A Glimpse Inside the Vanity Project

the course outline and the concepts data file

Claude and I have the wiring working for what we call the Vanity Project. You can see the working demo here. The dialogues that are not up to the quality that I have in mind, especially the first two chapters. I plan to leave the demo up in its sketchy form for now, while I work on putting an entire plan together. Then I will come back to tackle improving the dialogues.

For the Vanity project, the goal is to create an immersive, interactive seminar in which a student can learn about human interdependence as I would teach it. To complete the project, I have to revise and expand the course outline and flesh out the concepts data file. Then I instruct Claude to construct the dialogues. That takes very little time—I give some basic instructions, get a result from Claude, then see how I need to change the instructions, get a new result, and so on. Usually only 3 or 4 iterations, is completed in less than an hour.

Since organizing the Vanity Proect is mostly my job, not Claude’s, and because I would rather spend time with my children and grandchildren, this could take several weeks.

The larger project, of Substack for Professors, is deferred until the Vanity Project is complete. Substack for Professors would have Claude interview a professor about a course and then convert the answers to the interview into a seminar like mine. This would require many enhancements, to give professors the option to have different students, different classroom dynamics, articulate how students are to be assessed, and so on. I probably would want funding for this project, so that I could hire a developer to work on it with Claude.

Below is the format in which I am keeping the course outline. It consists of chapters and concepts. This is for my own benefit. Claude does not need it.

course outline

chapter: Problems with Studies

concept: spurious correlation

concept: reverse correlation

concept: p-hacking

concept: Bryan Caplan’s notion of social desirability bias

concept: first iron law

[more]

chapter: personality psychology

concept: Big Five

concept: HEXACO

concept: Dark Triad

chapter: evolutionary psychology

concept: The Blank Slate

concept: Warriors and worriers

concept: The Moral Dyad

chapter: cultural evolution

concept: humans' unique ability to transmit and expand knowledge

concept: advantages of copying

concept: we decide what to believe by deciding who to believe

chapter: game theory

concept: individual rationality can lead to bad collective outcome

concept: Dunbar's number

concept: informal cooperation mechanisms and formal cooperation mechanisms

chapter: theories of motivation

concept: economic man

concept: Darwinian man

concept: status man

concept: sociological man

concept: the intimate world and the remote world

chapter: forms of social order

concept: the rule of the clan

concept: Big Gods

concept: the natural state

concept: the open-access order

concept: hard to transition between the two orders

chapter: 21st century disorder

concept: the revolt of the public

concept: donscription

concept: Haidt and Twenge on smart phones

chapter: social epistemology

concept: misinformation and fact-checking

concept: naive realism

concept: knowledge/power discrepancy (J. Friedman, Kling)

concept: concept: FITS

concept: concept: Crediable substack

chapter: tribalism

concept: three languages of politics

concept: the law of asymmetric insight

concept: elite vs. populists

concept: brokenists vs. institutionalists

concept: luxury beliefs

chapter: government pathologies

concept: the intention heuristic

concept: trust and accountability

concept: selection for leaders with bad traits

concept: selection for bureaucrats with bad traits

chapter: government and banks

concept: commitments (see state and commitments)

concept: The Cash Nexus

concept: Regulators' calculation problem and 2008

(this is not the end of the seminar, just how far along I am)

Concept cross-reference table

Concepts often get mentioned in multiple chapters. When a concept is mentioned, it will be highlighted so that a user can click and see a definition and the chapter where it is discussed in the most depth. In the current demo, this is only active in the test chapter.

Below are the concepts whose “home chapter” will be the first chapter on Problems with Studies. This is the format that Claude understands and that will be stored as a JSON file. For each concept, “short” is a phrase that will show up in a seminar dialogue, and “long” is a description that will come up when the user clicks on it.

short: binary data

long: data from situations that can be either/or, such as Democrat or Republican

short: categorical data

long: data that from situations that can have multiple discrete categories, such as Democrat, Republican, or Green Party.

short: scalar data

long: data that come from situations where there can be more or less, such as the number of votes a candidate received, or the percentage of votes a candidate received

short: spurious correlation

long: correlation that appears in a small sample and does not hold up in a larger population

short: reverse causation

long: when someone asserts that X causes Y, when it is possible or even likely that Y causes X

short: third-factor causation

long: also called Confounding Variable. when X and Y are correlated because a third factor, Z, is causing both

short: First Iron Law

long: First Iron Law of Social Science. Sometimes it's this way, and sometimes it's that way. Reminds us not to over-generalize.

short: p-hacking

long: in layman's terms p-hacking is when a researcher keeps tweaking the analysis to get a result that is likely to be worthy of publication

short: social desirability bias

long: Economist Bryan Caplan's claim that it is easier to get a paper published if its results support opinions about the topic that are popular with academics.

short: replication crisis

long: replication means getting similar results with a similar study done by another researcher. The crisis is that many studies, in psychology, epidemiology, and other fields, including studies heavily cited, have failed to replicate

short: Hill criteria

long: Bradford Hill's criteria for strengthening the case for causality in relationship found in epidemiology. includes consistency of findings (not just one study), plausible causal mechanism, and larger values of the causal variable leading to larger effects in the result variable.

short: Measurement error

long: measurement error reduces correlation. Conversely, if there is strong correlation, we can infer that measurement error is not a major issue

[next will come concepts whose “home chapter” will be the Personality Psychology chapter]