fbpx
How to know if you’re talking to a wokescold: A scientific method for preventing IRL flame wars

[FYI: If you’re interested in data-blogging like this, I’m offering a little free course on it.]

We’ve all been there: You’ve had a couple drinks, you’re having fun talking with someone, then you blurt out a controversial opinion and everything goes belly up. Maybe your interlocutor scolds you, maybe they just walk away, or maybe nothing happens but there’s gossip a week later…

If you have controversial opinions, what you need is a method for knowing — in advance — whether your conversation partner can handle them. It needs to be simple and quick enough to be practical, but it needs to be scientific enough to offer real predictive validity.

It recently occurred to me that there exists a statistical technique that solves exactly this problem. It’s called recursive partitioning, and the practical tool it produces is called a decision tree. If you have data on public opinion and other demographic variables, you can use statistics to determine which chain of questions will give you the best guess about someone’s position on any given issue. If we create a decision tree to predict their position toward suppressing naughty opinions, then we have a simple, practical, and scientifically valid “life hack” for avoiding IRL flame wars.

Analysis

I did this last week and the results are very interesting. If you’re interested in the statistical details, or you’d like to run the code yourself (perhaps on a different outcome variable), you can find all of that here. In this post, I’ll focus on the social and practical implications.

Here’s all you need to know about the stats. In this analysis, “being a wokescold” is proxied by whether or not someone thinks racist speakers should be allowed or disallowed. For possible predictor variables, I included a handful of variables that are reasonable to ask someone about or easy to observe yourself.

Specifically:

  • sex/gender = variable named sex
  • race = variable named race
  • left/right identification = variable named pol
  • family income = variable named realinc
  • college attendance = variable named college
  • word knowledge or verbal skill (proxy for IQ) = variable named wordsum

I then conducted recursive partitioning, which breaks the data down into the sequence of branches giving the most predictive traction over the outcome variable.

Results

Figure 1 plots the resulting decision tree.

Figure 1

The graph is fairly intuitive, and if you’d like to understand the numbers better, see my more technical post over at jmrphy.net. Here I will give you a more concise and practical translation, resulting in a simple heuristic you can memorize.

If you meet a random person, there’s a 38% chance they’re a wokescold (defined as wanting to suppress racist speakers; one can debate this, but whatever, it’s a decent proxy).

The very first and most important question you can ask someone, to avoid a flame war, is: “Did you ever go to college?" If they say yes, the probability of them being a wokescold drops to 29% and that’s your best guess: They are probably not a wokescold. Nothing else will improve your guess from this point (at least from the variables we selected).

Now, many of you will say: But it’s the college-educated wokescolds one should be most afraid of! True. The limited utility of this analysis is also it’s primary social-scientific value: It reminds us that college-educated wokescolds remain a relatively minor anomaly, quantitatively speaking. Being educated still means you’re much more likely to support unsavory expression. It’s true that educated wokescolds are often the most dangerous landmines we’d like to tiptoe around, and unfortunately my particular analysis this week will not help you on this front. Fortunately, I have an alternative algorithm custom made for this use-case: If they went to college and they’re also a female with dyed hair, hold fire on your nuclear takes: They are probably a wokescold. Unless they’re Amber Frost.

If they never went to college, the next question you have to ask yourself is whether they're smart. You probably don't want to give them a vocabulary test, but conversation is pretty revealing. If they are smart, you infer they are not a wokescold (40% chance). If they are dumb, it's now a coin flip (50%).

Next, what is their race? This you can probably guess yourself. If white, this bumps them very slightly toward not being wokescolds (48%). If non-white, this bumps them toward being wokescolds (57%). From here:

If they are white and male, there's a 45% chance they’re a wokescold so you infer they are not — and that’s your final guess. If they are white and female, you should see if their family is rich or not. If rich, they are slightly less likely than a coin flip to be a wokescold (46%); if poor, they're slightly more likely than a coin flip to be a wokescold (54%).

If they are dumb and non-white, there is a 57% they’re a wokescold and that’s your best guess.

A heuristic you can memorize

(This only applies in America, mind you, the land of the free.)

  1. If they’re a female who signals creativity or virtue (e.g., dyed hair, bumper stickers), don’t share any edgy takes (this is post-hoc to the model, just a precaution in light of data limitations and researcher experience).

Otherwise:

  1. If they went to college, they’re probably not a wokescold. You may gradually begin to share your edgy takes.
  2. If they did not go to college, but speak more intelligently than average, they are probably not a wokescold. You may gradually begin to share your edgy takes.

For all others, the safest decision rule is to not share edgy takes. Bonus rule only if you can master the above 3-step algorithm and you have an appetite for risk:

  1. If they are rich white people, you may gradually begin to share your edgy takes.

What about ideological identification?

The most intriguing result here, to my mind, is that ideological identification totally drops out — it appears to have no predictive power! As I wrote in my technical post:

[That ideological identification has no predictive power] is fascinating, given that many people today tend to think of speech suppression as a fashion on the educated Left! And it is, but that's only a highly visible minority. Political scientists would not be surprised by this result: We've long known that leftists and educated people are always more supportive of free expression (you just don't hear about those people in the media right now).

Limitations

Please note that the model here does not provide especially satisfying statistical discrimination. It’s better than nothing, but one must still proceed carefully. Always begin by sharing mildly provocative takes, and watching your interlocutor’s reactions. Do not advance to nuclear takes until several acts of mild edgelording produce only smiles, laughter, or excited edgy reciprocity. With additional data and more sophisticated modeling, we may hope to derive more confident predictions for more ambitious social maneuvering. Until then, be careful.

If you're a content creator, you're a tiny startup (you just don’t know it yet)

All of the following terms from Silicon Valley startup culture have direct analogues in the content-creation game. In the examples below, I’ll discuss a blogger or video producer who earns income via Patreon, but the principles generalize much more widely.

Product-Market Fit

Product-market fit is when it becomes clear that a few of your random blog posts get more traction than all the rest — and that these few have something in common. Or when a set of your Youtube videos start earning a lot of positive comments revolving around a particular theme — and this set of videos has that theme in common. In short, it’s when you learn from data where your own autonomous creative tendencies (all your “product features”) intersect (fit) what people want (market). As with startups, this is one of the most crucial turning points in the lifecycle of a creative/intellectual endeavor on the internet. It’s not where amateur dabbling necessarily turns pro, but it is where amateur dabbling has an opportunity to turn pro.

Cost of Acquisition

Cost of acquisition (CAC) is the amount of effort/time it takes you to get a new patron. Consider the following example.

You publish one Youtube video per week.

You spend about 5 hours recording, editing, and posting each video

You gain one new patron per month

The cost of acquiring one customer — your CAC — is 4 x 5 = 20 hours. What is the value of your time? Estimate the highest hourly wage you could currently obtain from an employer. If the most you could earn per hour is $20, then you are effectively spending 20 x 20 = $400 to obtain one customer. Is that good or bad? Depends on how much you earn from a customer, and how much you value money relative to the intrinsic rewards of doing your work.

Churn Rate

Churn is the rate at which your patrons cancel their pledges. Every content creator has some sense of how frequently they pick up new patrons, because picking up a patron is delightful (and Patreon sends you an email). But many content creators pay zero attention to how frequently patrons disappear. Losing patrons is sad, so we’d rather not pay too much attention (and Patreon does not send you an email). But if you are serious about succeeding as a content creator, you need to be as aware of your churn as you are currently aware of your rate of acquisition.

Simply put, if your churn is too high, you will never be able to go full time (at least not via Patreon). Churn rate is also a useful signal in the short-term because it tells you if your subscribed patrons are content with what you’re putting out and how you’re putting it out. If churn is uncomfortably high, that means it will be worth it to allocate time and effort to improving what you post exclusively for patrons, how often, etc. If churn is low, don’t waste time revamping Patreon and rather focus on public content that will acquire new patrons. There are surprisingly many ways to calculate churn. Here is the one that makes the most sense for content creators. To calculate your churn rate, take the total dollar value of all cancelled and decreased pledges in the previous calendar month, and divide it by the total value of all your pledges at the beginning of that month. For example, assume the following.

Between Sept. 1 and Sept. 30, two $10 patrons canceled and one $50 patron decreased their pledge to $25. That’s a total of $45 exiting your recurring revenue stream.

On September 1, you had $800 in pledges.

Your churn rate is 45 / 800 = 0.06. Multiply by 100 for a nice percentage figure: 6%.

You could do this for a few of the most recent months and average across them for a more robust sense of your churn.

Communications from Patreon staff suggest that the average churn rate is about 5%. (They say that’s a “healthy rate,” which I’m assuming means “average.”) You might therefore decide that a churn rate higher than 5% deserves your attention and effort, whereas a churn rate at 5% or lower does not warrant any increased effort on Patreon management.

If it’s not obvious, your growth rate has to be higher than your churn rate to succeed in the long-run.

Lifetime Value

The lifetime value (LTV) of a patron is the size of their pledge multiplied by the number of months they stay subscribed. If the average patron gives $5, and continues to give for 6 months, the average LTV is 5 x 6 = $30. This hypothetical content creator is not a promising startup.

However, this basic formula somewhat underestimates LTV if you plan to sell additional products or services outside of Patreon, such as books or courses. You might assume a 2-3% conversion rate, such that 2 or 3% of patrons will buy whatever additional products you sell later. If you make 3 courses for $500 each, and every patron has a 2-3% chance of buying, then you can add to each patron’s LTV 1500 x .025 = 38. In this example, your average LTV is really more like 68.

Communications from Patreon staff suggest the average amount of time a patron remains a patron is about three months.

The ultimate benefit of understanding and calculating these metrics is to gain an honest, objective picture of your project’s financial viability. Adding these numbers to your understanding might produce some somber insights about your project at first, but they all suggest viable solutions and even tell you how to prioritize and sequence your next steps: increase your planned products outside of Patreon, shift effort from retaining to acquiring patrons, etc.

Was this helpful? You should let me know because I’m not sure how many of my readers are interested in this stuff. If you are, I could say much more on this…

The interaction of education and race on Trump approval

The correlation between education and support for Trump is very different across the black-white divide. The graphs below I have taken from Civiqs.

For white people with no college degree, a small majority approves of Trump:

white Trump support education

For white postgraduates, a small majority disapproves of Trump. Interestingly, this is more Trump support from white postgrads than I would have thought:

white postgraduates Trump approval

For black people with no college degree, a huge majority disapprove of Trump:

Approval of Trump for black people with no college degree

And for black postgraduates, the distribution of Trump approval is… about the same as it is for black people with no college degree.

Approval of Trump for postgraduate black people

This surprised me. At first I thought there was a glitch in the browser, I had to refresh it for the different subsets to make sure this wasn’t a mistake.

So what’s going on here? It’s genuinely unclear to me, but there are only a few plausible possibilities. One possibility is that this variation is just an artifact of other variables. But if education does have some effect on attitudes toward Trump, is there a reason why would it would be different for white and black folks? Who knows, but it’s interesting enough to hypothesize about. Scholarly literatures on the relationship between education and political attitudes sometimes debate whether education has an income effect (grads think differently because their market position is different), a learning effect (grads think differently because they have more information or knowledge), or a socialization effect (grads think differently because they enter into cosmopolitan social circles). Which one of these mechanisms could account for an educational effect on Trump support, conditional on race, where education shifts white people toward disapproval while shifting black people nowhere?

An income effect is conceivable, in which the better jobs and salaries won by white postgrads makes some of them change their mind toward disapproval of Trump. But other research suggests that income, apart from education, was not really an independent driver of Trump support.

A learning effect seems to me unlikely, in part because university education is probably not about learning, but also because I see no reason why black students would be less likely than white students to learn new reasons for disliking Trump. It’s possible that black people are so opposed to Trump that education doesn’t really have much room to exert a unique, additional effect; or that whatever university teaches, black people already know it from childhood, e.g. that White Supremacy is real. So education perhaps only affirms what black people already know; whereas many white children do not know that White Supremacy is real, but university teaches them the error of their youthful ways. But if this were the case, it would be unclear why black people bother to attend university; also, you’d have to believe that university teaching is, at least for white students, a hard change of course from 5th grade civics class, to have such an effect; but it seems to me that 5th grade and 15th grade teachers have a pretty unified message that racism is bad and that one should not grab women by their pussies, and that anyone who does or says such things should not be President. I don’t see what exactly university would teach white people that departs from what the education system already taught them. So I don’t see how an education effect could be a learning effect.

Personally, my priors are more in favor of the socialization mechanism. What university lecturers teach is not radically different from what 5th grade civics teachers teach, but the club is very different. If you got a 5th grade civics class, everyone you knew got a 5th grade civics class. There is no club. If you go to university, you leave behind the townies who do not go to university. It’s basic sociological knowledge that all clubs use symbols and rituals to distinguish members from outsiders, and members receive a premium of resources, care, and attention from other members. The culture of the university club is best defined by cosmopolitanism. Why cosmopolitanism is the culture of the university, and how the features of cosmopolitanism serve its members, are topics for a separate post. For now, suffice it to say that cosmopolitanism is the opposite of chauvinism, nationalism, aggression, etc. Cosmopolitanism is the sublimation of these drives into polite speech, which conquers inferiors through competitive subtlety rather than competitive… competition, which is brutish and too obvious. Anyway, it seems plausible that entry into the cosmopolitan social club would have a significant effect, in the direction consistent with the data: away from Trump. But why would the socialization effect be conditional on race, when above I argued there’s no reason a learning effect would be conditional on race? Well, I think there’s a good reason that university would socialize white students into Trump disapproval, while having no such socialization effect on black students. Cosmopolitanism includes compassion for the weaker ‘other.’ As black people in the United States suffer disproportionately from poverty and other ills, white students who enter the university club must become more compassionate toward America’s oppressed black population — as a ritual requirement of membership, mind you, not for any reason that has to do with information, knowledge, or learning. Black students who traverse the university system might become more compassionate for female garment workers in the Global South, but membership in the university club does not require them to increase their expressed compassion to black people in the United States. On the contrary, cosmopolitanism gives them an increased sense of their deserved seat at the table. In short, the cosmopolitan or extra-civilized gain symbolic power over the less civilized, by forfeiting their right to brute force, investing in the social club of advanced symbolic manipulators, and cultivating their symbolic facilities in lieu of their brute force facilities. The more ridiculous social justice fashions today — sometimes led by students of color and supported secondarily by white 'allies’ — are no better or worse than than social justice fashions popular among the educated white elite of any previous generation: cosmopolitanism always means telling refined fibs to secure resources away from the grabbing hands of those who are unable to tell refined fibs.

In summary, I hypothesize that education exerts a socialization effect on students, and that such an effect should alter Trump support only in the case of white students.

The content of this website is licensed under a CREATIVE COMMONS ATTRIBUTION 4.0 INTERNATIONAL LICENSE. The Privacy Policy can be found here. This site participates in the Amazon Services LLC Associates Program, an affiliate advertising program designed to provide a means for us to earn fees by linking to Amazon.com and affiliated sites.