Discourse Platforms
A Comment on Social Media, Higher Education, Virtuous Leadership and Lifelong Learning
This post was actually intended to be a comment here, but due to its length I’m parking it here for now. Keep in mind, that it’s not a polished piece. It’s just a comment.
From the podcast, Arnold says, “AI is now where the web was 30 years ago.” My takeaway from this isn’t to rush into the next gold rush, but rather to think carefully about the lessons learned from the past 30 years of internet history, focusing on a few key implementations of “discourse platforms.” Or rather precursors to discourse platforms. What is a “discourse platform?” It’s a town square or forum for Socratic discourse. Social media on the other hand is a look-at-me selfie-fest in which engagement is optimized to maximize profit. A discourse platform is more sophisticated. The best example of it is Substack; itself a precursor to a “Network Based University.” We must be careful here though, as the concept of a “university” automatically begins to limit our creativity. We don’t want to unnecessarily constrain ourselves with academic models of the past.
We should do a root cause analysis here on higher-education and another on social media. The one on higher education is probably mostly complete (see Richard Vedder’s book Restoring the Promise: Higher Education in America). We should ask why Substack is so much better than Facebook for Socratic dialogue and how it can be even better? Same with Twitter. How do we incorporate in-person discourse and community with online discourse and community.
The ultimate goal here is to promote virtuous leadership, virtuous governance and lifelong learning. As we’ve learned from FITs, AI grader, and the success with Three Languages, there’s a growing theme here pointing toward incentives.
The rest of this comment aims to devise better incentives for AI graders, discourse platforms, AI mentors and governance in general, including new ideas from Michael Huemer on the separation of powers.
Some possible next steps.
At some point I will do a more thorough lessons learned of important platforms and technologies such as Facebook, Twitter, and YouTube. This is just the back of the envelope version of that lessons learned. We might call this lessons learned “internet anatomy.” How can we take this internet anatomy to build better discourse platforms?
Probably the biggest improvement that Substack offers is the direct payment from student to teacher, from reader to writer, from learner to educator. It is the closest model to Adam Smith’s vision for students paying professors directly. This promotes good teaching. This is one of the most important facts we have when it comes to understanding what’s wrong with higher education. Government and other third-party funding has messed it all up. Reminds me of how we pay for medical care. In the words of Milton Friedman
"Two simple observations explain both the high level of spending on medical care and the dissatisfaction with that spending. The first is that most payments to physicians or hospitals or other caregivers for medical care are made not by the patient but by a third party—an insurance company or employer or governmental body. The second is that nobody spends somebody else's money as wisely or as frugally as he spends his own. No third party is involved when we shop at a supermarket. We pay the supermarket clerk directly: the same for gasoline for our car, clothes for our back, and so on down the line. Why, by contrast, are most medical payments made by third parties?” From Milton Friedman’s, "How to Cure Health Care.”
Here are some other observations regarding discourse platforms.
Ad-based models like Facebook and Twitter fail to promote virtuous discourse. We want virtuous Socratic dialogue. Same phenomenon as explained by Smith and Friedman. Woke corporations promote woke discourse. Again third-party payment is a problem.
Comment ranking as implemented on Substack is superior to the Twitter and Facebook models. How can it be better though?
Currently the only merit function we have uses “Like” or “null.” Combined with chronological. I imagine that we’ll probably look back on this method of ranking comments as crude and simplistic; worse than a skateboard with metal wheels. “Like” can mean dozens of things. It’s binary; no magnitude other than 1 or 0. Surely there are many times we want to promote a comment or sort it based on some emotion, some idea or some merit other than 0 or 1. This reminds me of the public choice arguments against one vote for one person. Using money to communicate desire has big benefits. We need sophisticated merit functions to promote comments and other components of discourse. We need a more sophisticated selection mechanism that uses emergence. Think of this an “impartial spectator” that promotes virtue.
Rather than just one axis to promote good comments with the most-liked comments at the top of the hierarchy, we might consider a multi-axis model in which comments can be pushed to the sides or diagonally, or in 360 degrees—sorted so to speak into categories chosen by the owner of the Substack. Think of Substack as using a single-axis “Best Work Board.” Up and down only. (I’m stealing this idea from elementary school classrooms in which the teacher motivates students to do their best work by elevating the status of students that do superior work). We want a multi-category “Best Virtue Board” with say 360 degrees of resolution, or even more advanced, 4*Pi steradians—fully spherical. A three-dimensional Best Work Sphere like this might allow for counter-balancing of virtues, like a gravitational force model, in which virtues can be positive or negative.
In order to implement this, we would need merit variables or virtue variables. Call them what you want. These merit variables could be incorporated into a merit function—a differential equation that solves (optimizes) for some desirable output such as leadership or beauty. A higher level optimization algorithm would solve these merit functions, to sort comments into categories onto the Best Merit Sphere. The winners of this contest would then be elevated in status, or showcased in a “Fantasy Intellectual League.” Awards, status, respect would ensue. The ultimate goal is to motivate, and promote virtuous leadership and lifelong learning. An excellent book has already been written on character education. See Bob Luddy’s book The Thales Way.
In a multi-axis comment ranking model, the platform authority (Substack) would allow the teacher (Arnold Kling) to choose the merit variables or virtue variables. We might have a merit sphere for IQ and a virtue sphere for character. Each of these merit variables would be discretized with a certain resolution say on a scale of 0 to 1, or 0 to 10, or 0 to 100. Zero is at the center of the sphere, 100 is on the surface, 50 would be halfway to the surface from the center.
Examples of merit variables (to sort and promote comments) within this sphere might be: creativity, brevity, poetry, or humor. Readers, or maybe just the teacher would have voting/grading power. Or an AI Grader might be trained to grade the comments per the teachers GPT programming interface. The teacher might want to promote other merits/virtues within the comments using merit variables such as “good question”, “best Turing test,” “best business idea,” “best understanding of topic,” and “most nuanced.” There might be others for younger students, such as “self-control,” “best grammar,” “best intro, body and conclusion.”
In the end, rather than a single-axis hierarchy for promoting “good” comments we now have a multi-axis hierarchy, possibly a sphere, in which the teacher customizes the criteria for comments and decides whether grading will be crowd-sourced, teacher-graded, or AI-graded.
Let’s return to our internet anatomy project.
I have not seen any politicians on Substack yet. Thoughts on this? One can argue that we don’t want them here, but we should want our best ideas to be easily communicated to our congresses. Feedback between citizen and government is good, whether that governance is public or private.
This brings up the issue of legislation. Shouldn’t we be discussing proposed and established legislation on a “discourse platform” like Substack?
Let’s imagine for a moment a Supreme Court Substack page in which people are connected in limited ways to discourse surrounding Supreme Court cases. Similarly there might be an Executive Branch Substack page and a Legislative Branch Substack page. These shouldn’t necessarily be created by those branches themselves. We should create them. We can create them. Carefully, based on the Constitution, and improvements to the Constitution. Thoughts?
We can even create a fourth branch of government. And fifth. As many as needed. I’m particularly fond of some of Michael Huemer’s suggestions.
“There are systematic reasons why the U.S. constitution has failed to limit the power of the federal government in the way that it was intended to do. After examining which kinds of constitutional provisions have been respected and which have not, we can devise alternative constitutional provisions that would have a greater chance of successfully limiting the power of government. In particular, (i) there should be supermajority rule for passage of all legislation, (ii) there should be a separate, “negative legislature” with the sole power of repealing laws, and (iii) there should be a separate “constitutional court” with stronger powers for enforcing the constitution than the current Supreme Court.”
Here’s one final question. How can we use the idea of a discourse platform—that promotes virtuous dialogue and lifelong learning—to improve governance, particularly the U.S. Constitution?