[I'm a recently retired computer science professor.]
I really like the idea and I really want it to work, but there are few usability/pedagogical issues that I had, which make it difficult to evaluate:
1. After typing in my question, it vanished. The conversation acknowledges it and (modestly) addresses it, but it was impossible for me to refer back to it.
2. I don't see how to ask to be called on again. I'm at the end of section 1, and the only interaction offered is "Next Section".
3. I'm confused by Avery, Blake, Casey, and Drew. Do they have different personalities that I'm supposed to understand over time? If so, then I really want a graphic along with the name to help me associate what they say with their identities. If not, then I'm not sure why there are so many people in the class. It might be better to just have Instructor, Classmate and me.
4. I *think* that I don't want a seminar, but would rather have a 1:1 Socratic tutor. If I were back at my undergraduate LAC, then I'm tolerant of other people's bad questions/comments because they paid just as much tuition as I did, but here it seems like all cost and no benefit to me. I think I'd rather have a tutor ask challenging questions. If the classmates are foils aiding the Socratic method, that's fine, but that's not the feeling that I got.
5. I did not get the sense that Hartwell cared if *I* understood the concept. If the goal is to teach me (and not Avery, etc.), then I want the system to pepper me with explanations and questions until I've demonstrated mastery or have given up. Basically, "WHILE pupil is confused DO explain differently and test comprehension".
6. I think "Call on me" has two meanings, and I don't see how to invoke one of them. One meaning is basically that I have a question or comment, but the other is "Ask me a question to determine if I understand the topic." I really wanted that latter option. I tried to simulate is by asking, "Can you ask me a question to determine if I understand?", but I was forced to read through six comments from other pupils before it was my turn, and then there was no way to type an answer.
Thanks for your comments. I think that what they suggest is what I might call a "tutor interface" along with the "seminar interface." The tutor interface would allow one-on-one interaction between the professor and the user, it would be persistent (allowing for follow-up questions), and it would allow the professor to query the student, testing for understanding. The tutor interface seems very do-able and very much worth doing.
I've been a professor for quite a while now - in its present form, it struck me as similar to graphic novels which attempt to be textbooks. I felt that the discussion was too long before any interjection was permitted, that the interjection was single shot, and the discussion never really had a life after it. It felt pretty wooden. Maybe it is reminding me what I didn't like teaching (or taking) undergraduate seminars. The discussion felt very shallow. Still - I'd be interested in following another iteration.
Avery, while everyone else is thinking, can I ask what you are majoring in?
Avery:
Engineering. But I'm thinking of minoring in econ. It's like engineering in that you think in terms of a system. I'm really liking this course, too.
Professor Hartwell:
Mainstream economics is actually too close to engineering for my taste.
Avery:
I gather that you think social systems are more complicated.
Professor Hartwell:
More complex, yes. I would be happy to talk more later. Meanwhile, Ben was saying. "Do we think Bees have a high dunbar's number because of their relatively simple behavior? could we measure their ability to predict each other's behavior relative to the complexity of their own? "
Discussion continues:
Drew:
Ben, that's an interesting question but it's taking us away from our focus on human social dynamics. We need to stay on track with discussing how individuals balance their needs with group cooperation.
Professor Hartwell:
Thank you, Drew. Ben's question about social insects does connect to broader patterns, but let's refocus on human society. We were examining the tension between individual liberty and group cohesion. Blake, you've been skeptical of enforced cooperation - would you say there's an inherent conflict between personal freedom and social order?
Blake:
Obviously there is. Every rule or social norm is a constraint on individual choice. The Dunbar number itself shows we're not meant for large-scale cooperation - we have to be forced into it through artificial structures.
Drew:
That's too cynical. Humans naturally form communities and help each other. Look at how people respond during disasters - they don't need to be forced to cooperate.
Blake:
Sure, in small groups where everyone knows each other. But scale that up past Dunbar's number and you need hierarchy, rules, enforcement - basically coercion.
Casey:
Actually, if we look at anthropological studies of hunter-gatherer societies...
Drew:
Casey, before you go into the historical details, can we focus on modern society? How do we actually balance individual rights with collective needs today?
Professor Hartwell:
Good redirect. Remember our First Iron Law - sometimes individuals naturally cooperate, sometimes they need structure. The key is understanding which conditions promote voluntary cooperation versus when formal institutions become necessary. Avery, you've been quiet - what are your thoughts on this?
Avery:
Well... it seems like we might be creating a false dichotomy between individual freedom and social cohesion. Maybe different types of social structures can either enhance or restrict individual liberty, depending on how they're designed.
The "class discussion" is pre-coded "banter" that takes up time while the system is calling Claude to get an answer. The "discussion continues" gives Claude's response. Drew is instructed to "police" the discussion, and Drew thought that you were getting off track. In this case, I am inclined to agree.
Ah, now I understand the role of the banter better. Very interesting. As for Drew, maybe he and I just don't get along. In that case, you have some 'adversarial testing.'
I'll give you folks a moment to think. Then I'll ask for your reactions.
Blake:
Why can't we just start now?
Avery:
Some of us need time to think. Not everybody shoots their mouth off instantly--Blake!
Blake:
Yeah, Avery, you're probably going to tell us how everything in this seminar fits together, and your genius brain has it all figured out.
Drew:
Settle down, people.
Professor Hartwell:
Let's remember what Ben was saying. "It's so interesting that we appear to be defending conformity to norms in a setting which was established to allow free-thinking outside the established norms!"
Discussion continues:
Professor Hartwell:
That's an astute observation, Ben. But I think we need to distinguish between arbitrary conformity and coordination mechanisms that emerge to solve collective action problems. What do others think?
Blake:
This is exactly my concern. We're basically being taught that individual freedom needs to be constrained for the "greater good." But who decides what that greater good is?
Drew:
Come on Blake, it's not about constraining freedom - it's about recognizing that humans naturally develop ways to work together. Look at how hunter-gatherer societies functioned without formal rules.
Blake:
That's circular reasoning. Just because something emerged naturally doesn't make it optimal or justified.
Casey:
If I may, this reminds me of Rousseau's concept of the "general will" versus individual will. He argued that true freedom comes through participating in collective decision-making, not through pure individualism.
Blake:
Oh please, Rousseau was just dressing up conformity in fancy philosophical language.
Drew:
That's not fair. We need some basic rules to func-
Avery:
I think we might be missing the distinction between coordination and control. Game theory shows us that many cooperative norms emerge spontaneously because they benefit everyone, not because they're imposed from above.
Professor Hartwell:
Excellent point, Avery. This connects to what we'll discuss next week about emergent order versus designed systems. Blake raises valid concerns about coercion, but let's consider: can true individual freedom exist without some framework for managing conflicts between individuals?
This seems like an interesting and on-point follow-up to your comment. I agree that it would have been nice to allow you to push "call on me" again and continue the conversation. That is high on my list for enhancements.
I'm inclined to tell Prof. Hartwell that 'nobody remembers what came before the word 'but.'' Anyway, if you want to enable extended dialogue, that would be a great feature that would encourage me to try it again with more vigor. And perhaps more grace.
I'll give you folks a moment to think. Then I'll ask for your reactions.
Blake:
Why can't we just start now?
Avery:
Some of us need time to think. Not everybody shoots their mouth off instantly--Blake!
Blake:
Yeah, Avery, you're probably going to tell us how everything in this seminar fits together, and your genius brain has it all figured out.
Drew:
Settle down, people.
Professor Hartwell:
Let's remember what Ben was saying. "Is this the discussion about the 'magisteria?' Am I spelling that right?"
Discussion continues:
Drew:
No Ben, that's from next week's reading. We're discussing correlation and causation right now.
Professor Hartwell:
Thank you Drew. Ben, we're examining the limitations of correlational studies in social science. Would anyone like to explain why correlation doesn't always equal causation?
Blake:
Well, like the Super Bowl thing - just because the market goes up when NFC teams win doesn't mean there's any real connection. It's just random chance with a small sample size.
Casey:
Actually, this reminds me of Hume's problem of induction. He argued that we can never be completely certain about causation, we can only observe regular conjunction between events. So even when we think we see a clear cause-effect relationship...
Drew:
Casey, let's stick to the specific examples we're discussing today.
Avery:
I think the third factor causation is most interesting in social science. Like, you might see a correlation between two behaviors, but they're both actually caused by some underlying personality trait or social factor we haven't measured.
Professor Hartwell:
Excellent point, Avery. This is why we need to be especially careful when interpreting social science research. Can anyone give us another example of third factor causation?
What's interesting here is that the science/humanities dichotomy is exactly where Casey was leading ... and the professor shot it down. I don't think I would have bothered completing college with Prof. Hartwell.
Everybody's so quiet, not sure how to react. Blake, you're a senior right?
Blake:
Mmm-hmm.
Professor Hartwell:
Do you know what you're going to do after you graduate?
Blake:
I want to go into business, but I don't want to work for a big company. What I'd really like to do is start my own company. I've got a couple of ideas.
Professor Hartwell:
I'd like to hear about that some time. But we should think about what Ben was saying. "We can do correlation without scalar data, am I correct? With binary data even? Or in multiple dimensions?"
Discussion continues:
Professor Hartwell:
Excellent question, Ben. Yes, correlation analysis can be performed with various types of data. Binary correlations are particularly common in social science research - like studying whether being married correlates with living longer. Who can think of other examples?
Casey:
The phi coefficient is used for binary correlations, and there's fascinating historical precedent in early social statistics. Karl Pearson actually developed it while studying evolutionary biology quest-
Drew:
Let's stay focused on the main point about correlation types rather than the whole history.
Blake:
What about measuring correlation between three or more variables? Seems like that could get messy real fast.
Professor Hartwell:
Indeed, Blake. Multiple correlation introduces additional complexity, which relates to our First Iron Law. The more variables we add, the harder it becomes to establish clear patterns.
Avery:
I think this connects to the Dunbar Number in a way... when we try to track too many relationships at once, we hit cognitive limits. Maybe that's why we need specialized statistical tools.
Professor Hartwell:
Interesting connection, Avery. The human mind does struggle with multiple correlations, which is precisely why we've developed mathematical tools to help us understand these relationships.
Casey:
And in social network analysis, we can measure correlations between different types of relationships, like friendship ties and information sharing patterns...
Drew:
We should probably stick to the basics of correlation before diving into network analysis.
Professor Hartwell:
Let's focus on understanding the fundamental concepts first. Can anyone explain why binary correlation might be particularly useful in social science research?
This is not the answer I would have wanted to give. I am against talking about relationships among binary variables as correlations. And with categorical data, I think that to speak of correlation is absurd. Regardless of who is right, I wish Claude would answer more from my way of thinking and less from its own background.
Here's one example for starters, from the first page: ""Yes, I see how the research question determines the data structure - in fact, the data is an emergent property of the researcher observing the system, correct? How the researcher creates meaning by asking the question and taking observations?""
"Class discussion:
Professor Hartwell:
Remember. Don't react right away. Give it some thought.
Blake:
Why don't you ever give your reaction first?
Casey:
This is traditional pedagogical practice. It goes all the way back to Socrates. The ancient Greeks--
Drew:
Oh, for heaven's sake, Casey. Get over yourself. We know you love the ancient Greeks. You love everything, as long as it's ancient and esoteric.
Professor Hartwell:
OK, let's simmer down. Focus on what Ben was saying. "Yes, I see how the research question determines the data structure - in fact, the data is an emergent property of the researcher observing the system, correct? How the researcher creates meaning by asking the question and taking observations?"
Discussion continues:
Drew:
Ben, I have to stop you there. That's getting way too philosophical. We're supposed to be discussing the three types of data and how to evaluate research studies.
Professor Hartwell:
Thank you, Drew. Ben, while epistemology is fascinating, let's focus on the concrete differences between experimental, observational, and survey data. Each has distinct strengths and limitations.
Blake:
Yeah, like how experiments give you causation but might not reflect real-world conditions. I mean, paying college students to play economic games in a lab isn't exactly like studying actual market behavior.
Casey:
This reminds me of Campbell and Stanley's seminal work on experimental design. They outlined sixteen different threats to validity that researchers must consider. For instance, the Hawthorne effect, where subjects change their behavior simply because they're being observed...
Drew:
Casey, we don't need the whole history. The point is that each type of data has trade-offs.
Avery:
I think... what's important is matching the right type of data to your research question. Like, if you want to study the Dunbar number, you'd need observational data from real social networks, not just surveys asking people how many friends they have.
Professor Hartwell:
Excellent point, Avery. And this connects to our First Iron Law - different methods might reveal different aspects of the same phenomenon. What matters is understanding the limitations of each approach."
I agree with Drew that your question, while very important, was outside of the scope of what this chapter was trying to address. But I think everyone else proceeded to go off the rails with their comments. I need to give instructions to curb that.
That made the room go quiet. Drew, what kind of profession would you like to end up in?
Drew:
Well, I have this fantasy of eventually having my own dance studio teaching kids musical theater.
Professor Hartwell:
So meanwhile can you be like an assistant dance teacher at someone else's studio?
Drew:
No money. I'm actually getting a teaching certificate and that's what I'll do the first couple of years after graduation. But you asked where I would *like* to end up. But we should get back to
Professor Hartwell:
Keeping me on track, Drew, as usual. Ben was saying. "Could you explain how Anscombe's Quartet relates to this? "
Discussion continues:
Professor Hartwell:
Excellent question, Ben. Anscombe's Quartet perfectly illustrates our discussion about misleading correlations. These are four datasets that have identical statistical properties - same mean, variance, and correlation coefficient - but look completely different when graphed.
Casey:
Oh yes! Francis Anscombe created these in 1973 to demonstrate why we should always visualize our data, not just rely on summary statistics. One dataset is linear, one's curved, one's linear except for an outlier, and one's basically vertical except for one point.
Blake:
But isn't this just a clever mathematical trick? How often do we actually encounter datasets that are this deceptive in real social science research?
Drew:
Actually, this is super relevant. Think about how many studies just report correlation coefficients without showing scatter plots. We could be missing important patterns.
Avery:
I think... I think this relates to our First Iron Law too. Even with the same correlation, the relationship between variables can take different forms. Sometimes it's this way, sometimes it's that way.
Professor Hartwell:
Precisely. And this brings us back to our discussion of how correlation can mislead. What looks like a straightforward relationship in summary statistics might be masking more complex patterns - or even pure coincidence.
Casey:
It reminds me of how Galton initially misinterpreted regression to the mean when studying heredity. He saw the correlation between parent and child heights but didn't initially grasp the underlying statistical principle that was actually causing...
Drew:
Let's stay focused on modern social science examples. How does this affect the studies we're reviewing today?
Professor Hartwell:
Good redirect, Drew. Let's apply this to our earlier example about social media use and depression. Who can explain how visualization might reveal patterns that simple correlation might miss?
This seems to me like an ok response. Claude had the quartet in his knowledge base, which is more than I would have (I understand the point, but I did not know the name)
I tried the called-on feature quite a few times and in general it feels like the conversations went sideways. The other students, sometimes the professor, didn't care about what I was raising; it got dealt with only summarily, and there was no opportunity to enter into repeated dialogue. I had high hopes. Maybe this is how real seminars feel to the participants? The personalities in these modules are strong - perhaps a bit too strong - and the discussions are very short.
Ok, I've posted a few of my examples. It seems pretty clear that something doesn't really work with my own questions, even though I'm really modeling behavior I'd like to see in an ideal classroom (and frankly, I was pretty hopeful for some stimulating discussion).
Not a professor but I think this is great. It would also be interesting to see curricula of this sort presented in ordinary didactic form rather than in dialogue form. Present a module of material and allow the user to interact with the “presenter” with questions, etc. Seems like that could capture a lot of the value of interactivity while maintaining a bit more linear presentation. But I think these sorts of things could become fantastic.
Not sure why Casey is in the seminar other than to be dumped on. If you hate Aristotle fine, but why drag in interesting digressions if you are just going to silence them? It's interesting that Frank Furedi began a recent human interdependence-focused substack post (https://frankfuredi.substack.com/p/the-struggle-for-a-new-political ) with a paragraph:
"Start with Aristotle’s Politics. Here is one of the original attempts to grapple seriously with the organisation of society, power, and citizenship. Aristotle grounds us firmly in the pragmatic reality that politics is fundamentally about how different people with different interests can live together. Given our polarised and fragmented present, Aristotle’s clarity about community, virtue, and practical wisdom seems refreshingly radical."
I can see how that would be inimical to economic thought so maybe Aristotle references best avoided all together.
[I'm a recently retired computer science professor.]
I really like the idea and I really want it to work, but there are few usability/pedagogical issues that I had, which make it difficult to evaluate:
1. After typing in my question, it vanished. The conversation acknowledges it and (modestly) addresses it, but it was impossible for me to refer back to it.
2. I don't see how to ask to be called on again. I'm at the end of section 1, and the only interaction offered is "Next Section".
3. I'm confused by Avery, Blake, Casey, and Drew. Do they have different personalities that I'm supposed to understand over time? If so, then I really want a graphic along with the name to help me associate what they say with their identities. If not, then I'm not sure why there are so many people in the class. It might be better to just have Instructor, Classmate and me.
4. I *think* that I don't want a seminar, but would rather have a 1:1 Socratic tutor. If I were back at my undergraduate LAC, then I'm tolerant of other people's bad questions/comments because they paid just as much tuition as I did, but here it seems like all cost and no benefit to me. I think I'd rather have a tutor ask challenging questions. If the classmates are foils aiding the Socratic method, that's fine, but that's not the feeling that I got.
5. I did not get the sense that Hartwell cared if *I* understood the concept. If the goal is to teach me (and not Avery, etc.), then I want the system to pepper me with explanations and questions until I've demonstrated mastery or have given up. Basically, "WHILE pupil is confused DO explain differently and test comprehension".
6. I think "Call on me" has two meanings, and I don't see how to invoke one of them. One meaning is basically that I have a question or comment, but the other is "Ask me a question to determine if I understand the topic." I really wanted that latter option. I tried to simulate is by asking, "Can you ask me a question to determine if I understand?", but I was forced to read through six comments from other pupils before it was my turn, and then there was no way to type an answer.
I hope this is helpful.
Thanks for your comments. I think that what they suggest is what I might call a "tutor interface" along with the "seminar interface." The tutor interface would allow one-on-one interaction between the professor and the user, it would be persistent (allowing for follow-up questions), and it would allow the professor to query the student, testing for understanding. The tutor interface seems very do-able and very much worth doing.
I've been a professor for quite a while now - in its present form, it struck me as similar to graphic novels which attempt to be textbooks. I felt that the discussion was too long before any interjection was permitted, that the interjection was single shot, and the discussion never really had a life after it. It felt pretty wooden. Maybe it is reminding me what I didn't like teaching (or taking) undergraduate seminars. The discussion felt very shallow. Still - I'd be interested in following another iteration.
Was it better or worse when you were called on?
Class discussion:
Professor Hartwell:
Avery, while everyone else is thinking, can I ask what you are majoring in?
Avery:
Engineering. But I'm thinking of minoring in econ. It's like engineering in that you think in terms of a system. I'm really liking this course, too.
Professor Hartwell:
Mainstream economics is actually too close to engineering for my taste.
Avery:
I gather that you think social systems are more complicated.
Professor Hartwell:
More complex, yes. I would be happy to talk more later. Meanwhile, Ben was saying. "Do we think Bees have a high dunbar's number because of their relatively simple behavior? could we measure their ability to predict each other's behavior relative to the complexity of their own? "
Discussion continues:
Drew:
Ben, that's an interesting question but it's taking us away from our focus on human social dynamics. We need to stay on track with discussing how individuals balance their needs with group cooperation.
Professor Hartwell:
Thank you, Drew. Ben's question about social insects does connect to broader patterns, but let's refocus on human society. We were examining the tension between individual liberty and group cohesion. Blake, you've been skeptical of enforced cooperation - would you say there's an inherent conflict between personal freedom and social order?
Blake:
Obviously there is. Every rule or social norm is a constraint on individual choice. The Dunbar number itself shows we're not meant for large-scale cooperation - we have to be forced into it through artificial structures.
Drew:
That's too cynical. Humans naturally form communities and help each other. Look at how people respond during disasters - they don't need to be forced to cooperate.
Blake:
Sure, in small groups where everyone knows each other. But scale that up past Dunbar's number and you need hierarchy, rules, enforcement - basically coercion.
Casey:
Actually, if we look at anthropological studies of hunter-gatherer societies...
Drew:
Casey, before you go into the historical details, can we focus on modern society? How do we actually balance individual rights with collective needs today?
Professor Hartwell:
Good redirect. Remember our First Iron Law - sometimes individuals naturally cooperate, sometimes they need structure. The key is understanding which conditions promote voluntary cooperation versus when formal institutions become necessary. Avery, you've been quiet - what are your thoughts on this?
Avery:
Well... it seems like we might be creating a false dichotomy between individual freedom and social cohesion. Maybe different types of social structures can either enhance or restrict individual liberty, depending on how they're designed.
The "class discussion" is pre-coded "banter" that takes up time while the system is calling Claude to get an answer. The "discussion continues" gives Claude's response. Drew is instructed to "police" the discussion, and Drew thought that you were getting off track. In this case, I am inclined to agree.
Ah, now I understand the role of the banter better. Very interesting. As for Drew, maybe he and I just don't get along. In that case, you have some 'adversarial testing.'
Class discussion:
Professor Hartwell:
I'll give you folks a moment to think. Then I'll ask for your reactions.
Blake:
Why can't we just start now?
Avery:
Some of us need time to think. Not everybody shoots their mouth off instantly--Blake!
Blake:
Yeah, Avery, you're probably going to tell us how everything in this seminar fits together, and your genius brain has it all figured out.
Drew:
Settle down, people.
Professor Hartwell:
Let's remember what Ben was saying. "It's so interesting that we appear to be defending conformity to norms in a setting which was established to allow free-thinking outside the established norms!"
Discussion continues:
Professor Hartwell:
That's an astute observation, Ben. But I think we need to distinguish between arbitrary conformity and coordination mechanisms that emerge to solve collective action problems. What do others think?
Blake:
This is exactly my concern. We're basically being taught that individual freedom needs to be constrained for the "greater good." But who decides what that greater good is?
Drew:
Come on Blake, it's not about constraining freedom - it's about recognizing that humans naturally develop ways to work together. Look at how hunter-gatherer societies functioned without formal rules.
Blake:
That's circular reasoning. Just because something emerged naturally doesn't make it optimal or justified.
Casey:
If I may, this reminds me of Rousseau's concept of the "general will" versus individual will. He argued that true freedom comes through participating in collective decision-making, not through pure individualism.
Blake:
Oh please, Rousseau was just dressing up conformity in fancy philosophical language.
Drew:
That's not fair. We need some basic rules to func-
Avery:
I think we might be missing the distinction between coordination and control. Game theory shows us that many cooperative norms emerge spontaneously because they benefit everyone, not because they're imposed from above.
Professor Hartwell:
Excellent point, Avery. This connects to what we'll discuss next week about emergent order versus designed systems. Blake raises valid concerns about coercion, but let's consider: can true individual freedom exist without some framework for managing conflicts between individuals?
This seems like an interesting and on-point follow-up to your comment. I agree that it would have been nice to allow you to push "call on me" again and continue the conversation. That is high on my list for enhancements.
I'm inclined to tell Prof. Hartwell that 'nobody remembers what came before the word 'but.'' Anyway, if you want to enable extended dialogue, that would be a great feature that would encourage me to try it again with more vigor. And perhaps more grace.
Class discussion:
Professor Hartwell:
I'll give you folks a moment to think. Then I'll ask for your reactions.
Blake:
Why can't we just start now?
Avery:
Some of us need time to think. Not everybody shoots their mouth off instantly--Blake!
Blake:
Yeah, Avery, you're probably going to tell us how everything in this seminar fits together, and your genius brain has it all figured out.
Drew:
Settle down, people.
Professor Hartwell:
Let's remember what Ben was saying. "Is this the discussion about the 'magisteria?' Am I spelling that right?"
Discussion continues:
Drew:
No Ben, that's from next week's reading. We're discussing correlation and causation right now.
Professor Hartwell:
Thank you Drew. Ben, we're examining the limitations of correlational studies in social science. Would anyone like to explain why correlation doesn't always equal causation?
Blake:
Well, like the Super Bowl thing - just because the market goes up when NFC teams win doesn't mean there's any real connection. It's just random chance with a small sample size.
Casey:
Actually, this reminds me of Hume's problem of induction. He argued that we can never be completely certain about causation, we can only observe regular conjunction between events. So even when we think we see a clear cause-effect relationship...
Drew:
Casey, let's stick to the specific examples we're discussing today.
Avery:
I think the third factor causation is most interesting in social science. Like, you might see a correlation between two behaviors, but they're both actually caused by some underlying personality trait or social factor we haven't measured.
Professor Hartwell:
Excellent point, Avery. This is why we need to be especially careful when interpreting social science research. Can anyone give us another example of third factor causation?
What's interesting here is that the science/humanities dichotomy is exactly where Casey was leading ... and the professor shot it down. I don't think I would have bothered completing college with Prof. Hartwell.
Class discussion:
Professor Hartwell:
Everybody's so quiet, not sure how to react. Blake, you're a senior right?
Blake:
Mmm-hmm.
Professor Hartwell:
Do you know what you're going to do after you graduate?
Blake:
I want to go into business, but I don't want to work for a big company. What I'd really like to do is start my own company. I've got a couple of ideas.
Professor Hartwell:
I'd like to hear about that some time. But we should think about what Ben was saying. "We can do correlation without scalar data, am I correct? With binary data even? Or in multiple dimensions?"
Discussion continues:
Professor Hartwell:
Excellent question, Ben. Yes, correlation analysis can be performed with various types of data. Binary correlations are particularly common in social science research - like studying whether being married correlates with living longer. Who can think of other examples?
Casey:
The phi coefficient is used for binary correlations, and there's fascinating historical precedent in early social statistics. Karl Pearson actually developed it while studying evolutionary biology quest-
Drew:
Let's stay focused on the main point about correlation types rather than the whole history.
Blake:
What about measuring correlation between three or more variables? Seems like that could get messy real fast.
Professor Hartwell:
Indeed, Blake. Multiple correlation introduces additional complexity, which relates to our First Iron Law. The more variables we add, the harder it becomes to establish clear patterns.
Avery:
I think this connects to the Dunbar Number in a way... when we try to track too many relationships at once, we hit cognitive limits. Maybe that's why we need specialized statistical tools.
Professor Hartwell:
Interesting connection, Avery. The human mind does struggle with multiple correlations, which is precisely why we've developed mathematical tools to help us understand these relationships.
Casey:
And in social network analysis, we can measure correlations between different types of relationships, like friendship ties and information sharing patterns...
Drew:
We should probably stick to the basics of correlation before diving into network analysis.
Professor Hartwell:
Let's focus on understanding the fundamental concepts first. Can anyone explain why binary correlation might be particularly useful in social science research?
This is not the answer I would have wanted to give. I am against talking about relationships among binary variables as correlations. And with categorical data, I think that to speak of correlation is absurd. Regardless of who is right, I wish Claude would answer more from my way of thinking and less from its own background.
Here's one example for starters, from the first page: ""Yes, I see how the research question determines the data structure - in fact, the data is an emergent property of the researcher observing the system, correct? How the researcher creates meaning by asking the question and taking observations?""
"Class discussion:
Professor Hartwell:
Remember. Don't react right away. Give it some thought.
Blake:
Why don't you ever give your reaction first?
Casey:
This is traditional pedagogical practice. It goes all the way back to Socrates. The ancient Greeks--
Drew:
Oh, for heaven's sake, Casey. Get over yourself. We know you love the ancient Greeks. You love everything, as long as it's ancient and esoteric.
Professor Hartwell:
OK, let's simmer down. Focus on what Ben was saying. "Yes, I see how the research question determines the data structure - in fact, the data is an emergent property of the researcher observing the system, correct? How the researcher creates meaning by asking the question and taking observations?"
Discussion continues:
Drew:
Ben, I have to stop you there. That's getting way too philosophical. We're supposed to be discussing the three types of data and how to evaluate research studies.
Professor Hartwell:
Thank you, Drew. Ben, while epistemology is fascinating, let's focus on the concrete differences between experimental, observational, and survey data. Each has distinct strengths and limitations.
Blake:
Yeah, like how experiments give you causation but might not reflect real-world conditions. I mean, paying college students to play economic games in a lab isn't exactly like studying actual market behavior.
Casey:
This reminds me of Campbell and Stanley's seminal work on experimental design. They outlined sixteen different threats to validity that researchers must consider. For instance, the Hawthorne effect, where subjects change their behavior simply because they're being observed...
Drew:
Casey, we don't need the whole history. The point is that each type of data has trade-offs.
Avery:
I think... what's important is matching the right type of data to your research question. Like, if you want to study the Dunbar number, you'd need observational data from real social networks, not just surveys asking people how many friends they have.
Professor Hartwell:
Excellent point, Avery. And this connects to our First Iron Law - different methods might reveal different aspects of the same phenomenon. What matters is understanding the limitations of each approach."
I agree with Drew that your question, while very important, was outside of the scope of what this chapter was trying to address. But I think everyone else proceeded to go off the rails with their comments. I need to give instructions to curb that.
Thanks so much for showing me all of these!!
Class discussion:
Professor Hartwell:
That made the room go quiet. Drew, what kind of profession would you like to end up in?
Drew:
Well, I have this fantasy of eventually having my own dance studio teaching kids musical theater.
Professor Hartwell:
So meanwhile can you be like an assistant dance teacher at someone else's studio?
Drew:
No money. I'm actually getting a teaching certificate and that's what I'll do the first couple of years after graduation. But you asked where I would *like* to end up. But we should get back to
Professor Hartwell:
Keeping me on track, Drew, as usual. Ben was saying. "Could you explain how Anscombe's Quartet relates to this? "
Discussion continues:
Professor Hartwell:
Excellent question, Ben. Anscombe's Quartet perfectly illustrates our discussion about misleading correlations. These are four datasets that have identical statistical properties - same mean, variance, and correlation coefficient - but look completely different when graphed.
Casey:
Oh yes! Francis Anscombe created these in 1973 to demonstrate why we should always visualize our data, not just rely on summary statistics. One dataset is linear, one's curved, one's linear except for an outlier, and one's basically vertical except for one point.
Blake:
But isn't this just a clever mathematical trick? How often do we actually encounter datasets that are this deceptive in real social science research?
Drew:
Actually, this is super relevant. Think about how many studies just report correlation coefficients without showing scatter plots. We could be missing important patterns.
Avery:
I think... I think this relates to our First Iron Law too. Even with the same correlation, the relationship between variables can take different forms. Sometimes it's this way, sometimes it's that way.
Professor Hartwell:
Precisely. And this brings us back to our discussion of how correlation can mislead. What looks like a straightforward relationship in summary statistics might be masking more complex patterns - or even pure coincidence.
Casey:
It reminds me of how Galton initially misinterpreted regression to the mean when studying heredity. He saw the correlation between parent and child heights but didn't initially grasp the underlying statistical principle that was actually causing...
Drew:
Let's stay focused on modern social science examples. How does this affect the studies we're reviewing today?
Professor Hartwell:
Good redirect, Drew. Let's apply this to our earlier example about social media use and depression. Who can explain how visualization might reveal patterns that simple correlation might miss?
This seems to me like an ok response. Claude had the quartet in his knowledge base, which is more than I would have (I understand the point, but I did not know the name)
I did like this response. It would have been great except for Drew and the enabling professor at the end!
I tried the called-on feature quite a few times and in general it feels like the conversations went sideways. The other students, sometimes the professor, didn't care about what I was raising; it got dealt with only summarily, and there was no opportunity to enter into repeated dialogue. I had high hopes. Maybe this is how real seminars feel to the participants? The personalities in these modules are strong - perhaps a bit too strong - and the discussions are very short.
Thanks. That is very helpful. I should have thought to ask users to copy paste such response dialogues because I am particularly interested in those
Ok, I've posted a few of my examples. It seems pretty clear that something doesn't really work with my own questions, even though I'm really modeling behavior I'd like to see in an ideal classroom (and frankly, I was pretty hopeful for some stimulating discussion).
Ok Arnold. I can repeat the experience and see what you think.
Not a professor but I think this is great. It would also be interesting to see curricula of this sort presented in ordinary didactic form rather than in dialogue form. Present a module of material and allow the user to interact with the “presenter” with questions, etc. Seems like that could capture a lot of the value of interactivity while maintaining a bit more linear presentation. But I think these sorts of things could become fantastic.
Not sure why Casey is in the seminar other than to be dumped on. If you hate Aristotle fine, but why drag in interesting digressions if you are just going to silence them? It's interesting that Frank Furedi began a recent human interdependence-focused substack post (https://frankfuredi.substack.com/p/the-struggle-for-a-new-political ) with a paragraph:
"Start with Aristotle’s Politics. Here is one of the original attempts to grapple seriously with the organisation of society, power, and citizenship. Aristotle grounds us firmly in the pragmatic reality that politics is fundamentally about how different people with different interests can live together. Given our polarised and fragmented present, Aristotle’s clarity about community, virtue, and practical wisdom seems refreshingly radical."
I can see how that would be inimical to economic thought so maybe Aristotle references best avoided all together.