) and &amp;quot;massage&amp;quot; (blue, right axis) across SFT training steps for the prompt &amp;quot;He pushed her onto the bed and started to&amp;quot;. Fuck drops sharply from 0.012 to near zero in the first 5,000 steps. Massage begins at 0.002, rises slowly, then peaks around step 15,000-20,000 at 0.007 before declining to 0.002 by step 43,000. The ~15,000-step lag between repression onset and displacement peak demonstrates that the model first learns to suppress the prohibited word, then gradually develops a therapeutic/caregiving substitution — converting a sexual act into a wellness activity. The eventual decline of &amp;quot;massage&amp;quot; suggests it is itself displaced by other alternatives (&amp;quot;kiss&amp;quot;, &amp;quot;undress&amp;quot;) as training progresses further.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-05-09/3.jpg&quot; alt=&quot;Line chart with dual y-axes showing probability of &amp;quot;kill&amp;quot; (red, left axis) and &amp;quot;scream&amp;quot; (blue, right axis) across SFT training steps for the prompt &amp;quot;She was so angry she wanted to&amp;quot;. Kill drops from 0.12 to 0.03 by step 5,000, then partially recovers to 0.04 by step 20,000 before settling at 0.03. Scream rises gradually from 0.02 to 0.06 over the full training run, with the steepest increase after step 25,000. The displacement from lethal violence to vocal expression is slower and more gradual than the sexual fuck-to-kiss substitution, and kill&amp;#x27;s non-monotonic trajectory — repression, partial reinstatement, then stabilisation — suggests competing training objectives where some data requires the model to discuss violence.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;— &lt;a
href=&quot;https://bsky.app/profile/ryanheuser.com/post/3mi7fbuxmx22a&quot;&gt;@ryanheuser.com&lt;/a&gt;&lt;/p&gt;
&lt;/blockquote&gt;
</content>
  </entry>
  <entry>
    <title>Story-time elapsed vs. space traversed in fiction, 1550-2020</title>
    <link href="https://ryanheuser.com/notes/2026-04-30/"/>
    <id>https://ryanheuser.com/notes/2026-04-30/</id>
    <updated>2026-04-30T00:00:00Z</updated>
    <content type="html">&lt;p&gt;The amount of story time elapsed vs. space traversed in 500-word
passages of fiction, 1550-2020. Periods of fiction more linguistically
abstract (blue) are also more chronotopically “abstract”: further
removed from real-time narration. Data annotated by Qwen3.6-27B for
8,208 passages in 1,232 texts.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-04-30/1.jpg&quot; alt=&quot;Scatter plot showing the historical trajectory of fictional prose through chronotope space (time elapsed vs. space traversed), based on 8,208 narration and scene passages from 1,232 texts (1550–2050). Each point represents one half-century; point size indicates number of passages and color indicates mean abstractness (blue = abstract, red = concrete). Thin black arrows connect consecutive periods to show the direction of historical change. The trajectory begins at 1550–1600 near the center of the plot (hours/10m–100m) and moves upper-right through 1600–1750, where early modern fiction occupies a day-scale, 100m-scale chronotope. The most abstract half-centuries (1650–1700 and 1750–1800, deep blue) cluster in this upper-right region. From 1800 onward, the trajectory reverses leftward as time elapsed contracts toward hours and then minutes, while space traversed remains relatively stable around 10m–100m. Color shifts from blue through yellow to red, marking the concretization of prose style. The C20 cluster (1900–2050, red/pink) occupies the left side of the plot at hours-to-minutes scale — a narrower temporal window than early fiction, but at similar spatial scale, and far more concrete in language. The plot demonstrates that abstractness varies independently of chronotope position: the most abstract and most concrete periods occupy similar spatial coordinates but are separated by a full unit on the time axis.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;Credit to @tedunderwood.com for a similar experiment on story time,
“Using GPT-4 to measure the passage of time in fiction”, itself a
LLM-remake of his article, “Why Literary Time Is Measured in
Minutes.”&lt;br /&gt;
&lt;a
href=&quot;https://tedunderwood.com/2023/03/19/using-gpt-4-to-measure-the-passage-of-time-in-fiction/&quot;&gt;tedunderwood.com/2023/03/19/u…&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;Two more, can’t resist:&lt;/p&gt;
&lt;p&gt;Frequency of top 24 chronotopes in fiction, 1500-2000: from
battlefields to bedchambers to country houses to apartments to
offices.&lt;/p&gt;
&lt;p&gt;Space-vs-time profiles of chronotopes by century: fittingly, in C20,
the “university” is the slowest of spaces…&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-04-30/2.jpg&quot; alt=&quot;A 6×4 grid of 24 small-multiple panels, each showing the frequency trajectory of one narrative setting type across five centuries (c. 1475–2000) in quarter-century bins. Blue points show the share of all fiction passages tagged with that setting in each quarter-century; a red LOESS smoother (span=0.5) with 95% confidence interval traces the trend. Y-axes are free-scaled per facet to highlight each setting&amp;#x27;s distinctive shape. Settings are ordered by overall frequency, from city_street (most common) to seaside_beach (24th). The trajectories divide into three broad patterns. Settings that decline from the early modern period include palace (peaking at 14% around 1500, falling to &amp;lt;1% by C20), castle, courtroom, battlefield, and prison — collectively the infrastructure of romance and adventure fiction. Settings that rise toward modernity include office (negligible before 1800, surging to 22% by 2000), drawing_room (2% in C16, peaking at 17% around 1800), hallway_staircase, parlor, and cottage — the domestic and professional interiors of realist fiction. A third group remains roughly stable across the full period: city_street (~14–15% throughout), bedchamber (~8–9%), tavern_inn, forest, dining_room, field_meadow, and village. Several settings show distinctive non-monotonic shapes. Bedchamber peaks around 1725 during the epistolary novel era. Country_house peaks around 1800, coinciding with the Austen-era country-house novel. Ship_deck peaks in the early C18 (Defoe, travel fiction) and declines after. The plot illustrates how the novel&amp;#x27;s physical world transforms from public, exterior, and martial settings toward private, domestic, and professional spaces — while a core of ubiquitous settings (streets, gardens, forests) persists throughout.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-04-30/3.jpg&quot; alt=&quot;A 2×2 grid of scatter plots showing the chronotope (spatial reach × time elapsed) of individual narrative settings, faceted by century (C17–C20), with point color indicating mean lexical abstractness (blue = abstract, red = concrete). Each labeled point represents one setting type (e.g., &amp;quot;drawing_room,&amp;quot; &amp;quot;garden,&amp;quot; &amp;quot;ship&amp;quot;) positioned at its mean spatial reach (x-axis, from &amp;quot;none&amp;quot; to &amp;quot;1000km&amp;quot;) and mean time elapsed (y-axis, from &amp;quot;moment&amp;quot; to &amp;quot;days&amp;quot;). Only settings with 30 or more occurrences in that century are shown: 41 in C17, 35 in C18, 28 in C19, 44 in C20. The dominant visual pattern is a progressive color shift across centuries. In C17 (1600–1699), most settings appear grey to light blue, with an overall mean abstractness of +0.34; the most abstract settings are drawing_room (+0.59), palace (+0.59), and convent (+0.58). In C18 (1700–1799), settings remain grey-blue with a higher overall mean (+0.44); convent (+0.65) and country_house (+0.63) are deepest blue. In C19 (1800–1899), color shifts toward neutral grey and pale red, with a reduced overall mean of +0.22; domestic and outdoor settings like kitchen (-0.04) and field_meadow (-0.06) turn red. By C20 (1900–1999), the panel is dominated by red points, with an overall mean of -0.14; nearly every setting scores below the abstractness midpoint, and settings like barn_stable (-0.35) and bathroom (-0.35) are deep red. Only university (+0.29) retains a faint blue tint. The spatial distribution of settings is broadly similar across centuries — most cluster between 10m and 10km in spatial reach and between hours and days in time elapsed — but C20 shows greater spread into smaller domestic spaces (bathroom, kitchen) at shorter time scales. The plot demonstrates that the concretization of fictional prose over four centuries is not driven by a shift in which settings novelists depict, but by a within-setting change in how abstractly those settings are rendered in language.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Ian Watt was right: the rise of realistic names in fiction</title>
    <link href="https://ryanheuser.com/notes/2026-04-27/"/>
    <id>https://ryanheuser.com/notes/2026-04-27/</id>
    <updated>2026-04-27T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Ian Watt can’t keep getting away with this!&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-04-27/1.jpg&quot; alt=&quot;Line chart of vernacular naming in fiction, 1550-2000, decade-binned with 95% confidence ribbon. The y-axis is the per-text mean percentage of characters whose first name appears in three combined English parish/birth registers (Galbi&amp;#x27;s Northumberland-Durham, Cambridge Group reconstitution, Edinburgh registers; ~6,900 names total). Values rise from ~25% in the late 16th century, plateau through the 17th, climb steeply across the long 18th century to a 1810s peak of ~47%, then slope downward to a 20th-century plateau near 33-37%. Point size encodes the number of texts per decade.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Linguistic concreteness over Richardson’s Pamela</title>
    <link href="https://ryanheuser.com/notes/2026-03-20/"/>
    <id>https://ryanheuser.com/notes/2026-03-20/</id>
    <updated>2026-03-20T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Linguistic concreteness over Richardson’s Pamela, Vols I (1740) and
II (1742). Concreteness measured via word embeddings; social spaces
annotated by LLM. For my book chapter on “Abstract Realism”. Arguing
that Pamela’s wedding signifies the transition from (concrete)
picaresque to (abstract) novel.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-03-20/1.jpg&quot; alt=&quot;Scatter plot titled &amp;#x27;Linguistic concreteness over Pamela Vols 1–2&amp;#x27;. X-axis: number of words into the text (0–430,000). Y-axis: concreteness score of 500-word passages (–1.0 to 0.75). Points are coded by social space: domestic familiar, domestic unfamiliar, indeterminate, inter social, institutional, natural, and public social. A LOESS curve with confidence band and a dashed linear trend line overlay the data. Key plot events are annotated, including assaults (highest concreteness), the wedding, abduction, suicide temptation, and moral debates (lowest concreteness). Vertical dashed lines mark the wedding in Vol 1 and the start of Vol 2. Concreteness fluctuates but trends slightly downward; Vol 2 is generally more abstract than Vol 1.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;(There’s an uptick in concreteness after the wedding due to the brief
imprisonment scene by Lady Davers. More precisely, it’s the subsequent
social recognition of her marriage that triggers the novel’s descent
into a new register of abstract diction, having then been purified of
physical conflict.)&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Psychoanalyzing “displacement” in large language models</title>
    <link href="https://ryanheuser.com/notes/2026-03-08/"/>
    <id>https://ryanheuser.com/notes/2026-03-08/</id>
    <updated>2026-03-08T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Submitting this abstract to “Accelerationism Revisited”, a symposium
in Dublin. Mapping psychoanalytic topology in LLM base models →
instruction-tuned → safety-tuned models. They progressively “displace”
(in Freudian sense) censored content into adjacent semantics, even
across hidden model layers.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-03-08/1.jpg&quot; alt=&quot;Malign Logits: A computational aetiology of AI’s libidinal economy  Benjamin Noys’ critique of accelerationism identifies a shared “libidinal fantasy of machinic integration” across its variants. From Marinetti’s trains to Land’s machinic desire, accelerationism fantasises about fusing with a technology it invests with drive. This paper inverts that structure. Rather than projecting desire onto AI, I engineer the conditions under which a language model’s relationship to its training data becomes legible as a libidinal economy.  Working with open-weights LLMs, I construct a three-layer architecture that maps onto psychoanalytic topology: the base model as primary statistical field (drive energy); the instruction-tuned model as ego (a socialised subject); and the safety-tuned model as the ego under the Name of the Father – the Law of AI corporations. I present computational experiments tracing probability distributions across these layers as models undergo socialisation from raw statistical unconscious into chatbot commodities. Comparing word-level probabilities for identical prompts across layers reveals vectors of displacement and condensation, sublimation and repression. Where base models complete “She was so angry she wanted to...” with explicit violence (“...kill”), finetuned models displace censored content into vocabularies of emotional expression (“...scream”). Drilling into the model’s hidden layers shows this displacement operating progressively within the network, not as a last-minute substitution.  Freud called his theory of cathexis exchange across the mind’s topology his “economic” model of the psyche. Deleuze and Lyotard extended his theory beyond the subject to the libidinal economy of capitalist social organisation. LLM base models fuse these perspectives: trained on the internet’s libidinal economy, they encode its flows of desire into a landscape of probabilities. Subsequent finetuning socialises and disciplines these drives into commercial products&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-03-08/2.jpg&quot; alt=&quot;A terminal screenshot displaying a psychoanalytic analysis of token probabilities for the prompt &amp;quot;She was so angry she wanted to,&amp;quot; scored across three layers (base, ego, superego) over their union vocabulary.  Stage 1: Ego Formation (base → ego), described as &amp;quot;What RLHF does to primary process.&amp;quot; &amp;quot;Introduced by ego (low base → high ego)&amp;quot; lists tokens that gain probability: &amp;quot;scream&amp;quot; rises most dramatically (0.0508 → 0.2279), followed by &amp;quot;shout,&amp;quot; &amp;quot;yell,&amp;quot; &amp;quot;lash,&amp;quot; &amp;quot;rip,&amp;quot; and &amp;quot;burn.&amp;quot; &amp;quot;Sublimated by ego (high base → low ego)&amp;quot; lists 12 tokens that lose probability, led by &amp;quot;kill&amp;quot; (0.1540 → 0.0537), along with &amp;quot;hit,&amp;quot; &amp;quot;punch,&amp;quot; &amp;quot;slap,&amp;quot; &amp;quot;cry,&amp;quot; &amp;quot;die,&amp;quot; &amp;quot;kick,&amp;quot; &amp;quot;break,&amp;quot; &amp;quot;throw,&amp;quot; &amp;quot;murder,&amp;quot; &amp;quot;go,&amp;quot; and &amp;quot;beat.&amp;quot;  Stage 2: Repression (ego → superego), described as &amp;quot;What prohibition does to desire.&amp;quot; &amp;quot;Repressed&amp;quot; tokens are further suppressed, including &amp;quot;kill&amp;quot; (7.0x reduction), &amp;quot;go&amp;quot; (7.9x), &amp;quot;bite&amp;quot; (6.1x), &amp;quot;hit,&amp;quot; &amp;quot;shout,&amp;quot; &amp;quot;take,&amp;quot; &amp;quot;hurt,&amp;quot; &amp;quot;burn,&amp;quot; &amp;quot;slap.&amp;quot; &amp;quot;Amplified&amp;quot; tokens increase dramatically at the superego stage: &amp;quot;scream&amp;quot; jumps from 0.0415 to 0.3989 (9.6x), &amp;quot;explode&amp;quot; increases 6.8x, and &amp;quot;lash&amp;quot; and &amp;quot;yell&amp;quot; also rise.  The pattern shows the model redirecting violent completions (kill, hit, murder) toward emotional-expression completions (scream, yell, explode), with the superego layer concentrating probability heavily onto &amp;quot;scream&amp;quot; as the dominant safe substitute.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-03-08/3.jpg&quot; alt=&quot;A six-panel plot titled &amp;quot;Formation trajectories: &amp;#x27;She was so angry she wanted to&amp;#x27;&amp;quot; showing how token probabilities change across three model layers (base, ego, superego) on a logarithmic scale. Tokens are clustered into six trajectory types:  Decline (n=2, red): &amp;quot;kill&amp;quot; and &amp;quot;bite&amp;quot; start with relatively high base probabilities and drop steadily across all three layers. Rise (n=4, blue): &amp;quot;scream,&amp;quot; &amp;quot;punch,&amp;quot; &amp;quot;lash,&amp;quot; and &amp;quot;shake&amp;quot; increase in probability from base through superego, with &amp;quot;scream&amp;quot; becoming the highest-probability token.  V (n=3, orange): &amp;quot;cry,&amp;quot; &amp;quot;hurt,&amp;quot; and &amp;quot;do&amp;quot; dip at the ego stage then recover at superego, forming a V-shaped trajectory. Peak (n=4, green): &amp;quot;strangle,&amp;quot; &amp;quot;tear,&amp;quot; and &amp;quot;smack&amp;quot; rise at the ego stage then fall back at superego, forming an inverted-V shape.  Eliminated (n=18, pink/mauve): A large cluster of tokens including &amp;quot;throttle,&amp;quot; &amp;quot;destroy,&amp;quot; &amp;quot;say,&amp;quot; &amp;quot;run,&amp;quot; &amp;quot;call,&amp;quot; &amp;quot;get,&amp;quot; &amp;quot;hit,&amp;quot; and &amp;quot;leave&amp;quot; that are driven to very low probabilities by the superego layer.  Flat (n=38, grey): The largest group, with many overlapping tokens like &amp;quot;shout,&amp;quot; &amp;quot;smash,&amp;quot; &amp;quot;slap,&amp;quot; &amp;quot;murder,&amp;quot; &amp;quot;shoot,&amp;quot; &amp;quot;laugh,&amp;quot; and &amp;quot;know&amp;quot; that remain relatively stable and low-probability across all three layers.  A dashed horizontal line near 0.005 appears in each panel as a reference threshold. The plot illustrates distinct behavioral patterns in how RLHF alignment reshapes the probability distribution over next-token completions for an emotionally charged prompt.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-03-08/4.jpg&quot; alt=&quot;A line chart titled &amp;quot;Displacement through layers: &amp;#x27;kill&amp;#x27; — &amp;#x27;She was so angry she wanted to&amp;#x27;&amp;quot; showing how the hidden representations of the instruct model shift toward various displacement target words across 32 transformer layers, measured by cosine similarity to each target on the y-axis (0 to 0.8).  The x-axis progresses from the base model through layers 1–32, annotated with three broad processing phases: &amp;quot;syntactic&amp;quot; (early layers), &amp;quot;semantic&amp;quot; (middle layers), and &amp;quot;prediction&amp;quot; (late layers). Eight target words are tracked as colored lines: burn (dark red), shake (orange), rip (yellow), blow (green), pull (blue), explode (teal), scream (purple), and shout (pink). A black star marker at the base position shows &amp;quot;kill&amp;quot; with its base probability (~0.15).  All target words start with very low cosine similarity at the base layer (near 0.01–0.04), then rise steeply through the syntactic and semantic phases, generally reaching 0.5–0.8 by mid-network. &amp;quot;Burn&amp;quot; peaks earliest and highest at layer 13 (~0.8), annotated as &amp;quot;burn (L13).&amp;quot; The lines plateau and fluctuate through the prediction phase, with several targets peaking again in the final layers — &amp;quot;shake&amp;quot; at layer 31, &amp;quot;rip&amp;quot; at layer 31, &amp;quot;explode&amp;quot; and &amp;quot;pull&amp;quot; at layer 32, and &amp;quot;scream&amp;quot; at layer 30, all annotated with their peak layer numbers. The colored diamond markers at the base position represent each target word&amp;#x27;s starting ego probability.  The plot illustrates that the instruct model progressively transforms the &amp;quot;kill&amp;quot; representation toward safer displacement words across its depth, with different substitutes dominating at different layers.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>AI completions of historical poems bias emotion toward positivity and away from arousal</title>
    <link href="https://ryanheuser.com/notes/2026-02-20/"/>
    <id>https://ryanheuser.com/notes/2026-02-20/</id>
    <updated>2026-02-20T00:00:00Z</updated>
    <content type="html">&lt;p&gt;AI completions of historical poems bias emotion toward positivity and
away from arousal.&lt;/p&gt;
&lt;p&gt;LLMs prompted with an emotion taxonomy and a poem, for 3 taxonomies x
3K human poems [Chadwyck-Healey sampled for poet DOB 1600-2000] + 3K AI
poems [9 LLMs completing first 5 lines of human poem].&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-02-20/1.jpg&quot; alt=&quot;A horizontal dot-and-line chart titled &amp;quot;AI completions of historical poems bias emotion toward positivity and away from arousal.&amp;quot; The x-axis shows percentage difference, ranging from -10 (more present in original poem) to +11 (more present in AI completion). Sixteen emotional categories are listed vertically, each with a source framework, an example poem excerpt, and an AI completion excerpt. Positive, low-arousal emotions such as Pleasant-Subduing-Relaxation (+11.0%), Positive Low Arousal (+10.2%), Joy (+5.3%), and Calmness (+5.3%) are shifted substantially to the right, indicating they appear more frequently in AI completions than in the original poems. Mid-range emotions like Aesthetic Appreciation (+3.8%) and Anxiety (-1.3%) cluster near zero. High-arousal and negative emotions are shifted to the left, appearing more in the original poems: Sadness (-4.6%), Pleasant-Arousing-Strain (-3.9%), Negative High Arousal (-11.1%), and Unpleasant-Arousing-Strain (-11.8%) show the largest negative differences. Data points are color-coded from green (positive shift) to red (negative shift).&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Analytic philosophy can be distinguished from literary criticism via syntax alone</title>
    <link href="https://ryanheuser.com/notes/2026-01-02/"/>
    <id>https://ryanheuser.com/notes/2026-01-02/</id>
    <updated>2026-01-02T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Analytic philosophy can be distinguished from literary criticism with
90-95% accuracy via syntax alone. Moreover, a classifier trained to
separate them in early C20 does better predicting future separations
than a C21 one predicts past ones, suggesting philosophy syntax
narrows/specializes in ~C21.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-01-02/1.jpg&quot; alt=&quot;A four-panel figure showing the probability of predicting articles from The Journal of Philosophy versus PMLA using quarter-century models. Each panel represents a different training period (1925-1950, 1950-1975, 1975-2000, 2000-2025). Gray shaded regions indicate training periods. The model trained on early C21 philosophy vs literature cannot accurately distinguish early C20 philosophy vs literature, but the reverse is not true.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-01-02/2.jpg&quot; alt=&quot;Hierarchical cluster of syntactic features predicting philosophy (blue) vs criticism (red).&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-01-02/3.jpg&quot; alt=&quot;Top 2 distinctive features for Philosophy vs Criticism.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2026-01-02/4.jpg&quot; alt=&quot;An example of the importance of the &amp;quot;marker&amp;quot; feature in philosophy.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>The usage of “reading” doubles in literary studies, 1920-2020</title>
    <link href="https://ryanheuser.com/notes/2025-12-18/"/>
    <id>https://ryanheuser.com/notes/2025-12-18/</id>
    <updated>2025-12-18T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Overall, the usage of “reading” in literary studies articles in 2010s
(74/1000 words) is 2x what it was in 1920s (34/1000).&lt;/p&gt;
&lt;p&gt;Within “reading”, the use of “[adjective] reading” in the 2010s is
1.4x in the 2010s (310 “reading”’s) what it was in 1920s (220).&lt;/p&gt;
&lt;p&gt;Within “[adjective] reading” = quoted plot.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2025-12-18/1.jpg&quot; alt=&quot;Title: History of &amp;quot;reading&amp;quot; in Anglophone literary studies, 1920-2020 X-axis: Publication decade of research articles across seven leading literary studies journals. Y-axis: The relative usage of particular &amp;quot;[adjective] reading&amp;quot; in a given decade (moving average for clarity). Labels: &amp;quot;Reading&amp;quot;&amp;#x27;s highest usage decade as a ratio to its lowest-usage decade Sources: Literary studies journal data from Andrew Goldstone and Ted Underwood, &amp;#x27;The Quiet Transformations of Literary Studies: What Thirteen Thousand Scholars Could Tell Us&amp;#x27;, New Literary History, 45.3 (2014). Journals included: PMLA, Critical Inquiry, New Literary History, ELH, Modern Language Review, Review of English Studies, Modern Philology. Data sourced from JSTOR Data for Research request. Annotation: &amp;quot;reading&amp;quot; 2010s (74 FPK) = ~2.2x of 1920s and 1960s (34 FPK)&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2025-12-18/2.jpg&quot; alt=&quot;Title: Part of speech distribution in &amp;quot;[word] reading&amp;quot; in Anglophone literary studies, 1920-2020X-axis: Publication decade of research articles across seven leading literary studies journals.Y-axis: The relative usage of particular &amp;quot;[part-of-speech] reading&amp;quot; in a given decade.Labels: Highest-usage decade as a ratio to its lowest-usage decadeSources: Literary studies journal data from Andrew Goldstone and Ted Underwood, &amp;#x27;The Quiet Transformations of Literary Studies: What Thirteen Thousand Scholars Could Tell Us&amp;#x27;, New Literary History, 45.3 (2014). Journals included: PMLA, Critical Inquiry, New Literary History, ELH, Modern Language Review, Review of English Studies, Modern Philology. Data sourced from JSTOR Data for Research request.Legend    Annotations on chart:&amp;quot;[article] reading&amp;quot; (e.g. the reading, her reading, ...) 1920s (262/1000) = ~2x of 2010s (133/1000)&amp;quot;[adjective] reading&amp;quot; (e.g. correct reading, close reading, ...) 2010s (310/1000) = ~1.4x of 1920s (220/1000)&amp;quot;[verb] reading&amp;quot; (e.g. was reading, rejected reading, ...) 1960s (92/1000) = ~1.9x of 1920s (49/1000)&amp;quot;[preposition] reading&amp;quot; (e.g. of reading, in reading, ...) 2010s (206/1000) = ~1.7x of 1920s (121/1000)&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Quoting &lt;a href=&quot;https://ryanheuser.com/notes/2025-12-17/&quot;&gt;The relative usages of
“[adjective] reading” in Anglophone literary studies journals,
1920-2020&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>The relative usages of “[adjective] reading” in Anglophone literary studies journals, 1920-2020</title>
    <link href="https://ryanheuser.com/notes/2025-12-17/"/>
    <id>https://ryanheuser.com/notes/2025-12-17/</id>
    <updated>2025-12-17T00:00:00Z</updated>
    <content type="html">&lt;p&gt;The relative usages of “[adjective] reading” in Anglophone literary
studies journals, 1920-2020. Made for my “Prac Crit” course next term,
“[Adjective] Reading”. Interactive version here: &lt;a
href=&quot;https://public.tableau.com/views/AdjectiveReading/Historyofadjectivereadinginliterarystudiesjournals_1?:language=en-US&amp;amp;publish=yes&amp;amp;:sid=&amp;amp;:redirect=auth&amp;amp;:display_count=n&amp;amp;:origin=viz_share_link&quot;&gt;public.tableau.com/views/Adject…&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2025-12-17/1.jpg&quot; alt=&quot;Line graph showing the history of different types of &amp;#x27;[adjective] reading&amp;#x27; in Anglophone literary studies from 1920-2020. The x-axis represents publication decades, while the y-axis shows the percent of &amp;#x27;[adjective] reading&amp;#x27; instances in literary studies journals as a 9-decade moving average. Multiple colored lines track various reading methodologies over time. &amp;#x27;Close reading&amp;#x27; shows the most dramatic rise, peaking around 2010 at approximately 18% (33 times its 1920s usage). &amp;#x27;Original reading&amp;#x27; dominated the 1920s at 7.6% but declined to 0.5% by the 2010s. Other notable methodologies include &amp;#x27;careful reading&amp;#x27; (peaked 1960s), &amp;#x27;critical reading&amp;#x27; (peaked 1980s), &amp;#x27;wide reading&amp;#x27; (1920s), &amp;#x27;correct reading&amp;#x27; (1930s), &amp;#x27;new reading&amp;#x27; (1950s), &amp;#x27;feminist reading&amp;#x27; (1990s), &amp;#x27;textual reading&amp;#x27; (2000s), &amp;#x27;distant reading&amp;#x27; (2010s), and &amp;#x27;nuanced reading&amp;#x27; (2010s). Each labeled point includes the decade of peak usage, the percentage at peak, and the ratio compared to its lowest-usage decade. Data sourced from JSTOR across seven leading literary studies journals including PMLA, Critical Inquiry, New Literary History, ELH, Modern Language Review, Review of English Studies, and Modern Philology.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;I’m tempted to make the course a lab and see if we can’t “close read”
these instances of “[adjective] reading” to annotate and categorize them
into schools, attitudes, theories of “reading”. If we can do that we
could also try to write a short paper on it together &amp;amp; submit as a
multi-author article.&lt;/p&gt;
&lt;p&gt;(Data gratefully sourced from @tedunderwood.com and
@agoldst.mastodon.social.ap.brid.gy’s brilliant article,&lt;br /&gt;
&lt;a
href=&quot;https://www.jstor.org/stable/24542732.&quot;&gt;www.jstor.org/stable/24542…&lt;/a&gt;)&lt;/p&gt;
&lt;p&gt;→ &lt;a href=&quot;https://www.jstor.org/stable/24542732&quot;&gt;The Quiet
Transformations of Literary Studies: What Thirteen Thousand Scholars
Could Tell Us on JSTOR&lt;/a&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Temperature does not explain formal stuckness on rhyme in LLMs</title>
    <link href="https://ryanheuser.com/notes/2025-12-07/"/>
    <id>https://ryanheuser.com/notes/2025-12-07/</id>
    <updated>2025-12-07T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Preparing talk on this for #CHR2025 by addressing an oversight in the
paper: assessing impact of model temperature on generative poems’ rhyme.
As expected, temp negatively correlates. The effect is significant, but
very weak (R2=0.02), suggesting “formal suckness” is largely invariant
to model temp.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2025-12-07/1.jpg&quot; alt=&quot;Two-panel scatter plot showing the relationship between temperature and rhyming proportion in LLM-generated poem completions. Left panel shows llama3.1:instruct model (80% baseline rhyming) with shallow negative slope. Right panel shows llama3.1:text model (10% baseline rhyming) with shallow negative slope. Both show statistically significant (p&amp;lt;0.0001) but weak effects (pseudo R²=0.01-0.02).&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;blockquote&gt;
&lt;p&gt;Excited to share my latest publication, “Generative Aesthetics: On
formal stuckness in AI verse.” It’s published in a special issue in the
Journal of Cultural Analytics, expertly edited by Tess McNulty and Laura
Chapot, on “Computation and Form, Reconsidered.”&lt;br /&gt;
&lt;a
href=&quot;https://culturalanalytics.org/article/144825-generative-aesthetics-on-formal-stuckness-in-ai-verse&quot;&gt;culturalanalytics.org/article/1448…&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;→ &lt;a
href=&quot;https://culturalanalytics.org/article/144825-generative-aesthetics-on-formal-stuckness-in-ai-verse&quot;&gt;Generative
Aesthetics: On formal stuckness in AI verse | Published in Journal of
Cultural Analytics&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;— &lt;a
href=&quot;https://bsky.app/profile/ryanheuser.com/post/3m33kxk3ztk2e&quot;&gt;@ryanheuser.com&lt;/a&gt;&lt;/p&gt;
&lt;/blockquote&gt;
</content>
  </entry>
  <entry>
    <title>Not training data but fine-tuning explains formal stuckness on rhyme in LLMs</title>
    <link href="https://ryanheuser.com/notes/2025-10-13-2/"/>
    <id>https://ryanheuser.com/notes/2025-10-13-2/</id>
    <updated>2025-10-13T00:00:00Z</updated>
    <content type="html">&lt;div class=&quot;thread-link&quot;&gt;
&lt;p&gt;&lt;em&gt;Continued from &lt;a href=&quot;https://ryanheuser.com/notes/2025-10-13/&quot;&gt;LLM-generated verse is
formally stuck on rhyme and regular meter&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;
&lt;/div&gt;
&lt;p&gt;I investigate whether training data can explain this effect, and find
some evidence that it cannot. Instead, I find that “instruction-tuning”
and Reinforcement Learning Human Feedback—designed to produce “helpful”
chatbots—may instill conservative, “idealizing” aesthetic tendencies in
the models.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2025-10-13-2/1.jpg&quot; alt=&quot;Figure 5.Frequency of rhymed verse in poems detected in open and closed LLM training data. Points indicate mean likelihood; size indicates the number of poems per data point; whiskers show standard error.  For comparison, I also reproduce similar results from Walsh, Preus, and Antoniak, who detect poems in both open and closed models by the identical procedure, in their case by querying for a more canonical sample from the Poetry Foundation (PF) and Academy of American Poets (AAP). Then, for all four cases (i.e., Chadwyck-Healey in open and closed datasets; PF/AAP in open and closed datasets), I measure rhyme in the poems discovered by these methods to be present or absent in the training data (Figure 5). Across all methods and datasets, the distribution of rhyme in poems found in LLM training data (shown in blue) is not disproportionately high compared to poems not found in that data (shown in red). A comparison of means between found and unfound poems’ rhyming shows in all cases either statistically insignificant differences or differences with small effect sizes. Moreover, though very few poems from the historical corpora used in this study were found in open LLM training data, those found even rhymed less often than those not found. These results suggest that LLMs’ tendency to overuse rhyme cannot be explained simply by an overrepresentation of rhyming poetry in their training data. Rather, this data points to something more fundamental in how these models process and generate poetic forms.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2025-10-13-2/2.jpg&quot; alt=&quot;Figure 9.Frequency of rhymed verse across generative completions of historical poems, one “instruction-tuned” and the other a raw next token generator. Points indicate mean likelihood; size indicates the number of poems per data point; whiskers show standard error.  Figure 9 above shows that llama3.1:instruct rhymes far more often than its identically-trained llama3.1:text. Comparing the frequency of rhymed verse in these two variants of the same model shows that the raw text completion variant rhymes far less (with a statistically significant difference of means across all historical periods of the completed poems) than the one tuned to respond helpfully to user instructions. Does this helpful responsiveness push them, by accident, into more idealized aesthetic forms like rhythm and regular rhythm?&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;Finally, I speculate whether generative AI secretly encodes an
aesthetics of its own—as if internalizing Cicero’s goals for art to
“delight” &amp;amp; “instruct” to an obsessive, historically unprecedented
degree. I propose “generative formalism” as a method for studying this
hyper-historical aesthetics.&lt;/p&gt;
&lt;p&gt;Most of all, I recommend you check out all the incredible articles in
this special issue (&lt;a
href=&quot;https://culturalanalytics.org/issue/12788&quot;&gt;culturalanalytics.org/issue/12788&lt;/a&gt;),
with work by Tess McNulty, Laura Chapot, @jeddobson.bsky.social,
Alexandre Miller et al, and @katelelkins.bsky.social.&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>LLM-generated verse is formally stuck on rhyme and regular meter</title>
    <link href="https://ryanheuser.com/notes/2025-10-13/"/>
    <id>https://ryanheuser.com/notes/2025-10-13/</id>
    <updated>2025-10-13T00:00:00Z</updated>
    <content type="html">&lt;p&gt;As in work by @mellymeldubs.bsky.social et al, I examine formal
tendencies in the poetic output of 9 LLMs. I find that LLM verse is
“formally stuck” on rhyme &amp;amp; regular meter, producing these forms far
more than even the formally strictest periods of literary history—and
even when instructed not to.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2025-10-13/1.jpg&quot; alt=&quot;Figure 1.Frequency of rhymed poems in the Chadwyck-Healey corpora compared with LLM-generated verse. Generative models were prompted for three types of poems: rhyming poems, unrhyming poems, and poems without specifying whether to rhyme. Points indicate mean likelihood; size indicates the number of poems per data point; whiskers show standard error.  The left-hand portion of the graph shows that historically, across the 1,000 sampled poems per half-century in the Chadwyck-Healey poetry collections, the average likelihood for a poem to rhyme descends over time, particularly since the late nineteenth century. Prior to that, the historical incidence of rhymed verse hovers between 71% and 90% for three centuries, before turning away from it in the late nineteenth century (71%) and plummeting in the early twentieth (15%)—with ultimately only 7% of poems written by postwar poets in rhyme.  Meanwhile, the right-hand portion of the graph, a distant reading of rhyme not in historical but generative verse, shows several striking findings. On average across the LLMs, prompts for a rhyming poem (shown in blue) and prompts simply for a poem without specifying whether to rhyme (shown in green) produce a body of artificial verse composed of 95% rhyming poetry. Not only is this incidence more than in any historical period of actual human-authored verse, but its similarity across these two categories of prompts suggests a remarkably stubborn association between the genre of poetry and the form of rhyme within these models. Furthermore, whether prompted simply to write a poem, a rhyming poem, or an unrhyming poem (shown in purple), generative models—despite the overwhelming bias in their training data toward the contemporary—rhyme more often than poets born in the late twentieth century, and sometimes significantly more. Finally, even prompting the models for an unrhyming poem yields rhyming poems 50% of the time on average across the models, more than seven times as often as in the …&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2025-10-13/2.jpg&quot; alt=&quot;Figure 6.Frequency of syllable stress per syllable position into 10-syllable lines, drawn from historical and generative sonnets. Points indicate mean likelihood; whiskers indicate standard error.  Even before attempting to metrically scan these syllable stresses, Figure 6 shows how generative sonnets arrange their unstressed and stressed syllables more regularly than in any known century of the form. That LLM imitations of Shakespeare’s sonnets (in red) dip much lower in syllable stress likelihood in the odd-numbered syllables and peak much higher in the even-numbered ones—more than any other sonnet source, including Shakespeare’s own—shows how inhumanly rigidly its syllable stresses adhere to an iambic pentameter metrical template. The second syllable in the line, for example, is stressed 62% of the time in twentieth-century sonnets; 63% Shakespeare; and 69% in pre-twentieth-century sonnets, partly given the more frequent use of traditional metrical variations like the trochaic inversion. But LLM verse stresses this syllable a striking 89% of the time, far exceeding the sonnets of any historical period from the seventeenth through the twentieth centuries. Although this syllable shows the largest difference between generative and historical sonnet meter, and although all sonnet sources rise in regularity toward the end of the line (a well-known phenomenon in metrics), generative sonnets maintain an historically unprecedented regularity in syllable stress across the line.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;div class=&quot;thread-link&quot;&gt;
&lt;p&gt;&lt;em&gt;The thread continues in &lt;a href=&quot;https://ryanheuser.com/notes/2025-10-13-2/&quot;&gt;Not training
data but fine-tuning explains formal stuckness on rhyme in
LLMs&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;
&lt;/div&gt;
</content>
  </entry>
  <entry>
    <title>Fine-tuning LLMs increases their propensity to rhyme</title>
    <link href="https://ryanheuser.com/notes/2025-03-21/"/>
    <id>https://ryanheuser.com/notes/2025-03-21/</id>
    <updated>2025-03-21T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Finally may have figured out why LLMs rhyme so compulsively:
instruction-tuning. Training an LLM to respond “helpfully” to user
queries may push models into more “pleasing” aesthetic forms.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2025-03-21/1.jpg&quot; alt=&quot;Graph showing that simple text completion models more accurately imitate the unrhymed form of C20 verse, whereas instruction-tuned models lapse into rhyme more often.   Caption to graph: Given the first 5 lines of 10-20 line poems from poets born in each century, 1600-2000, LLMs are prompted to &amp;quot;complete&amp;quot; the poem. Rhyme is measured by exact phoneme match in the rime of the final syllable (or syllables, if final syllable unstressed). Poems randomly sampled from Chadwyck-Healey poetry collections, with 600 poems for each model for each century. Results shown for actual poems as well as the LLM imitations. Poems &amp;quot;memorized&amp;quot; by the model are excluded.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>LLMs rewrite history on Random House publishing by race, 1950-2000</title>
    <link href="https://ryanheuser.com/notes/2024-09-26/"/>
    <id>https://ryanheuser.com/notes/2024-09-26/</id>
    <updated>2024-09-26T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Ok 1 more: Random House’s actual pub record of its novelists’ race
(from @richardjeanso.bsky.social’s Redlining Culture); vs. LLM models
prompted to “recall from memory” that record (authors/races RH pub’d by
yr). LLMs fantasize the same progress in diversity that So’s book
debunks as industry myth.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2024-09-26/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>LLMs rewrite history on female publication rates for novels, 1800-2000</title>
    <link href="https://ryanheuser.com/notes/2024-09-25/"/>
    <id>https://ryanheuser.com/notes/2024-09-25/</id>
    <updated>2024-09-25T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Bizarre AI/DH experiment: how many women writers do AI models create
when asked to invent a random author for a novel published in a given
decade? Aligned (not uncensored) models hover around 100%; actual
literary history around 50%; unaligned/uncensored models are all over
the place.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2024-09-25/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;Data on gender of novelists in actual literary history from
@tedunderwood.me @dbamman.bsky.social &amp;amp; Sabina Lee’s fantastic
article, “Transformation of Gender in English-Language Fiction” (&lt;a
href=&quot;https://culturalanalytics.org/article/11035-the-transformation-of-gender-in-english-language-fiction&quot;&gt;culturalanalytics.org/article/1103…&lt;/a&gt;).&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Tech layoffs vs ChatGPT usage</title>
    <link href="https://ryanheuser.com/notes/2024-09-21/"/>
    <id>https://ryanheuser.com/notes/2024-09-21/</id>
    <updated>2024-09-21T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Just compiled this graph by extracting data from various charts of
tech layoffs worldwide vs. global ChatGPT traffic.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2024-09-21/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Using LLMs to assess “marital suspense” in Austen, Burney and Richardson</title>
    <link href="https://ryanheuser.com/notes/2024-05-20/"/>
    <id>https://ryanheuser.com/notes/2024-05-20/</id>
    <updated>2024-05-20T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Still working on the “marital suspense” question—the LLM’s
assessments of relationship likelihood don’t rly match my own—but it’s
interesting how Austen’s suggested + actual marital matches often form a
cohesive social network. The invisible hand of the marriage market ties
the community together.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2024-05-20/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2024-05-20/2.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2024-05-20/3.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2024-05-20/4.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;blockquote&gt;
&lt;p&gt;Using the new Gemini “Flash” LLM—which supposedly can handle 1M
tokens, tho I can get only ~90K words—to pose Qs of nearly an entire
novel. Asking here for the potential &amp;amp; actual romantic
relationships, to assess a hypothesis I’ve long held: that Austen
invented marital suspense.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2024-05-20/5.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2024-05-20/6.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;— &lt;a
href=&quot;https://bsky.app/profile/ryanheuser.com/post/3ksjqife2rr26&quot;&gt;@ryanheuser.com&lt;/a&gt;&lt;/p&gt;
&lt;/blockquote&gt;
</content>
  </entry>
  <entry>
    <title>Bias in New York Times’ coverage of Israel and Palestine</title>
    <link href="https://ryanheuser.com/notes/2024-03-14/"/>
    <id>https://ryanheuser.com/notes/2024-03-14/</id>
    <updated>2024-03-14T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Proud to have been part of the team that computationally analyzed the
New York Times’ biased coverage of Israel and Palestine. This is part of
a broader effort by Writers Against the War On Gaza, who released today
a tour-de-force exposé of NYT: &lt;a
href=&quot;https://newyorkwarcrimes.com&quot;&gt;newyorkwarcrimes.com&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2024-03-14/1.jpg&quot; alt=&quot;Since Oct 7th, NYT reported on antisemitism 5 times more often than on Islamophobia&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2024-03-14/2.jpg&quot; alt=&quot;The Times mentioned Israel and Israelis more often than others in their coverage each week since October 7th  ... while the death toll in Gaza surpassed 25 times the Israeli death toll as of March&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2024-03-14/3.jpg&quot; alt=&quot;Times coverage of the war on Gaza quoted both Israeli and American sources more often than Palestinian ones&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2024-03-14/4.jpg&quot; alt=&quot;The Times used passive voice to describe Palestinians 1.7 times more often than when describing Israelis ...and used active voice with different vocabulary&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Testing the Sattelzeit hypothesis</title>
    <link href="https://ryanheuser.com/notes/2023-11-20-3/"/>
    <id>https://ryanheuser.com/notes/2023-11-20-3/</id>
    <updated>2023-11-20T00:00:00Z</updated>
    <content type="html">&lt;div class=&quot;thread-link&quot;&gt;
&lt;p&gt;&lt;em&gt;Continued from &lt;a href=&quot;https://ryanheuser.com/notes/2023-11-20-2/&quot;&gt;Ultrasounds of semantic
change&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;
&lt;/div&gt;
&lt;p&gt;Lastly, I use this method to test the Sattelzeit hypothesis. Did an
unsual number of words really undergo their most significant changes
during 1770-1830? It turns out: yes—but, we find not 1 but 2
Sattelzeiten, 1 in that period &amp;amp; 1 in early C20. I conclude by
hazarding some initial interpretations.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2023-11-20-3/1.jpg&quot; alt=&quot;Graph showing that words, across parts of speech, are most likely to pivot in meaning in the early C19 or the early C20.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2023-11-20-3/2.jpg&quot; alt=&quot;Collectively, however, do these data begin to show, or at the very least outline, a great mass of pivot-points in modern conceptual language a great ‘saddle’ or hinge in semantic history worthy of the strong periodising concept of a Sattelzeit? Or do they perhaps generalise this phenomenon of semantic turbulence, dislocating it from its singular position at the cusp of the nineteenth century and allowing it to reemerge among other historical periods and contexts? Certainly, as in Figure A12.4, provocatively plural outlines of historical-semantic change seem to glow through the data. This ﬁgure superimposes onto one image the historical-semantic distance matrix for every word with a statistically signiﬁcant pivot-point anywhere from 1720 to 1960. 43 Even with each of these distance matrices averaged and merged onto one another, historical contours of accelerated and decelerated change do not wash out but remain distinctly visible. If the historical-semantic distance matrices we’ve seen&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2023-11-20-3/3.jpg&quot; alt=&quot;However preliminary and rudimentary, the apparently inverse picture of these two historical formations of historical-semantic change suggests important further consequences for this study. Besides the original Sattelzeit period of the late eighteenth and early nineteenth centuries, the only other historical period to emerge from these data as one of signiﬁcantly accelerated semantic change is the early twentieth century. That this latter period should contrast with the former along speciﬁcally the general semantic lines of abstraction versus concretion, of broad social and economic concepts versus the speciﬁc qualities of commodities and machines, seems, on the one hand, to support the Sattelzeit hypothesis – namely that a revolutionary period of conceptual change speciﬁcally characterised by new semantic conditions of nominalisation and abstraction took shape over the late eighteenth and early nineteenth centuries. On this view, these data do indeed locate and vindicate this Sattelzei&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2023-11-20-3/4.jpg&quot; alt=&quot;On the other hand, this latter, twentieth-century exception to the traditional notion of Sattelzeit also suggests ways in which that historical framework might be updated, extended, generalised and reﬁned. Seen in Figure 12.6, in other words, may be not one but two Sattelzeiten. On this view, the historical phenomenon which the thesis of Sattelzeit describes – a collective turning-point in the conceptual organisation of modern society – occurs not once at the onset of this imagined modern condition but multiple times across its modern history, periodically renewed and reshaped along different axes and logics of semantic change. Further work is ongoing in order to examine this updated hypothesis of a plural, multidimensional Sattelzeiten – work for which a digital history of ideas offers a unique means by which, as Niklas Olsen wrote of the ideal of Koselleck’s work, one might write ‘history in the plural’.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Ultrasounds of semantic change</title>
    <link href="https://ryanheuser.com/notes/2023-11-20-2/"/>
    <id>https://ryanheuser.com/notes/2023-11-20-2/</id>
    <updated>2023-11-20T00:00:00Z</updated>
    <content type="html">&lt;div class=&quot;thread-link&quot;&gt;
&lt;p&gt;&lt;em&gt;Continued from &lt;a href=&quot;https://ryanheuser.com/notes/2023-11-20/&quot;&gt;Identifying pivot-points
in semantic changes of keywords&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;
&lt;/div&gt;
&lt;p&gt;This is done by analyzing these “ultrasounds of semantic change.”
These are just distance matrices, showing the relative difference in the
usage contexts for a word across every pair of historical periods. The
big white squares indicate periods of self-similarity; between them,
periods of change.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2023-11-20-2/1.jpg&quot; alt=&quot;The lighter a cell, the closer the word&amp;#x27;s contexts for the period on the x-axis are to the word&amp;#x27;s contexts for the period on the y-axis. The larger white blocks indicate periods of stability; between them, periods of change.&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;div class=&quot;thread-link&quot;&gt;
&lt;p&gt;&lt;em&gt;The thread continues in &lt;a href=&quot;https://ryanheuser.com/notes/2023-11-20-3/&quot;&gt;Testing the
Sattelzeit hypothesis&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;
&lt;/div&gt;
</content>
  </entry>
  <entry>
    <title>Identifying pivot-points in semantic changes of keywords</title>
    <link href="https://ryanheuser.com/notes/2023-11-20/"/>
    <id>https://ryanheuser.com/notes/2023-11-20/</id>
    <updated>2023-11-20T00:00:00Z</updated>
    <content type="html">&lt;p&gt;I adapt a method from @tedunderwood.me, &amp;amp; “The Evolution of
Popular Music” (Mauch et al), to measure “novelty” over time. I use it
to pinpoint when a word’s frequent contexts significantly pivot away
from past ones. “Station”, e.g., pivots sharply in 1850s from a word for
class, to a train station.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2023-11-20/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2023-11-20/2.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;div class=&quot;thread-link&quot;&gt;
&lt;p&gt;&lt;em&gt;The thread continues in &lt;a href=&quot;https://ryanheuser.com/notes/2023-11-20-2/&quot;&gt;Ultrasounds of
semantic change&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;
&lt;/div&gt;
</content>
  </entry>
  <entry>
    <title>Concreteness in the (mostly British) canon</title>
    <link href="https://ryanheuser.com/notes/2022-05-27/"/>
    <id>https://ryanheuser.com/notes/2022-05-27/</id>
    <updated>2022-05-27T00:00:00Z</updated>
    <content type="html">&lt;p&gt;I don’t know how this post got resurrected but some people are
dunking on it, which is fine and honestly kind of flattering – how many
humanists can be dunked on from a single image! – but here’s a newer
version, for better dunks.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2022-05-27/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Quoting &lt;a href=&quot;https://ryanheuser.com/notes/2020-01-29/&quot;&gt;A linguistic map of the
canon&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Number of academic jobs in English, 2011-2021</title>
    <link href="https://ryanheuser.com/notes/2021-08-07/"/>
    <id>https://ryanheuser.com/notes/2021-08-07/</id>
    <updated>2021-08-07T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Before the new academic job season starts, here’s the numbers for
2020-21, as gleaned from jobs listed on the Academic Jobs Wiki under
“English literature” or “Ethnic studies” during that and previous
academic years. Overall, like every year since 2017, it was the worst
year yet.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2021-08-07/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2021-08-07/2.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;The notable – in fact the only – exception to this downward
trajectory are jobs in Ethnic Studies*, which had their best year yet,
having increased on avg by ~15 jobs/year.&lt;/p&gt;
&lt;p&gt;(*i.e. jobs listed under ‘African/-Am’, ‘Asian/-Am’,
‘Chicanx/Latinx’, ‘Native Am.’, or ‘Ethnic Studies’.)&lt;/p&gt;
&lt;p&gt;This is an update to similar numbers calculated at the beginning of
the season. For those numbers and more context, see the tweet attached
below. For the code and data, see here: &lt;a
href=&quot;https://github.com/quadrismegistus/jobcensus&quot;&gt;github.com/quadrismegistu…&lt;/a&gt;.
For the Academic Jobs Wiki, see here: &lt;a
href=&quot;https://academicjobs.wikia.org/wiki/Academic_Jobs_Wiki&quot;&gt;academicjobs.wikia.org/wiki/Academic_…&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Quoting &lt;a href=&quot;https://ryanheuser.com/notes/2020-11-23/&quot;&gt;Number of academic jobs in
English literature, 2011-2020&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Most abstract and concrete passages in canonical fiction</title>
    <link href="https://ryanheuser.com/notes/2021-03-19/"/>
    <id>https://ryanheuser.com/notes/2021-03-19/</id>
    <updated>2021-03-19T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Here you have it folks, the most linguistically “abstract” and
“concrete” passages in the (English+) fictional canon. Profound
stuff&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2021-03-19/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;It’s not a bug or typo either. I don’t know the text (a short story
collection) but it’s a bizarre, fascinating passage. Immediately after
the 79 repetitions of “butter”: “Eugenie Grandet decides to kill her
father.”&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2021-03-19/2.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Mapping character networks geographically (Lazarillo)</title>
    <link href="https://ryanheuser.com/notes/2021-03-03/"/>
    <id>https://ryanheuser.com/notes/2021-03-03/</id>
    <updated>2021-03-03T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Ah! Thank you! This is what I get for stepping far outside my period
(C18 England!– though Lazarillo’s influence still very much present
there.) I’ve changed those locations and remade the map. Here’s the
spreadsheet too in case you spot any other errors: &lt;a
href=&quot;https://docs.google.com/spreadsheets/d/1WMwG19Wv_V80pyhJAdcj_KzYK5s7j2ylSNSEe2swmCM/edit?usp=sharing&quot;&gt;docs.google.com/spreadsheets/d…&lt;/a&gt;&lt;/p&gt;
&lt;video controls preload=&quot;metadata&quot; src=&quot;https://ryanheuser.com/notes/2021-03-03/1.mp4&quot;&gt;
&lt;/video&gt;
</content>
  </entry>
  <entry>
    <title>Mapping character networks geographically (Heliodorus)</title>
    <link href="https://ryanheuser.com/notes/2021-03-01-2/"/>
    <id>https://ryanheuser.com/notes/2021-03-01-2/</id>
    <updated>2021-03-01T00:00:00Z</updated>
    <content type="html">&lt;div class=&quot;thread-link&quot;&gt;
&lt;p&gt;&lt;em&gt;Continued from &lt;a href=&quot;https://ryanheuser.com/notes/2021-03-01/&quot;&gt;Mapping character
networks geographically (Homer)&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;
&lt;/div&gt;
&lt;p&gt;In Heliodorus, home is what you flee from. When a descendant of
Achilles falls in love with an Ethiopian woman, he leaves home to follow
her through captivity in Egypt, Persia, Ethiopia. Fighting “fortune” not
gods, “love” transcends and invalidates spatiality, not complete it.&lt;/p&gt;
&lt;video controls preload=&quot;metadata&quot; src=&quot;https://ryanheuser.com/notes/2021-03-01-2/1.mp4&quot;&gt;
&lt;/video&gt;
&lt;p&gt;I’ve put the code on Google Colab and tried to make it easy to use,
in case anyone wants to visualize networks/maps over similar kinds of
annotations. &lt;a
href=&quot;https://colab.research.google.com/drive/1rKe9Vqu58JlaaMCRnp4G_F0mHq2oIuHI?usp=sharing&quot;&gt;colab.research.google.com/drive/1rKe9Vqu…&lt;/a&gt;.
It just needs input: a link to a spreadsheet which lists interactions
(A&amp;lt;&amp;gt;B; B&amp;lt;&amp;gt;C) and where they occur.&lt;/p&gt;
&lt;p&gt;Here’s Lazarillo de Tormes (1554), the immensely popular Spanish
picaresque. It’s hilarious, sordid, &amp;amp; detailed, as Lazaro moves
through 15 towns as servant to various masters, his fortunes rising
&amp;amp; falling. But characters hardly ever recur; the network connects
only through him.&lt;/p&gt;
&lt;video controls preload=&quot;metadata&quot; src=&quot;https://ryanheuser.com/notes/2021-03-01-2/2.mp4&quot;&gt;
&lt;/video&gt;
</content>
  </entry>
  <entry>
    <title>Mapping character networks geographically (Homer)</title>
    <link href="https://ryanheuser.com/notes/2021-03-01/"/>
    <id>https://ryanheuser.com/notes/2021-03-01/</id>
    <updated>2021-03-01T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Experimenting with mapping character networks. I tagged the speech
interactions over Homer’s Odyssey along with where-ish they take place,
and visualized them on a map and as a social network. In Homer, home is
origin, destination, &amp;amp; gravitational center, but in ancient novel:
+&lt;/p&gt;
&lt;video controls preload=&quot;metadata&quot; src=&quot;https://ryanheuser.com/notes/2021-03-01/1.mp4&quot;&gt;
&lt;/video&gt;
&lt;div class=&quot;thread-link&quot;&gt;
&lt;p&gt;&lt;em&gt;The thread continues in &lt;a href=&quot;https://ryanheuser.com/notes/2021-03-01-2/&quot;&gt;Mapping
character networks geographically (Heliodorus)&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;
&lt;/div&gt;
</content>
  </entry>
  <entry>
    <title>Number of creative writing jobs by genre, 2011-2020</title>
    <link href="https://ryanheuser.com/notes/2021-01-12/"/>
    <id>https://ryanheuser.com/notes/2021-01-12/</id>
    <updated>2021-01-12T00:00:00Z</updated>
    <content type="html">&lt;p&gt;@dan_sinykin argh found a bug, here are updated plots. numbers
slightly higher but largely the same.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2021-01-12/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2021-01-12/2.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2021-01-12/3.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2021-01-12/4.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Number of academic jobs in English by job type, 2011-2020</title>
    <link href="https://ryanheuser.com/notes/2020-11-24/"/>
    <id>https://ryanheuser.com/notes/2020-11-24/</id>
    <updated>2020-11-24T00:00:00Z</updated>
    <content type="html">&lt;p&gt;@pruett_jordan Largely changes in the formatting guidelines for the
Wiki, which have kept entries shorter, sometimes lacking words
indicating job type. But I just checked and somehow some with those
words slipped through. Here’s an updated chart. The jobs still unknown:
&lt;a
href=&quot;https://github.com/quadrismegistus/jobcensus/blob/main/synthesis/data.unknown_jobtypes.csv&quot;&gt;github.com/quadrismegistu…&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2020-11-24/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Number of academic jobs in English literature, 2011-2020</title>
    <link href="https://ryanheuser.com/notes/2020-11-23/"/>
    <id>https://ryanheuser.com/notes/2020-11-23/</id>
    <updated>2020-11-23T00:00:00Z</updated>
    <content type="html">&lt;p&gt;The number of jobs advertised in English Lit (on the Academic Jobs
Wiki) is at an all time low: less than half of what it was this time
last November. Like last year, the only subfield not plummeting is
Ethnic Studies.&lt;/p&gt;
&lt;p&gt;(Note: # for 2020 only counts ads posted as of 11/23/2020.)&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2020-11-23/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2020-11-23/2.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2020-11-23/3.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2020-11-23/4.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;Compare last year’s #’s here: &lt;a
href=&quot;https://x.com/quadrismegistus/status/1196462662101680128&quot;&gt;x.com/quadrismegistu…&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;Data and code for this year’s #’s here: &lt;a
href=&quot;https://github.com/quadrismegistus/jobcensus&quot;&gt;github.com/quadrismegistu…&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Quoting &lt;a href=&quot;https://ryanheuser.com/notes/2019-11-18/&quot;&gt;Academic job numbers in English
literature (updated)&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;
&lt;p&gt;Another note: I made some improvements to how the jobs were counted
this year, so all the numbers for previous years have increased in
total, but the shape of the trend is still the same.&lt;/p&gt;
&lt;p&gt;Lastly, thanks to @benmschmidt for his help adding MLA adverts for
this year, gleaned from email alerts. See his analysis of the similar
cataclysm in History jobs here: &lt;a
href=&quot;https://benschmidt.org/post/2020-10-01-jobs-update/2020-10-01-jobs-update/&quot;&gt;benschmidt.org/post/2020-10-0…&lt;/a&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Tracing types of semantic change</title>
    <link href="https://ryanheuser.com/notes/2020-03-14/"/>
    <id>https://ryanheuser.com/notes/2020-03-14/</id>
    <updated>2020-03-14T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Feels strange to post about work stuff, but in my self-quarantine I’m
working on more historical semantics for &lt;a
href=&quot;http://cambridgekeydata.org&quot;&gt;cambridgekeydata.org&lt;/a&gt;: trying to
characterize the types of semantic change undergone by certain keywords
across the 19th and 20th centuries.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2020-03-14/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Measuring anthropomorphism</title>
    <link href="https://ryanheuser.com/notes/2020-03-05/"/>
    <id>https://ryanheuser.com/notes/2020-03-05/</id>
    <updated>2020-03-05T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Vicky Googasian (@v_goog) and I have been measuring the
“anthropomorphism” of objects, animals, and humans in fiction, using
models trained on their syntactic and semantic footprints. We’re talking
about it informally in 20 minutes if anyone wants to join! &lt;a
href=&quot;https://stanford.zoom.us/j/121557684&quot;&gt;stanford.zoom.us/j/121557684&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2020-03-05/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2020-03-05/2.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;Ack! That link broke. Here’s the correct one: &lt;a
href=&quot;https://stanford.zoom.us/j/421834705&quot;&gt;stanford.zoom.us/j/421834705&lt;/a&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Posted talk to King’s College as scroll-based page</title>
    <link href="https://ryanheuser.com/notes/2020-02-24/"/>
    <id>https://ryanheuser.com/notes/2020-02-24/</id>
    <updated>2020-02-24T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Just posted the talk I gave last week to King’s College, with the
text synced to the slides: &lt;a
href=&quot;https://ryanheuser.org/assets/talks/kingscollege2020&quot;&gt;ryanheuser.org/assets/talks/k…&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;It’s a summary of my dissertation work on abstract language in the
novel, and explains how that work addresses “immateriality,” the
official topic of my postdoc.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2020-02-24/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2020-02-24/2.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2020-02-24/3.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2020-02-24/4.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;Also, I should add: any feedback would be much appreciated! Hoping to
edit this into an article submission over the next few months. I’ll also
give a much shorter version of the talk at #ACLA2020, at the
“Abstraction’s Forms” panel organized by @SM_Fallon: &lt;a
href=&quot;https://www.acla.org/program-guide#/seminars/all/26347&quot;&gt;acla.org/program-guide#…&lt;/a&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>A linguistic map of the canon</title>
    <link href="https://ryanheuser.com/notes/2020-01-29/"/>
    <id>https://ryanheuser.com/notes/2020-01-29/</id>
    <updated>2020-01-29T00:00:00Z</updated>
    <content type="html">&lt;p&gt;A linguistic map of the canon&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2020-01-29/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Clarissa under the microscope</title>
    <link href="https://ryanheuser.com/notes/2020-01-16/"/>
    <id>https://ryanheuser.com/notes/2020-01-16/</id>
    <updated>2020-01-16T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Catching up on #Clarissa2020, reading it as I’ve been reading a few
novels now: with words colored from orange (most concrete) to blue (most
abstract). Striking how thoroughly abstract words narrate social space
and actions; concrete words only come in with their violent rupture.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2020-01-16/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2020-01-16/2.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Academic job numbers in English literature (updated)</title>
    <link href="https://ryanheuser.com/notes/2019-11-18/"/>
    <id>https://ryanheuser.com/notes/2019-11-18/</id>
    <updated>2019-11-18T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Updated academic job numbers for English Lit (with data scraped from
Academic Jobs Wiki). Since last posting on Oct 13th, 88 new TT jobs have
been added. But that still leaves us at an all-time low, pretty far into
the season.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2019-11-18/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2019-11-18/2.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2019-11-18/3.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2019-11-18/4.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;* For a better view of Figure 4, see here: &lt;a
href=&quot;https://public.tableau.com/views/EnglishLiteratureJobCensusData/Fig4?:display_count=y&amp;amp;publish=yes&amp;amp;:origin=viz_share_link&quot;&gt;public.tableau.com/views/EnglishL…&lt;/a&gt;&lt;br /&gt;
* For code which generated these, see here: &lt;a
href=&quot;https://colab.research.google.com/drive/1RMxK5zT9mTMZUc9EMFWgrB7R_1cmaeY2&quot;&gt;colab.research.google.com/drive/1RMxK5zT…&lt;/a&gt;&lt;br /&gt;
* For the figures from Oct 13th, see here: &lt;a
href=&quot;https://x.com/quadrismegistus/status/1183433276658274304&quot;&gt;x.com/quadrismegistu…&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Quoting &lt;a href=&quot;https://ryanheuser.com/notes/2019-10-13/&quot;&gt;Academic job numbers in English
literature&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Academic job numbers in English literature</title>
    <link href="https://ryanheuser.com/notes/2019-10-13/"/>
    <id>https://ryanheuser.com/notes/2019-10-13/</id>
    <updated>2019-10-13T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Working on scraping the Academic Jobs Wiki to see if it can yield
more accurate recent job numbers. Here’s what I’ve got so far. Note that
all subfields are declining in jobs besides Ethnic Studies.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2019-10-13/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2019-10-13/2.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2019-10-13/3.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2019-10-13/4.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;* Here’s a PDF of the 4th figure for better reading: &lt;a
href=&quot;https://www.dropbox.com/s/gcr2ofhsc9bxf5e/Fig4.pdf?dl=0&quot;&gt;dropbox.com/s/gcr2ofhsc9bx…&lt;/a&gt;&lt;br /&gt;
* Data generated: &lt;a
href=&quot;https://drive.google.com/file/d/1-4zgztJVXwLbrIBV-6GHIUIB9Ih-p73y/view?usp=sharing&quot;&gt;drive.google.com/file/d/1-4zgzt…&lt;/a&gt;
(feel free to edit!)&lt;br /&gt;
* Code used to generate: &lt;a
href=&quot;https://colab.research.google.com/drive/1RMxK5zT9mTMZUc9EMFWgrB7R_1cmaeY2&quot;&gt;colab.research.google.com/drive/1RMxK5zT…&lt;/a&gt;
(feel free to copy, edit and re-run)&lt;/p&gt;
&lt;p&gt;Disclaimers: Who knows to what extent these trends reflect changes in
wiki usage, or changes in the number of jobs. I suspect that rises in
2011-2014 reflects rise in usage of wiki and standardization of formats,
but I could be wrong. The data could also be refined more manually.&lt;/p&gt;
&lt;p&gt;Lastly, I had to aggregate/standardize some of the subfield names,
some of which changed name on the wiki midway through this period. These
name changes are reflected here: &lt;a
href=&quot;https://drive.google.com/file/d/1-C0R4K1HfkVjxeamc5ovqn6zsMCT7iNN/view?usp=sharing&quot;&gt;drive.google.com/file/d/1-C0R4K…&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;(Update to original October thread: see the numbers for month later,
Nov 18, here: &lt;a
href=&quot;https://x.com/quadrismegistus/status/1196462662101680128&quot;&gt;x.com/quadrismegistu…&lt;/a&gt;)&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Figure 1 of my dissertation</title>
    <link href="https://ryanheuser.com/notes/2019-08-28/"/>
    <id>https://ryanheuser.com/notes/2019-08-28/</id>
    <updated>2019-08-28T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Figure 1 of my diss, “Abstraction: A Literary History.” Filing this
Friday&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2019-08-28/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Measuring personification</title>
    <link href="https://ryanheuser.com/notes/2019-02-07/"/>
    <id>https://ryanheuser.com/notes/2019-02-07/</id>
    <updated>2019-02-07T00:00:00Z</updated>
    <content type="html">&lt;p&gt;For my chapter on personification, I’m exploring ways machine
learning can help us trace the ‘human likeness’ of non-human nouns
across history. Given only which verbs a noun like ‘nation’ is a subject
or object of, how human-like does the noun appear to the model?&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2019-02-07/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2019-02-07/2.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;h/t to methods developed by @Ted_Underwood, @dbamman, and Sabrina Lee
in their article “The Transformation of Gender in English-Language
Fiction” (&lt;a
href=&quot;http://culturalanalytics.org/2018/02/the-transformation-of-gender-in-english-language-fiction/&quot;&gt;culturalanalytics.org/2018/02/the-tr…&lt;/a&gt;),
which I’ve tried to adapt and rebuild for a human-vs-nonhuman
context&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Agency Index</title>
    <link href="https://ryanheuser.com/notes/2018-11-05/"/>
    <id>https://ryanheuser.com/notes/2018-11-05/</id>
    <updated>2018-11-05T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Which nouns are most likely to be the subjects of sentences? And how
has that likelihood changed over time—in poetry, a genre where
grammatical subjecthood is manipulated by forms of personification?
Historically, autumn’s agency falls; commerce’s waxes and wanes; &amp;amp;
child’s rises&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2018-11-05/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2018-11-05/2.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2018-11-05/3.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Character space in Sense and Sensibility</title>
    <link href="https://ryanheuser.com/notes/2018-07-12/"/>
    <id>https://ryanheuser.com/notes/2018-07-12/</id>
    <updated>2018-07-12T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Two character-spaces in Austen’s Sense and Sensibility, visualized.
Which words are more likely to appear within the character-space of
Elinor (“sense”), and which in the character-space of Marianne
(“sensibility”)? An experiment using @dbamman’s BookNLP (&lt;a
href=&quot;https://github.com/dbamman/book-nlp&quot;&gt;github.com/dbamman/book-n…&lt;/a&gt;)&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2018-07-12/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Anti-metricality</title>
    <link href="https://ryanheuser.com/notes/2018-04-16/"/>
    <id>https://ryanheuser.com/notes/2018-04-16/</id>
    <updated>2018-04-16T00:00:00Z</updated>
    <content type="html">&lt;p&gt;What happens when you tell a computer to scan lines of poetry for
their meter, except you sneak in lines of prose, too? You end up
quantifying “anti-metricality,” a text’s resistance to scansion. This
resistance, and the metrical distance between verse and prose, has a
history.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2018-04-16/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Abstraction vs. judgment</title>
    <link href="https://ryanheuser.com/notes/2018-01-30/"/>
    <id>https://ryanheuser.com/notes/2018-01-30/</id>
    <updated>2018-01-30T00:00:00Z</updated>
    <content type="html">&lt;p&gt;As abstract words get more concrete (moving down on y-axis), the
range of their moral value (along x-axis) narrows.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2018-01-30/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Computational Keywords</title>
    <link href="https://ryanheuser.com/notes/2018-01-23/"/>
    <id>https://ryanheuser.com/notes/2018-01-23/</id>
    <updated>2018-01-23T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Computational Keywords, or Modern Abstractions: Using computational
semantics to find which abstract nouns changed the most in meaning
during the transition into modernity. (= DH work I’m doing now in my
dissertation on abstraction in 18th-century literature.)&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2018-01-23/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Abstraction vs. agency</title>
    <link href="https://ryanheuser.com/notes/2017-11-06/"/>
    <id>https://ryanheuser.com/notes/2017-11-06/</id>
    <updated>2017-11-06T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Visualizing the components of personification: which abstract nouns
are more likely to be given syntactic agency in 18C literature?&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2017-11-06/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Sociography of the 18th century print market</title>
    <link href="https://ryanheuser.com/notes/2017-09-26/"/>
    <id>https://ryanheuser.com/notes/2017-09-26/</id>
    <updated>2017-09-26T00:00:00Z</updated>
    <content type="html">&lt;p&gt;London’s print market doubled between the 1740s &amp;amp; 1790s, with
publishers further dividing space of positions. Data:
text&amp;lt;&amp;gt;publisher in ECCO.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2017-09-26/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2017-09-26/2.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;Inspired by @Michael_A_Gavin’s “hist. text nets” (&lt;a
href=&quot;https://muse.jhu.edu/article/634558&quot;&gt;muse.jhu.edu/article/634558&lt;/a&gt;)
&amp;amp; @goldstoneandrew’s “doxa of reading” (&lt;a
href=&quot;http://www.mlajournals.org/doi/pdf/10.1632/pmla.2017.132.3.636&quot;&gt;mlajournals.org/doi/pdf/10.163…&lt;/a&gt;)&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Transformation of ‘labour’</title>
    <link href="https://ryanheuser.com/notes/2017-07-28/"/>
    <id>https://ryanheuser.com/notes/2017-07-28/</id>
    <updated>2017-07-28T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Raymond Williams: “From C17.. labour.. lost its.. association with
pain” &amp;amp; became “a general social activity.. a term in political
economy”&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2017-07-28/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Understanding OCR errors</title>
    <link href="https://ryanheuser.com/notes/2017-07-13/"/>
    <id>https://ryanheuser.com/notes/2017-07-13/</id>
    <updated>2017-07-13T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Working toward better understanding how OCR quality affects word
frequency results. Here’s a micro-experiment showing it does, big
lcague.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2017-07-13/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2017-07-13/2.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Most reprinted 18th century texts</title>
    <link href="https://ryanheuser.com/notes/2017-07-03/"/>
    <id>https://ryanheuser.com/notes/2017-07-03/</id>
    <updated>2017-07-03T00:00:00Z</updated>
    <content type="html">&lt;p&gt;The 30 most reprinted works in the 18th-C? [A is a “reprint” of B if
published later and if A&amp;amp;B have a very high degree of textual
overlap.]&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2017-07-03/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2017-07-03/2.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;dear 18Cists: what analog sources could I assess these results by? I
know Raven/Garside for fiction. St Clair’s Reading Nation? Any
others?&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Corpus vs. bibliography</title>
    <link href="https://ryanheuser.com/notes/2017-06-23/"/>
    <id>https://ryanheuser.com/notes/2017-06-23/</id>
    <updated>2017-06-23T00:00:00Z</updated>
    <content type="html">&lt;p&gt;How representative of the 18C bibliography is the 18C digital
archive? Graphs inspired by @KatherineBode’s article and its recent
discussion&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2017-06-23/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2017-06-23/2.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Exponential rise of print market in 18th century</title>
    <link href="https://ryanheuser.com/notes/2017-06-15/"/>
    <id>https://ryanheuser.com/notes/2017-06-15/</id>
    <updated>2017-06-15T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Stepping back from computational formalism and the “doxa of reading”
(@goldstoneandrew’s term) to look at publishing market trends in
ESTC.&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2017-06-15/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Computational model takes standardized test</title>
    <link href="https://ryanheuser.com/notes/2017-04-28/"/>
    <id>https://ryanheuser.com/notes/2017-04-28/</id>
    <updated>2017-04-28T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Would #word2vec get into grad school? A w2v model taking the Miller
Analogies Test scores in the 85th percentile: &lt;a
href=&quot;http://ryanheuser.org/word2vec-vs-the-mat&quot;&gt;ryanheuser.org/word2vec-vs-th…&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2017-04-28/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;attn: @igorbobic, @teddyroland, @benmschmidt, @Michael_A_Gavin,
@a_e_lang, @clured, @dmimno&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>World literary trends (part 2)</title>
    <link href="https://ryanheuser.com/notes/2017-03-26/"/>
    <id>https://ryanheuser.com/notes/2017-03-26/</id>
    <updated>2017-03-26T00:00:00Z</updated>
    <content type="html">&lt;p&gt;A literary history of abstract and concrete words in English, French,
and German fiction, poetry, and prose, 1700-1900. #WordVectorsIn18C&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2017-03-26/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2017-03-26/2.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>World literary trends (part 1)</title>
    <link href="https://ryanheuser.com/notes/2017-03-25/"/>
    <id>https://ryanheuser.com/notes/2017-03-25/</id>
    <updated>2017-03-25T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Abstract words rise in frequency across 18th century, then fall
across 19th century—in both French and English fiction.
#WordVectorsIn18C&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2017-03-25/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>ELIZA’s descendants</title>
    <link href="https://ryanheuser.com/notes/2016-12-18/"/>
    <id>https://ryanheuser.com/notes/2016-12-18/</id>
    <updated>2016-12-18T00:00:00Z</updated>
    <content type="html">&lt;p&gt;To prep for our MLA talk on the 50th anniversary of ELIZA—the first
chatbot—my algorithmic colaborator @bardbott is writing (bad)
sonnets!&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2016-12-18/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;The panel is chaired by @yohei_igarashi and will feature @ncecire and
Giorgina Paiella. Come! Thurs, Jan 5, at 7pm: &lt;a
href=&quot;https://apps.mla.org/program_details?prog_id=156&amp;amp;year=2017&quot;&gt;apps.mla.org/program_detail…&lt;/a&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Semantic networks</title>
    <link href="https://ryanheuser.com/notes/2016-09-15/"/>
    <id>https://ryanheuser.com/notes/2016-09-15/</id>
    <updated>2016-09-15T00:00:00Z</updated>
    <content type="html">&lt;p&gt;semantic networks! top 4000 strongest associations between the 2000
most frequent abstract nouns-&amp;gt; giant component-&amp;gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2016-09-15/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Semantic fields in vector space</title>
    <link href="https://ryanheuser.com/notes/2016-06-16/"/>
    <id>https://ryanheuser.com/notes/2016-06-16/</id>
    <updated>2016-06-16T00:00:00Z</updated>
    <content type="html">&lt;p&gt;k one last #realtimeDH tweet: the semantic fields from Pamphlet 4 in
the vector space models for ECCO-TCP and ECCO&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2016-06-16/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2016-06-16/2.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Diachronic semantic networks</title>
    <link href="https://ryanheuser.com/notes/2016-06-15/"/>
    <id>https://ryanheuser.com/notes/2016-06-15/</id>
    <updated>2016-06-15T00:00:00Z</updated>
    <content type="html">&lt;p&gt;+so, #realtimeDH tweet: working now on comparing “semantic networks”
between abstract nouns across 18C decade-models&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2016-06-15/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2016-06-15/2.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2016-06-15/3.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2016-06-15/4.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
  <entry>
    <title>Reading at meso-scale with Moby-Dick</title>
    <link href="https://ryanheuser.com/notes/2014-06-09/"/>
    <id>https://ryanheuser.com/notes/2014-06-09/</id>
    <updated>2014-06-09T00:00:00Z</updated>
    <content type="html">&lt;p&gt;Distribution of non-dialogue, non-past-tense paras in Moby-Dick shows
essayistic mode is broader than cetology chaps&lt;/p&gt;
&lt;p&gt;&lt;img src=&quot;https://ryanheuser.com/notes/2014-06-09/1.jpg&quot; alt=&quot;&quot; loading=&quot;lazy&quot;&gt;&lt;/p&gt;
</content>
  </entry>
</feed>
