Progress in form-based coding

Вот отформатированный транскрипт вебинара: Robert Studville: On the Park Bench presents interactive conversations with thought leaders in urbanism and allied fields related to the built environment. Today we have "Progress in Form-Based Coding" with Emily Talon and Arianna Salazar Miranda. My name is Robert Studville, and I'll be moderating the session. So, share your thoughts on hashtag #ontheparkbench, www.tinyurl.com/OTPBfeedback. And join us for the upcoming webinar, "The Art of the New Urbanism," which is going to be at the same time next week, May 13th. And this is going to be a fantastic show all about some of the great art that's been created in service of the new urbanism. So, go to cnu.org/resources/ontheparkbench to find out more and register. And of course, register for CNU 33. CNU 33 will be in Providence, Rhode Island, on June 11th through 14th. We have our educational offerings on the website now, and the agenda has eight tracks: climate, development, housing policy/regulation, public engagement, regional context, transportation, and urban design. So, there's something for everybody. The City of Providence is really fantastic. It's one of the great small to mid-sized cities in America. And if you haven't been there, you really need to go. Go to cnu.org/CNU33. And today, we're going to talk about a nationwide study that was made on form-based coding, making use of AI as a research tool, and what this study says about how zoning has changed in America since the New Urbanism movement was launched. Emily Talon is Professor of Urbanism at the University of Chicago, where she teaches urban design and directs the Urbanism Lab. She is a Fellow of the American Institute of Certified Planners and the recipient of a Guggenheim Fellowship. She is also a Senior Fellow and former CNU board member. Talon's books include "New Urbanism in American Planning," "Designed for Diversity," "Urban Design Reclaimed," "City Rules," and "Neighborhood." Her most recent book is "What Cities Say," published by Oxford Press in 2024. Arianna Salazar Miranda is an Assistant Professor of Urban Planning and Data Science at the Yale School of the Environment. Her research focuses on the relationship between urban planning, the built environment, human behavior, and sustainability. She recently launched her Livable Cities Lab at Yale. I'm Rob Studaville, editor of CNU's Public Square. First, Emily and Arianna are going to present, and then we're going to have a discussion among the three of us, and then we're going to go to Q&A from the audience. There's going to be plenty of time for Q&A, so please ask your questions as they occur to you using the Q&A tool of Zoom. And we'll get to those more or less in the order that they are asked. So, welcome, Emily and Arianna, to On the Park Bench. And I have to stop sharing. Emily Talon: Thank you so much for having us. So, let me go ahead and share my slides. Wonderful. Okay, thanks so much, Rob. I'm going to start things off here with some intro, and then Ari is going to take over. And I'll circle back at the end with some concluding thoughts. So, you know, everyone knows that traditional zoning codes are bad. They segregate land uses, they increase car dependence, they're creating more urban sprawl, social disconnection, bad for the economy, bad for the environment. We know all this. And in the last couple of decades, really, there has been widespread agreement about the disaster that conventional zoning codes are. That if we're going to have regulations governing how we build cities, then we need a much different approach. A regulatory approach that does not segregate land uses, that mixes uses and housing types, and that allows compact development. This is well known to probably everyone on this call right now. So, planners, and especially New Urbanists, have known about this problem for decades, as I said. And one solution – and one solution, a fair alternative – is pretty obvious to a lot of us, and it's fallen under the general heading of form-based coding or sometimes form-based zoning, which in its simplest terms is about employing rules that support pedestrian-based cities rather than car-based cities. So, that's all the things I just mentioned: mixing uses, promoting urban form that supports walkability. And you should know, and maybe many people on this call do know, that modern form-based codes really started to gain steam in the 1990s. However, I do want to say that regulations directed at controlling form have been around for centuries. And so the researchers among us started to investigate: where are these form-based codes located, and what might their impact be? What can we study about them? And the slide that you're looking at right there is one of the very earliest form-based codes in the modern era, and this is, of course, the urban code for the town of Seaside by DPZ, and they were really trailblazers in this whole effort. So, some of the earliest research, I think next slide. Some of the earliest research was conducted by Hazel Borys and Placemakers, who launched the Code Study, which was the first robust accounting of form-based code adoption in the US and globally. And these codes – that's tracking these codes in this Code Study – really relied on crowdsourced material as a starting point. And then there was a whole process of applying a set of criteria established by the Form-Based Codes Institute, which was created in 2004. And it's now called the Center for Zoning Solutions, which I only recently learned. And the approach was to determine whether a code, use the Form-Based Codes Institute criteria to determine whether a code was appropriate for inclusion. And as of a few years ago, there were 728 codes that met the criteria. So, it was a painstaking process tracking these codes. I got involved by requiring students in my classes to track these codes as sort of free student labor to kind of add to the database. I think we can go to the next slide. There we go. That's just a screenshot of this herculean effort by Hazel Borys and Matt Lambert was involved too in really trying to track these codes using this crowdsourced approach, but then having to manually go in and do some searching, some Google searching on the codes, where they were at. And it's a lot of effort. And it was a great start to what we need to do to track form-based codes, but pretty labor intensive. There have been, so what has the other research been? There have been some qualitative studies done about form-based codes: do form-based codes help place quality? Do they incorporate sustainability? Scale quantitative evaluation has been pretty rare. The problem is that, and one of the problems with tracking all this, is that the text of zoning codes is unstructured. It's very variable. So, it makes it difficult to assess codes quantitatively and in a way that you can compare across regions. And a major challenge has been in trying to identify form-based codes in the first place. Municipalities might adopt certain form-based code elements such as mixed-use and maybe having shorter setbacks. Maybe they've been relaxing their parking requirements. But without necessarily labeling it as, "Hey, here is our form-based code." So, not knowing that is kind of missing some of the nuance of these changes that have been taking place in zoning reform. So, I think there's definitely a need to employ other approaches to tracking form-based coding. Enter AI. And as a subset of AI about interpreting human language, natural language processing, NLP, as it's called. And that offers an enormous and enormous potential to analyze zoning documents. So, that's my introduction. Ari, I'll turn it over to you. Arianna Salazar Miranda: Thank you. All right. So, let's get into the details of how we actually did this. So, as Emily was saying, we're interested in how form-based codes can shape cities to be more sustainable. And our anvil here is going to be to use natural language processing or NLP, which essentially allows us to analyze zoning codes systematically across thousands of documents. Now, NLP is also very valuable in this specific context because, as Emily mentioned, and you all probably already know, zoning codes rarely label themselves as either form-based or use-based, right? Most of them fall somewhere in between. And so, by using NLP, we can detect these sort of subtle differences in language to measure how closely each city aligns with form-based code language, even if the city itself doesn't really explicitly say so. So, let me briefly explain how we collected our data. We gathered zoning documents from Municode. Municode is one of the largest online sources of municipal codes. And our first task was to figure out a way to keep only the chapters that relate to zoning. Municipal documents include many more topics than just zoning. They cover things like garbage collection, traffic rules, and many other regulations about the built environment. And so, the way that we did this was selecting chapters whose titles contain zoning-related keywords, things like zoning, land use, and the other keywords that I'm highlighting here on the slide. So then, once we have that subset selected, we then matched each of these zoning documents to their corresponding US Census place. And that created our final data set, which is around 2,723 municipalities across the entire US. And you can see here on this map their geographic distribution, which we're showing in blue. And then in red, you also see what we're calling like our ground truth documents. So, these are a separate set of 70 zoning codes that are known to be form-based. And these are the ones that we sourced from placemakers.com and from the Form-Based Codes Institute. Now, the reason why we need these is because these ground truth examples are going to give us a baseline for what a typical form-based code looks like. And then we can measure how closely the language of any zoning document matches the language of those form-based codes. So, how do we do that? I'm going to walk you through the process, which I think is very interesting. So, here on the left, you see an example of a zoning document. So, this is our raw input. Now, because these documents are typically very long and our NLP model has a limit on how much it can process all at once, we have to split each document into smaller segments or what we call chunks. Then each chunk is passed through this NLP model, creating a numeric representation called an embedding. You can think of an embedding as sort of a numeric summary of the content of the text. And then we average the embeddings from all the chunks in the same document, which gives us essentially a single measure per document. Now, the output of this process is shown in the top right diagram. So, each gray dot is the document embedding. And we follow the same exact process for the ground truth form-based codes, creating an embedding for them. In the diagram, we're showing these as red dots. Now, recall that our goal is to measure how similar each zoning document, so each gray dot, is to a typical form-based code. Now, to do that, of course, we need to first define what a typical form-based code actually means, right? And so, the way that we define this is as the average of all of the red dots, so all of our ground truth true form-based codes, which I'm showing with an X. This is the centroid, and it typically represents sort of the average or typical language that is found across all the true form-based codes that we have. Now, once we have this baseline, then we can measure similarity to it. And we use a cosine similarity, which essentially calculates the angle between two lines: one from the origin to our true form-based code centroid, and another from the origin to each zoning document or gray dot. So, for example, point A has a higher similarity because it has a smaller angle. And you can see that it's closer in distance to the centroid. And then in contrast, you have, for example, point B, which has a lower similarity because it has a much larger angle, right? You can see that point B is located a lot further from the centroid of the FBC zoning documents. So, we do this for every zoning document in our sample, and this gives us a continuous measure of what we call FBC or form-based code similarity. Now, once we have our FPC similarity measure, the first thing that we did was to check whether this measure actually captures meaningful differences in zoning language. And you can see here, these are word clouds that are showing the words that are most distinctive in each of the three groups. So, let me start with the word cloud on the left. These are zoning documents that ranked high according to the FPC similarity score. And you can see that these typically emphasize design-related words, right? You can see "landscape," you can see "architectural." Now, if we move to the middle word cloud, these are our reference group or our ground truth. These are the true form-based codes. And you can see that there is actually quite a close alignment with the left word cloud, right? You also see that it emphasizes sort of similar design-oriented language. Now on the right, you see codes with low form-based code similarity scores. And in contrast, you can see that these tend to emphasize sort of procedural, more regulatory terms. So, there's things like "performance" or "administrative." Here, the focus is more on rules rather than design. And the reason why we think this exercise was a fruitful one is because it's reassuring that it's, you know, it suggests that our measure is actually pretty good at capturing the relevant thematic differences that we would expect. So, now we can use our measure to learn about FPC adoption. And so, the first takeaway of our work is that FPC language is widespread across the US. And so, to give you a sense of this, we set three thresholds along our FPC similarity measure, which, as you can see in the figure, ranges from zero to six. If we stand at a cutoff of 3.5, this is going to be sort of a lower threshold indicating some similarity to FBCs. About 89% of municipalities meet or exceed this threshold. Now, if we stand at four, this might indicate like a more stronger similarity. We see that around 72% of municipalities adopt FPC language. And at a threshold of 4.5, which indicates sort of a stronger similarity, we see that roughly one-third of municipalities have zoning codes that are closely aligned with form-based code language. The second takeaway is that adoption of FPCs isn't limited to specific regions of the US. So, here you can see the distribution of our FPC similarity broken down across four regions. And you can see that every region includes municipalities that score both high and low similarity. So, there's a lot of variation everywhere. Now, one thing to point here is that the South, which we're showing in red, tends to have a slightly higher score overall. And you can see this by the shifting of the red distribution slightly towards the right. So, the takeaway here is that FBCs appear across the entire country and not just in particular areas of it. Now, the third takeaway is that FBCs emphasize urban design, and that doesn't necessarily lead to more restrictive zoning regulations. So, let me walk you through this figure that illustrates this point. So, here we're plotting on the horizontal axis the Wharton Regulatory Index, which some of you might have heard of, but we can talk more about it in the Q&A. And on the vertical axis, we have our FBC similarity measure. So, essentially, zoning restrictiveness here means how difficult or lengthy it is to build. Higher values mean stricter regulations. Now, if form-based codes were inherently more restrictive, we would see bunching or clustering of all the cities in the upper right area, right, where there's the quadrant with higher similarity and higher restrictiveness. But we actually don't see that, right? Instead, we see virtually no correlation between FPC similarity and restrictiveness. So, for example, if we dig a little deeper, we can see that cities like San Antonio and Oklahoma in the top right, these are cities that have bold, strong form-based and also are very restrictive. But on the other hand, you have cities like Seattle and Minneapolis in the bottom right. These are cities that have restrictive zoning, but actually have low FPC similarity. And so, the takeaway here is that adopting form-based codes doesn't automatically mean stricter zoning regulations. Now, in the last couple of minutes, I'll turn to our last task, which was to see whether FPCs affect urban form in practice. And this is important because it tells us how effective they are in shaping the built environment or shaping development. And so, to do this, we constructed detailed measures of urban form using very granular spatial data on individual plots, buildings, and streets for the entire US. And we focused on four key measures. So, first we measure how far buildings are from the street, so setbacks. We also measured how consistently they align, which you measured using setback deviations. We also measured how densely the buildings are in each plot. And here we use Floor Area Ratio or FAR. And our fourth measure was minimum plot sizes. So, we computed all these measures plot by plot, as you can see here in the maps of Chicago. And then we average them to get a single measure of urban form for each municipality. So, these essentially tell us the results for the built environment. So, it shows that municipalities with stronger adoption of form-based codes tend to have more compact urban forms. Specifically, places that adopt FPCs have smaller and more consistent setbacks. You can see that in panels A and B. They're also denser as measured by FAR, which you can see in panel C. And they have smaller minimum plot sizes, which you can see in panel D. Now, to ensure that this relationship that we observe is in fact FPC adoption, we also controlled for other potential factors like regional differences, the size of the municipality, these zoning vintage. So, all of these sort of different controls are shown in the different colors in each panel. Now, I just showed you that FBCs have the potential to shape urban form. Of course, we also care about whether these changes actually translate into differences in behavior, particularly how much people walk, how much they commute, and also just like lifestyle in general. And so, using the same approach as before, we find some support for the idea that municipalities with stronger FPC adoption have higher walkability scores. They have shorter commutes and a slightly higher share of multifamily housing. Emily, with that, over to you. Emily Talon: Okay, thanks so much, Ari. Okay. I think some of, I'm just going to wrap up with highlighting a few of the main points. Reforms are successive in a matter of gradation, zoning reforms like form-based codes. And I think a big point and justification for this study is that they aren't necessarily labeled form-based codes. These form-based code elements infiltrate zoning codes incrementally. It might be through setback decreases, allowing for mixed-use, allowing for higher density, reduced parking. And reforms are kind of layered and embedded in codes in a lot of different ways. So, what we have shown here is that natural language processing, NLP, can be leveraged to filter through zoning documents across the US to gauge FBC similarity. So, a couple of things that we learned. First, we found that FBC qualities are discernible in a distinct type of zoning code. So, we can, you know, this is not, there is a certain type of code out there, might not be labeled FBC, but it is distinct. And documents with high FPC characteristics were tightly clustered, as you saw in the scatter plot, which means that there's an internal thematic coherence to them. So, there's something to them. It isn't just that they have one-off regulatory dimensions to them that are sort of hit or miss FBC elements. There's a thematic internal coherence. Traditional zoning documents, in contrast to that, were much more dispersed and did not have that kind of thematic coherence to them. Another thing, another takeaway, is that codes with high FBC similarity incorporated themes related to physical urban form and placemaking. And as Ari showed, words like "architectural." But low FBC documents were more focused on sort of legal and administrative aspects of zoning with words like "performance" and "hazards." So, I think this difference is exactly what FBC proponents have been aiming for, which is to give more attention to the built environment, the form of the city, and less attention to the rules and procedures that could result in unpredictable outcomes, including car-based urban form. Let's see, next slide. Okay. Some of you may be thinking, "Oh my God, AI, it's destroying our lives." Does this AI classification align with assessments of actual human beings? Well, that was actually one of the critiques that we got when we went through the review process for this paper. It's always a danger, but we did conduct a validation exercise. Actually, Ari did. Thank you, Ari. We selected 10 zoning codes, each from the top and the bottom of the FBC similarity distribution, and extracted all paragraphs related to setbacks and FAR, and then developed a rubric and used a human evaluator, blinded to the FBC similarity scores, and applied the rubric. And we found good consistency. So, it's always important to do some human checks on things. AI can produce crazy results. But so you need to check that. And we did. Um, we were able to characterize the built environment associated with high versus low FBC. This was the other point that Ari made. So, as compared with low FBC places, high FBC places had shorter setbacks, higher FAR, smaller plot sizes, all of which support the goal of creating walkable, pedestrian-friendly environments. There was a positive association, as Ari showed, with walkability, commute time, and multifamily housing. FBCs are strongly associated with denser and more human-scaled environments. That's all great. You may be thinking though, correlation is not causation. And you would be right. But it's a very good sign. If FBCs had been shown to correlate with lower walk scores and longer commutes and less multifamily housing and deeper setbacks, that would be a cause for concern. So, associating codes and built form is really important because if the code differs substantially from the built environment that we're trying to affect, this could mean that the code is being tasked with making major hoped-for changes to place quality that aren't really panning out or that the code is not having the intended effect. Our methodology, like all methodologies, has limitations. So, let me just close by briefly suggesting that. Ari, do I just have one more slide? Okay. You know, ChatGPT can introduce bias. Um, and some of the metrics we use like Walk Score can be measured in many different ways. Walk Score has been critiqued a lot. And so, there's always room to extend and refine. We could further differentiate FBC places on the basis of urban intensity levels, like a transect kind of thing. We need to do more work on determining the appropriate time period for evaluating FBC outcomes. Having data on the timing of code reform is really hard to track. We tried. That's an area for future exploration. But overall, and in conclusion, I think planners and New Urbanists are justified in their optimism about form-based codes, but we do need to be realistic about the capabilities and limitations of form-based codes. And so, there's a lot being expected here of form-based codes, which is why I think we need to undergo regular evaluation, ongoing evaluation using all kinds of different tools. AI is one such tool. And actually, this is the last slide here. We have Ari developed a website that goes along with our study where you can kind of explore different outcomes. Tell us about the website a little bit, Ari. Arianna Salazar Miranda: This is just a summary for all of you that are interested in learning more about the project. It summarizes the key findings, and towards the end, you can also look up your own city or municipality of interest and compare and contrast across all the different sort of outcomes that we talked about. So, Walk Score, commute, the share of multifamily housing, and also see how those factors vary with our FBC similarity score. So, if you're interested, do go to the website. Emily Talon: Right. And I think that wraps it up for us. Happy to take a Q&A. Let me stop sharing my phone. Rob Studaville: Okay, well, let's have a… Yeah, I will remind everybody to use the Q&A function of Zoom. If you have questions, we're already getting some good questions in that. But I'd like to ask a few questions to begin with, and thank you very much, Emily and Ari, for a really informative presentation. I think the last form-based code study you did, Emily, was 2021. Um, I don't know that there has been much study nationwide of form-based codes. There might have been some in recent years, but um, over that last four years, I've noticed zoning reform taking on like a much higher profile in many respects nationwide. How does your study add to you, that conversation? Uh, does it uh, uh, does it indicate it possibly or add any depth to that? What are your thoughts on the zoning codes taking on a higher profile? Emily Talon: I think that the studies that are out there, I haven't seen any that have used AI for one thing. I think that there is a lot of research, not a lot, but the research that's out there tends to be either qualitative or looking at individual cities, which is also valuable. And it might be doing things like associating form-based codes with sustainability or looking at how our form-based codes, what do architects think about form-based codes? Do they love them or hate them? Survey kinds of things. But really tracking in a nationwide way to try to get a handle on the overall trajectory of how, how form-based codes are changing the way we build cities, the way we regulate cities. More research needs to be done. I think that this was kind of a newbie in terms of this particular approach. And so, the more the better. We need to keep on it. It could be monumentally important. Form-based coding in general has the potential to be incredibly important for turning the ship around on how we build our cities. Ari, you have any other thoughts? Arianna Salazar Miranda: I was just going to add that I think the strength of this approach is precisely what Emily was saying of allowing us to scale across the entire country. And I think we've just scratched the surface. I think, I mean, we were looking at just constructing this measure of similarity, but I think there's a lot more that we can do on the content side. One particularly sort of exciting and interesting thing that we could, you know, that could be done moving forward is looking at street design characteristics, right? Things that we know should be associated with form-based codes. So, you can think of, for example, using these kinds of techniques and tools to see how often zoning codes mention sidewalks or mention bike lanes or mention street trees, right? All these sort of features that we, that are commonly thought of as being pedestrian-friendly, and then relating those back to form-based codes or to zoning in general, right? So, this is sort of, I think, the first study to take a stab at that in this context, but hopefully it'll open up the avenue for many more exciting work because I think there's a lot of potential here. Rob Studaville: This study seems to indicate that form-based code adoptions are more widespread than I think a lot of New Urbanists might have suspected. Certainly, that I suspected. And that seems like good news, and yet I still keep hearing from practitioners that it's not a level playing field out there, that it's still difficult to get your form-based codes adopted. And so, I was wondering if you have any thoughts on that. Does it get to the capabilities and limitations of these form-based codes? Arianna Salazar Miranda: Yeah, I mean, that's a great point, right? Point one, yes, we show that the form-based code language is widespread. Now, one thing that I think is important here is that we don't find a correlation between adopting form-based codes and increased regulatory restrictiveness. In other words, cities that are adopting form-based code language aren't necessarily creating extra hurdles to build it, right? At least not from the evidence that we found. Of course, this doesn't mean that building that be form-based or traditional zoning isn't very challenging, right? I think we can all agree that in the US, it's just very difficult to build in general, form-based codes or not. And so, the point I think that you're making raises sort of a broader issue, right? That, you know, we should just policymakers should think about how to reduce those regulatory hurdles so that we can build more and increase supply so that we could build more efficiently, especially when it comes to housing. Rob Studaville: Another thought that I have is that zoning is really half of what creates urbanism. And the other half is streets and street networks. And I was wondering, first of all, if your study says anything about streets and street networks. Did you find out anything about that aspect of it? And whether progress is being made in that area, and could that, I mean, could that be something that's holding things up if there is less progress on the street side? Emily Talon: Ari, it sounds like a new research project to me. Arianna Salazar Miranda: I think so. Yeah. So like, I mean, in some sense, like some of the measures that we look at, you know, like setbacks, regularity of building frontages, density, plot sizes. I mean, arguably these are street design features, right? Or at least we, we think of them as factors that are important to shape the pedestrian experience at the street level scale. But I guess you're probably thinking about even more detailed street design features. So probably some of the things that I was mentioning before, like street trees, sidewalks, pedestrian crossings. So, I do think these are things that we don't measure, but that we could potentially measure. The only thing that perhaps I would want to say is that we do look at Walk Score as a behavioral outcome. And I know it's not a perfect measure of walkability. Emily and I talk about this a lot. But I mean, in some ways we do think that it captures some of the essential ingredients of pedestrian-friendly environments like availability of things to go to, visit, like amenities. And we find that places that have FBC adoption have higher Walk Score, right? And so, in some ways, this provides some evidence that there might be some of those pedestrian-friendly features in FBCs, even if measured indirectly through Walk Score. Do you want to add some? Emily Talon: I think Walk Score is pretty, is pretty accurate. But where it can go wrong, it goes really wrong. But I don't know. Rob Studaville: Emily, did you want to add something to? Emily Talon: Well, it's just that we were looking at documents in the Municode depository. So, that is going to be zoning, which is directed at private land. That is not directed at the public space necessarily. So, that would be a different set of documents that we could analyze in terms of their location and the degree to which they are, you know, trying to be more pedestrian-friendly. I don't know what source of documents that would be, but we will, we'll ask ChatGPT. I don't know. What do you think, Ari? Arianna Salazar Miranda: We can figure that out. But you can look at zoning codes and look for pedestrian-friendly features within them. Rob Studaville: Using the same sample of municipal code? Emily Talon: That's not going to regulate, that's not going to be regulating streets. Rob Studaville: Oh, and outside of zoning, outside of zoning. But in the broader municipal codes. Arianna Salazar Miranda: Perhaps not the zoning chapter. Exactly. Exactly. But we have all of the municipal chapters. And so, we could look for pedestrian-friendly keywords in all the municipal codes beyond the zoning one. Mm-hmm. Rob Studaville: I'm going to start getting to the Q&A because we already have a whole lot of questions, and I'm wondering whether we're even going to get through them, but a really quick one was: what program do you use to establish the text chunks, angle distance from the base form-based code? Can you talk a little bit about the programs used? Arianna Salazar Miranda: Yep. So, we use the combination of Python for everything that was the natural language processing, the processing of the text, the splitting of chunks. So, this was sort of the programming language that we used. And for all of the statistical analysis and actually constructing and measuring the cosine similarity, we used our RStudio for that. And maybe I should add that within Python, we used the NLP model is actually open source. It's called the BERT-RoBERTa model. Anyone can use it. It's free. And, you know, with some coding experience, you can get it running and extract insights yourself. Rob Studaville: Patricia and okay, go ahead. Emily Talon: And by the way, sorry, Rob, quick question. Will you have a record of all the comments and the questions? Because I'd like to look through those. Rob Studaville: Perfect. I think we can download that, yes. Because we won't get to them all now. Emily Talon: Okay, thank you. Rob Studaville: Yeah. As time progresses, Patricia asks: the term shift, and did you see shifting terms relating to form-based codes and, and different vocabularies, and how did that impact your outcomes? Emily Talon: Well, as I mentioned, we don't have a timestamp on these codes. So, we weren't really able to look at temporal changes. I think that is super important to do. So, we will be revisiting this. You know, going forward, we need to keep track of what's been happening. But we weren't able to date the codes, were we, Ari? Arianna Salazar Miranda: So, there's a small caveat on that. The problem is not so much that we weren't able to date the code. We were able to predict the date. The problem is that Municode only releases the latest version of that code. So, we don't know whether there were any reforms, any changes in language that happened prior to the last zoning code that we get. So, but I think it's a super interesting question, right, to see whether there are different trends in how the codes have talked about FBCs. And I think it would be super interesting to explore. But unfortunately, we just have a single point in time. Now, we could imagine that if we do this starting now and into the future and query the latest Municode every year, then we could start tracing the evolution of language over time. So, it's a great point. Mm-hmm. Rob Studaville: Karen asks: is plot size the footprint of the building or the lot size? Arianna Salazar Miranda: I think that was mentioned in plot size. Plot size is the, yeah, it's the dimensions of the, of the plot. Of the lot. Parcel, exactly. A lot. The parcel. Yeah. Rob Studaville: Yeah, okay. Did you measure relationships to city size or person density, or the density of people, uh, in a place? Arianna Salazar Miranda: We mostly controlled for it. Made sure that the relationships that we were seeing wasn't because, let's say, you think that larger or smaller cities adopt more FPCs. And so, we basically made the comparison fair in terms of controlling for differences in size so that we made sure that it wasn't just larger or smaller cities driving the effect, but actually you're comparing cities of a similar size that one ranks lower in terms of FBC adoption and one ranks higher. And so, we did look at municipality size and population. Size in terms of area and population. Rob Studaville: Marina asked: did you evaluate how fast projects are approved? We tend to try to streamline in exchange for better. So, that would be good to know. I mean, you did look at the level of restrictiveness, but whether that correlates to how long it takes to approve a project, I don't know. Should I take that one, Emily? Or do you want to say something about it? Emily Talon: Well, I'll just say about the restrictiveness index, you guys can look up that. It's pretty interesting data from the Wharton's, came out from the Wharton School at Penn. And that was based on a survey, right? Exactly. Right, Ari? A survey planners asking them questions to try to get a measure of the degree of restrictiveness. And it's a lot of it's, it's really about the dominance of single-family-only zoning is kind of a big part of that restrictiveness. Like where in the city can you build and how much of land is restricted to only single-family zoning? Arianna Salazar Miranda: Yep. So, I'll just add by saying that we, our form-based code similarity measure doesn't incorporate that. It's just a measure of similarity in terms of language to true form-based codes. But as Emily was saying, this Wharton Regulatory Index, in theory, it captures precisely that. So, you're going to rank higher or be more restrictive if you have lengthy processing, if it's difficult to build. You, they have a rubric of, I think it's like 20 or more dimensions that they rank. And they actually report the individual numbers for all of these dimensions and then construct the score. So, if you're interested in exploring that measure and the sort of the disaggregated rubric, you can also do that. It's publicly available. Rob Studaville: Now, you gave some examples of municipalities in the four quadrants. I think it would be very interesting to get a list of municipalities in those four quadrants so that you know if you're going into a place that's got form-based coding, but it's very restrictive, it may not be easy to build there at all. But if you go to a place where it's less restrictive and there's form-based codes, that's a different matter. Or the other two as well. You know what you're going into. Arianna Salazar Miranda: Exactly. I was actually, I think Emily would say the same. We were surprised to see how weakly these two dimensions are correlated in practice, right? That you can really find cities in all of the four quadrants, right? It's not sort of a one-to-one linear relationship, and I was, I was surprised by that. Mm-hmm. Rob Studaville: Is there any of that information that would be available? Arianna Salazar Miranda: Of course, we can share the list of municipalities in the four quadrants. Yeah, absolutely. Rob Studaville: That would be wonderful. Dwayne… Emily Talon: Yeah. And it would be great to hear from people. Sorry, Rob. It would be great to hear from people who are ground truthers for us to tell us that, "Hey, my city ended up in this quadrant, and that is not my experience at all. It's actually not restrictive at all." And it, or whatever. We need to keep training the model, right? Ari, more intelligence on this stuff, the better. And please let us know. You can email me at [email protected]. And I'd love to hear actual stories and how our data compares to the ground, the real world. Rob Studaville: Yeah, exactly. Absolutely. That's very interesting. And you've got a lot of people with real-world intelligence on this. So, Dwayne asks: how does the model distinguish between the form-based code adoption and mere use of similar design language, and what does that mean for interpreting form-based code similarity scores? You did a check on that, and you talked about that a bit. Do you have any further thoughts on this? I didn't quite understand that question. How does your model distinguish between the form-based code adoption and the mere use of similar design language? And what does that mean for interpreting the similarity scores if there is a distinction between those two, the adoption and, and the similar language? Emily Talon: Well, all the codes in our study have been adopted. They're law. They've reached the point of being codified in the Municode. But I think the process, just to reiterate the basic idea of taking that set of bonafide form-based codes, so says the Driehaus Award was one of the lists we used, right, Ari? Ones that everyone would agree. Arianna Salazar Miranda: Yeah, for sure, these are form-based codes. Emily Talon: Using that to understand and taking that through the NLP and then using that information to compare with all the other codes. So, that's the process. And I think that's having that set of bonafide codes was really key to making this work. Arianna Salazar Miranda: Yep, I'll just add that design language is part of our FBC similarity, right? So, if the true form-based codes speak about pedestrian-friendliness and speak about all of these different features of design in one way, and there is a zoning code that speaks about these same issues in a similar way, then the idea is that our measure of language will pick those up. And then the FPC similarity score is going to be high, right? Because they're going to be very similar. But I think this sort of points to some of the other point that we were talking about at the beginning. That we're not really getting at sort of the different topics that might matter for FPCs. So, think about, you know, pedestrian-friendly environments. You could list the elements that are important for those and then check whether FPCs speak about these things, how they speak about them, and then do that check for all the zoning codes, right, to try and understand that specific theme or aspect in a more sort of nuanced way. Rob Studaville: I don't know if your study really got into any aspect, a question of implementation of form-based codes, but we did have a question on that. And maybe you have some insight. What would you recommend is the best way to guide smaller cities who are interested in form-based coding, but have a difficult time wrapping their brains around it? Often these cities do not have certified planners, but are seeing the need to adjust their codes. And they don't know how to begin. Emily Talon: Well, there are a lot of good resources out there. You know, the Form-Based Codes Institute, which is now in, what is it called? I forget, but it's with Smart Growth America. And I was just looking on their web pages, and there's, there's some webinars to access. There's visual resources that are, we now have a couple of decades worth of trying to get the educational tools in line to that everybody can use and adopt. So, I would look to that. I could say you could come to CNU 33 as well, and you might find people who could help you answer that. Yeah. Rob Studaville: Let's see. Hazel asks: during this study, did you do any nuanced review of which zoning regulations with high form-based code correlations to see what percentages of them are aimed at mature communities of walkable, mixed-use, complete communities versus what percentage are aimed at suburban retrofit of malls and places like that? That's a good question. Emily Talon: We talked about that, Ari, about how maybe we're just capturing cities that are already walkable or form-based codes are already codifying what is already walkable. Hi, Hazel. I'm so glad you're on here. Hazel started this whole thing. So, thank you, Hazel. Ari, what do you think? Arianna Salazar Miranda: So, I think what we did to try and tease out or distinguish between the real effect of FBCs versus this other story of FBCs just being adopted in places that already had a different built environment. So, everything that we showed today is cross-sectional, right? It's a single point in time. Which means that essentially we cannot be completely sure whether it's that form-based codes caused these urban forms or if municipalities already had these types of built environments prior. To do this, of course, we would, uh, data like we were talking about before, we need longitudinal data, right? That tracks zoning reforms over time and also that tracks subsequent changes in urban form over time. We don't have that. So, what we did is we took advantage of the fact that most zoning rules were set before 1950. And so, we ran essentially the same regressions, but only for the subset of neighborhoods built after 1950, which we think of as places that were developed under the existing zoning rules, right? And so, that helps us address concerns about urban form influencing FBCs rather than the other way around. And reassuringly, our results were pretty much the same. So, we find comparable evidence which supports the idea that these associations might reflect the effect of FPCs and not the other way around. But it's a tricky one. And it's the most, you know, the most difficult. So, the more progress that we can make on this, the better. And so, I think we, it's on us to move forward and sort of start collecting this data over time. Rob Studaville: We're getting close to the hour. And can we continue? We've got so many more questions. Can we continue to answer questions after one o'clock? Do you guys have a hard stop at one? Ari has a plane to catch. Arianna Salazar Miranda: Yes, I have a plane to catch. Rob Studaville: Okay. Okay. I mean, I could go for a little bit longer. Let's see. Jen asks: instead of expanding form-based codes along our transit-oriented development, the planning department is using an overlay district. Do you have any, did you get any insight on the use of overlay districts and the pros and cons of that in your research? Emily Talon: Well, if they had overlay districts that were in the Municode, we didn't really separate that out. That was just part of our data set. So, if it's codified, an overlay district is possible in the code, then that will be analyzed as such. But we didn't go into the spatial particulars of where the overlay code within a municipality was designated. Arianna Salazar Miranda: That's exactly right. Oh. Rob Studaville: And so, that's just, that's just a different method of doing it. Pros and cons. I think, I think, I think… Arianna Salazar Miranda: Just to add to that, Rob, because I think it's an important point to clarify that it was basically a data limitation, right? We have the zoning document, but we don't know what aspects apply to which neighborhoods within that municipality. And so, our unit was basically the bigger municipality, but there might be a lot of variants and how things are getting actually done on the ground, right? And that's sort of part of the story that we're completely missing currently, and this is because we don't have a good mapping between what zoning, what parts of the zoning code affect what neighborhoods or zones within that municipality. So, all the efforts also that we can do to sort of link those two will also give us a better understanding of what's happening within each of these places. But, you know. Rob Studaville: I'm just going to say that, you know. So, we are at one o'clock. Emily Talon: Okay, go ahead, Emily. Data is getting better and better and more open and more open. And I'm coming across new data sources all the time, and AI is there, and it is getting better and better. These models are getting better. So, I feel like in the future, even five years, three to five years down the road, we might be in a position where we have a really much better handle on what's going on with these codes and where are they being implemented and what is the impact and how is it changing things. And I think that that's going to be really important. Arianna Salazar Miranda: Hi. I would also add, given that we have so much local expertise here today, I would love if you can put in the chat like, what is it about form-based codes that you would want to know and don't know? Like, what is it that is most interesting to all of you on the ground? We would love to know that. Emily Talon: Love to. Rob Studaville: I just had a thought that our form-based codes becoming like memes, and they're self-replicating. And if that is the case, are all codes going to have some elements of form-based codes sometime in the future? Emily Talon: Maybe. I mean, I was surprised at how widespread these form-based code dimensions, even if they're not called form-based at all, how widespread they were. So, I think that's very possible. Certainly the, what's happening with zoning in general, it's a national conversation. It's reached the kitchen table that people are actually talking about the impact of zoning and how bad it is, or some people holding on to it, but it's, um, that's different. That's so different than 10 years ago. So, I think there's something catching on there. I would be really interested to explore more about this aversion to the term "form-based," which, you know, that has been a critique. You know, we all know the story, architects, designers really being not happy about being told what to do and all the critiques about form-based. But is there some sort of insidious way that, or not insidious, but stealth, stealth way that the basic idea of paying attention to urban form is infiltrating, and that is what our results are kind of indicating. Rob Studaville: Tina has a comment and then a question. She says she really appreciates the rigorous effort to evaluate outcomes from form-based codes. That's very welcome. Thank you. The question is: did your study distinguish commercial versus residential outcomes? FAR and minimum lot size. Greater housing production is particularly of concern in terms of desired outcomes in Colorado and the United States in general. Arianna Salazar Miranda: That is a great point. We actually did not differentiate between residential and commercial FAR. This is in part because we have geometric information for all of the plots and buildings in the US. But we don't have the use for all of them. And we've learned progressively that even though in some cities we have it, the quality varies a lot. And so, we just kept it all about the built environment, regardless of its use. But I think there's a lot of future work that could, as you say, incorporate this dimension because I think it's an important distinction. Rob Studaville: We had a question to Professor Talon. Given that Euclidean zoning remains steeply embedded in the American planning system, both institutionally and culturally, what do you see as the most significant challenge to advancing form-based codes and practice? And from your perspective, what specific skills or paradigms should planning programs prioritize at the university level to better prepare the next generation of planners to implement and advocate for form-based approaches? Emily Talon: Well, on your second part of the question, I remember 20 years ago when I was in planning just trying so hard to get the planning academics to care more about built form, teaching built and form, teaching urban design. I have a book called "Urban Design for Planners." So, I fought the good fight. And now I'm not in a planning program. And so, I, but I, don't see that there's been a huge sea change in that. And I don't really, I can't really answer why that is. There's this annoying separation between architecture, design, and urban planning. And it didn't used to be that way. And like, you know, at the beginning of the profession in the early 1900s. And it just separated. And I don't see that it's really come back. However, that's the academy that it does its own thing on the ground. Planners who are practicing planners, I see that they are very interested in urban design issues, and there's a thirst for understanding all of this. They know that the built environment matters greatly for things like social justice and, and, you know, environmental stewardship and all kinds of important planning values. So, I have a lot of faith in planning practitioners, and maybe that's what we're seeing here with our, and now, you know, our surprise, all these form-based code type dimensions infiltrating. Maybe we have urban planners to thank for pushing that agenda. I think so. Rob Studaville: Christine is looking for guidance, tips, and tricks for presenting information, graphics, and data on form-based codes to elected officials in a way that they will understand and embrace. Particularly officials that emphasize the slow growth, no growth, focused on transportation issues, etc. Emily Talon: You know, one of the slides I showed toward the end there was from Urban Advantage. Steve Price is a, has sort of a cottage industry. I don't know. Steve, he's still out there doing these amazing visualizations of what a, a different kind of landscape or built environment does. And the visualization, and then the book you're talking about next week, Rob, with the webinar, "The Art of New Urbanism." There's just so much fantastic visualization. And so, I would get those visuals front and center, and a picture is worth a million words. In this landscape of talking about form-based codes, you need to make it all visual. Rob Studaville: I agree. Okay, okay. Well, thank you. Arianna Salazar Miranda: Rob, I have to run to catch my plane. But I'm seeing a whole lot of wonderful comments on the chat that we really need to save. Could you do that for us and share it afterwards? Rob Studaville: Yes. Yes, yes. Arianna Salazar Miranda: Thank you, Arianna. Safe travels. Emily Talon: Yeah. Thank you so much. And thank you everyone for joining us today. It was a real pleasure to share some of our work. Nice seeing you. Bye-bye. Rob Studaville: Thanks, Ari. Emily Talon: I'm going to sign off too. Thanks so much, Rob, for having us and looking forward to the follow-up. Rob Studaville: Okay, yeah, thank you both. And thank everybody for being on the Park Bench. It was a great webinar.