Thursday Things: Quantum of Civility Edition

6 February 2025. Vol 7 No 6. By Dan McGirt. #287

Feb 06, 2025

Thursday Things is here! This week AI teaches us to be civil and quantum physics confuses us.

If you enjoy this edition, please click the heart icon in the header or at the end of the post to let me know.

Awww, a basket of kittens! And … argument over. Take that, AI! Photo by Jari Hytönen on Unsplash

AI teaching civil discourse

Thursday Things doesn’t discuss politics, except for dog mayors, because everyone loves dog mayors. But we can discuss discussing politics, because that’s meta and we love going meta.

Civil discourse seems to have become a lost art in our society — or at least that’s the impression one gets if you go by the political, social, and policy arguments visible online and in other media. It has been my impression that in the real world it is still possible to have sane, informative, mutually respectful exchanges of views with people.

But I could be wrong about that.

A growing number of colleges and universities have started programs to teach their students (and perhaps their faculty) the skills and mindset of civil discourse. And some are using AI to do it:

AI Meets Civil Discourse

Since this is an essay about AI, I’ll use an AI-generated answer from Google Search for the definition (Google Search AI collects data from websites on a topic, in this case “civil discourse,” and summarizes the findings):
“Civil discourse” refers to the practice of having respectful and constructive conversations about public issues, where individuals actively listen to opposing viewpoints, express their own ideas clearly, and strive to promote understanding and knowledge expansion, even when there are disagreements, prioritizing civility over personal attacks.
In other words, civil discourse is not about winning an argument but about having a constructive dialogue to learn another person’s views, offer your own, and come to mutual understanding if not agreement.

Which I very much get. I enjoy a spirited debate on politics, or anything else, because I argue to learn. Rarely in a political discussion am I trying to change someone’s mind or “win” the argument. Rather, I want to understand why they think what they think or believe what they believe. And sometimes they don’t know. Sometimes people are just parroting ideas they heard on TV or reflexively taking the side of “their team”, which is rarely an interesting or useful discussion.

I also want to learn where I might be wrong — where there is a fact I’m missing or an idea I’ve never considered or a different way of looking at the question that makes sense to me. These are useful conversations, because then I then I can correct errors in my own thinking, or identify something to dig into and research further. Not that I always have time to do so.

But enough about me.

At the University of North Carolina School of Civic Life and Leadership (SCiLL), one professor decided to test an AI roleplaying tool called Copient.ai as a vehicle for teaching civil discourse.

There are multiple challenges in teaching civil discourse skills to colleges students:

During civil-discourse exercises, it can be difficult for students to openly express their sincere views on controversial topics such as race, gender, abortion, and immigration, as they may fear how they may be seen by classmates or the professor.
The number of students needing to be taught is in the thousands (UNC itself has 20,000 undergraduate students), and it’s difficult for a limited number of faculty to scale civil-discourse training to reach every student.
Scaling the instruction across multiple faculty may mean no standard approach to civil discourse can be instituted across the university.
Students don’t know how to perform civil discourse, so having them roleplay and provide one another feedback isn’t the best approach.
Since conversations are practiced orally, the faculty’s ability to assess individual students’ civil-discourse skills uniformly is limited due to time and location constraints, as well as faculty-member fatigue.

Using an AI discussion partner that generates an avatar students can practice discussions with ameliorates many of these problems. The discussion is private, so the students can freely express their true views without fear of blowback. The AI system can apply a uniform approach. It scales, so that many students can practice at once, while receiving personalized feedback. And so on.

The results of a pilot project were promising:

In the pilot, each student had a conversation with the AI six times: three times on the topic of abortion as practice rounds, followed by three times on the subject of affirmative action. Their final conversation on affirmative action was graded.
The results were very encouraging. In the practice rounds on abortion, the scores started low but rapidly increased. In the second set of conversations on affirmative action, the scores started high (since students had already practiced on the abortion topic) and went higher.
Students were also positive about the tool. Ninety-two percent thought it improved their civil-discourse skills. Eighty-four percent felt it helped them learn civil discourse. Eighty-eight percent felt it graded them fairly.

Read the article for more details. UNC will be expanding on the pilot project. There are also several other AI tools being used elsewhere to help teach civil discourse skills.

Who knows? Maybe instead of killing us all, Skynet will help us get along instead!

37 shades of quantum

I’ve read this 37 times and I still don’t understand it.

Extending a paradox: Quantum mechanics experiment measures a pulse of light in 37 dimensions

A team of physicists affiliated with multiple institutions in China has measured a pulse of light in 37 dimensions. In their paper published in Science Advances, the group explains that their experiment was meant to demonstrate that quantum mechanics is more nonclassical than thought.

Quantum mechanics is more post-modern physics is what I get from this.

In this new effort, the researchers in China sought to see how far nonclassical quantum mechanics differs from classical theory by carrying out an experiment to demonstrate the Greenberger–Horne–Zeilinger (GHZ) paradox.

That old thing?

The GHZ state, developed by David Greenberger, Michael Horne and Anton Zeilinger in 1989, is, in its simplest form, an entangled quantum state that involves a minimum of three subsystems. The GHZ state predicts outcomes that defy classical theory, such as mathematical impossibilities, like 1 = -1, giving rise to a paradox.

To put it in its simplest form. And … say what?

To demonstrate the GHZ paradox in a real-world setting, the research team found a way to produce photons that exist in 37 dimensions using coherent light from a laser, entanglement and a fiber-based photonic processor. This achievement involved far more dimensions than the three required to satisfy the GHZ paradox.

I’d double check this myself, but my fiber-based photonic processor is in the shop.

According to the researchers, the experiment helps to clarify some of the unknowns in quantum theory on a deeply fundamental level. … They conclude that their experiment opens the door to new avenues of research surrounding various aspects of quantum mechanics.

That’s scientist talk for “We don’t understand it either.” So now I don’t feel so bad about having no idea what just happened.

Read all the science details here: Zheng-Hao Liu et al, Exploring the boundary of quantum correlations with a time-domain optical processor, Science Advances (2025). DOI: 10.1126/sciadv.abd8080