Episode Transcript
[00:00:00] Speaker A: Imagine a justice system built on rigorous evidence, not gut instincts or educated guesses about what works and what doesn't.
More people could access the civil justice they deserve.
The criminal justice system could be smaller, more effective, and more humane.
The Access to Justice Lab here at Harvard Law School is producing that needed evidence. And this podcast is about the challenge of transforming law into an evidence based field.
I'm your host, Jim Greiner, and this is proof Over Precedent.
[00:00:35] Speaker B: I'm Michelle Bluein. I'm the communications manager at the Access to Justice Lab at Harvard Law School. And today we'll be talking with Renee Dancer and Matthew Steubenberg. Hello, Renee and Matthew. Can you both take a minute to introduce yourselves and tell us about your current job and what you were doing prior to that?
[00:00:50] Speaker C: Yeah. Hi, I'm Matthew Steubenberg. I used to work for the Access to Justice Lab. I'm currently the president of my own solo firm, Steubenberg Legal Group. And prior to that I was at the William S. Richardson School of Law as their innovator in residence.
[00:01:06] Speaker D: I'm Renee Dancer. I'm a researcher at the Access to Justice Lab. Been on the podcast a couple times. Very excited to be here with old friend Matthew Stubenberg because we like to.
[00:01:16] Speaker B: Always find out a little bit more about our guests and we know a lot about Renee at this point, but we're going to find out more. We like to ask about a fun fact that you can give us about yourself. So, Matthew, anything that you can share with us?
[00:01:28] Speaker C: My fun fact is for the past four years I lived in Hawaii and I just recently moved back to Maryland and I am adjusting to the weather much slower than I would have hoped.
[00:01:38] Speaker B: I think the follow up question to that is why? Why we turned back.
[00:01:43] Speaker C: That is everyone's follow up question.
[00:01:47] Speaker D: Just kidding.
[00:01:47] Speaker B: Well, that. That sounds fantastic. So, Renee, what about you?
[00:01:51] Speaker D: I will just editorialize Matthew's fun fact. He often refers to Maryland now as the Hawaii of the East.
So which I love, I think is funny and like to remind him of that.
Okay, so my fun fact is I have been giving you all fun facts about things that I like. And so I thought maybe I would go darker this time and give you.
[00:02:13] Speaker B: Things that I dislike.
[00:02:15] Speaker D: And so I think I've come to this realization perhaps too late in life, but I think I might be the Grim Reaper. Reaper for plants, I cannot keep them alive.
I try, but not really very much because I actually also don't find joy in gardening like some folks do. So not a green thumb.
[00:02:37] Speaker B: Okay, good to know.
Okay. Well, we are here today, we're discussing a publication that came out in the North Dakota Law Review and it addresses criminal justice record clearing. And both Rene and Matthew took part in writing this, as well as faculty director of the A to Day Lab, Jim Greiner.
Excuse me. And, and maybe we can start off with a discussion about expungement and record clearing and where that plays in at the lab because we currently have a study ongoing. And so let's, let's first discuss the study before we dive into the paper.
Renee, you want to take the lead?
[00:03:11] Speaker D: Yeah, sure, I'll take the lead on that. So we actually the longest study I've worked on at the lab is the record clearance study that we are winding down in Kansas. So that study will, you know more. We'll have a more fulsome discussion of that study I think on a later podcast episode. But essentially what we wanted to understand was what is the effect of record clearing. And since we can't randomly assign who gets their record cleared and who doesn't, that's outside of the scope of our authority here at the Access to Justice Lab. That scope is very narrow actually.
And so we had, what we needed to do was find an instrument that we could randomly assign that would allow us to infer the effect of expungement. And that instrument here was attorneys and full scope representation. And so since Kansas Legal Services is over subscribed, but this is one of the services that they do provide when they have the capacity. We were able to randomly assign full scope services representation for expungement and or record clearing or self help materials, actually a feature that Kansas Legal Services had already had in place prior to this with the recognition that at times they were over capacity. And then from there we hypothesized that attorneys of full scope representation will. Folks who are represented with full scope representation will have far away greater success at having their record cleared than those who just had the benefit of the self help materials. And if that proves true, then we might also be able to untangle how the effect of record clearing on certain dimensions. And in this study we looked at housing security and employment stability.
But that's the. And so we're still like we're in the data analysis stage where we will release a report at the end of this year. But we still have some data collection to do. We have a five year follow up period for our study participants. And so the final enrolled participants still have two years on their data collection follow up period. And so then we'll issue a truly final report thereafter. But we are about to close out one of the grants that funded that research. And so as a result, it's a good time to issue kind of a interim report, but it will be fairly robust because we have lots of data.
But that's just the record clearing study. And this paper is about the different kind of side quests that we did with the benefit of Matthew's technical expertise during his time with us at the lab.
[00:05:56] Speaker B: Yes. And we'll be sharing a link to the paper, paper itself in the show notes. So if anybody's interested in reading it in full, we have it provided. But Matthew, maybe you can expand a little bit about the paper and how that came about.
[00:06:11] Speaker C: So I've been involved, I'm a lawyer and a coder and I've been involved in a lot of data scraping projects over my career. And because we already had this study in Pennsylvania and Kansas, I took a look at the kind of data that was available on the court's kind of case lookup websites. And I realized that we could scrape a lot of this information and provide some extra about what cases are eligible, you know, how many cases might be eligible if we change certain rules. And especially in Pennsylvania, we could do an audit of whether or not clean slate was effectively implemented by the court system. And so we could back up a little bit and talk about what kind of the clean slate rule is. The clean slate rule went into effect in Pennsylvania and actually went into effect in about mid-2019, and it basically automatically sealed a number of eligible cases.
And so there's a number of requirements that have to be met before a case can be automatically sealed. But if those requirements are met, the court will automatically, on their own, seal those cases. And so what we were able to do is go in after that process was completed and take a look at did we find any cases that we identified as should have been sealed and were they actually sealed? And we actually found a lot of cases that by our metrics, should have been sealed, but were still showing up.
[00:07:35] Speaker B: Great. This is that awkward pause where I need them to edit up because I didn't look at my notes quick enough.
So based on your analysis, what kind of challenges did you come across in looking at Pennsylvania's clean slate law? What are the limitations of it?
[00:07:52] Speaker C: Yeah, so, you know, Pennsylvania, the clean state law doesn't just make everything sealable. It had a kind of complex mechanism to determine whether or not something was eligible.
And so for misdemeanor convictions, a person had to wait 10 years. They had to pay all their fines and restitution, and they had to avoid any charge that carried the potential of one year or more. So kind of a serious misdemeanor or felony within that 10 year window. It's commonly referred to as a subsequent conviction rule.
And if that was met, that charge was ineligible for sealing and the court should have removed it automatically without the defendant ever having to do anything. There were also some summary charge convictions that were eligible. It was a little bit easier to seal those and the non convictions were eligible, so not guilty, things like that.
For this study, we looked at four counties in particular. We looked at Butler, Beaver, Lawrence and Allegheny County. Allegheny county is by far the biggest county of those four.
And for our scraping project, we pulled cases from 1990 to about 2021, so about a 30 year time frame of cases.
And we pulled all these cases in early 2022. So this is about two years after the court had already implemented the clean slate. So everything should have been removed by that point.
We only looked at common plea cases in Pennsylvania.
There are other types of cases, but, you know, given the volume of cases, we had to kind of limit what we were looking at. You know, potentially projects in the future could expand on that. But this is kind of our first swing attempt at getting some kind of metric from these numbers. So the numbers were from Allegheny county, we pulled about 350,000 cases. And from the other three counties, Butler, Beaver and Lawrence, we pulled somewhere between 30,000 and 50,000 in those other counties.
And the interesting results is we found 200,000 charges, about 200,000 charges that we identified as being eligible for sealing, but that were still available at the time. We scraped this Data in early 2022, and so it brought up this question of how is the court determining which case is eligible for sealing? And, you know, when it came to certain, you know, data integrity issues, how are they coming down on whether or not a case is eligible for sealing or not?
[00:10:21] Speaker B: So was this one of the initial intentions of this, this analysis is to figure out where some of the limitations were, or did you assume that they the data was fairly clean coming in?
[00:10:33] Speaker C: You know, I always go in hopeful that the data is clean and I'm proven wrong every time. So this project had a number of kind of data quality issues that are very common in every court across the U.S.
one of the biggest issues we ran into were a lot of really odd dispositions that we couldn't really assign a value to in order to determine if it was eligible or not. And we're not sure exactly how the court made these determinations either. So some of them were things like held for court was one type of disposition.
Another one was proceed to court. These all sound like kind of active case type of case statuses, but that was the final disposition in the case. We couldn't find anything else in the record after those were introduced. We also had a really. We also had a lot of none disposition where we just couldn't find a disposition in the case.
And because a lot of the ceiling is dependent on what that disposition is, and a lot of these cases, we just couldn't make a determination as to whether it was eligible or not.
Now, you know, the court has more information than we do probably, and so maybe they have some other ways of determining whether or not this is eligible or not. But there were definitely some data quality issues from the data that we were able to pull publicly.
[00:11:50] Speaker D: And for either of you, did I.
[00:11:52] Speaker B: I think I also read in the paper something about some database migration issues that may have also caused some. Some problems.
[00:12:00] Speaker C: Yeah. So Pennsylvania moved to a single court management system in the early 2000s. And this occurred on a county by county basis. It took about a decade to do. And during that process, they would migrate from whatever the old system was that county was using onto this new system. And we found a lot of migration issues there that when a court would move on and migrate all their old historic data onto this new system, sometimes that's where we saw a lot of those really odd dispositions like held for court or migrated disposition was a disposition we saw a lot, which doesn't help us determine if it's a conviction or non conviction. And so, yeah, so we attribute these to just migration issues, which, you know, I do have sympathy for, that's trying to migrate years and years and years of historic court data onto a new system must be difficult. But it did hamper our ability to determine ceiling eligibility.
[00:12:56] Speaker B: So what happens to that data that falls and falls between the cracks?
What are you doing with that in terms of how to, how to move forward with your analysis?
[00:13:07] Speaker C: Yeah, largely we put those in kind of the unknown bucket. We, we didn't count them as eligible unless we could identify them one way or the other. And so it's possible that we under counted the number of eligible charges for ceiling that would be found. And so a further analysis could go back and try to determine if those cases resulted in one way or the other. And perhaps migrated disposition is just some, you know, placeholder for a disposition we could actually determine or if there was some new database where we could pull the dispositions for those cases. We could expand on this study.
[00:13:40] Speaker B: Renee. Maybe you can kind of put the data into a format that we can kind of visualize in terms of what type of records are we talking about here?
[00:13:49] Speaker D: So Matthew mentioned that these are common plea use cases. That's in Pennsylvania is the trial court level.
And so cases that are being held for court or instructed to proceed to court, which is a very old terminology, those are cases that are being bound over from the lower court, meaning they start there as an arrangement for their arraignment and then it's actually the jurisdiction of the trial court to dispose of them.
What the types of cases we're looking at are exclusively criminal cases. And for this so far, you know, in the expungement study that we discussed or the record clearing study that we discussed at the beginning, we're really only concerned with the types of cases that are eligible for record clearing. In this paper we really tried to interrogate that eligibility to understand if we expanded or contracted it in certain places, how would that change the population of people that may become eligible and is that a goal and a value of policymakers in these two jurisdictions as an example of what perhaps could be considered going forward? But the first hurdle, of course, is all of this, the data integrity. And we did try to be very clear in the paper about where we just couldn't know for sure.
I am a big proponent of regular auditing of court records. I think this is one of the things that courts don't spend time on probably and from because of resource constraints, courts and clerks of courts offices don't spend a lot of time on auditing their records to ensure that the data is clean. And that's what we see in when we do any sort of data analysis with, with court records. And so but this is, this becomes more important when you're relying on clean data to automate a process.
And so if you don't, because, because at that point you are letting your algorithm do that work.
And so you also then need to spot check that. But if you, if, if it's working with dirty data, that's what it's working with. It's not it, it's not trained to say this is dirty data. Right. And that would be a hard, hard to train it, I think. I mean again, this is like me diving into a world that is not my expertise. And so I'm so glad to have you here. To say that is just craziness is.
[00:16:36] Speaker B: Now the, the process of clearing a record involves a few different Potential obstacles, I believe, for, for people. And that could be fines and fees and waiting periods and, and wonder if, if either of you can address those challenges in terms of just obstacles towards, towards getting a record expunged.
[00:16:59] Speaker C: Yeah. So I, part of the requirement is that you have to pay all of your restitution if any was ordered.
And then there's a number of kind of other requirements of around waiting periods and whether you have subsequent convictions. And so we were able to dive in and take a look at if those rules were changed, how many cases would then become eligible.
And so you can actually go to the paper and you can see, well, if we reduce the 10 year waiting period to a 5 year waiting period, how many cases would be removed. And for the restitution, we broke it up into buckets of how many people owed $100 or less or $250 or less or $500 or less.
And so if the legislature still wanted to keep some kind of restitution provision in there, but maybe wanted to lower it, you know, they could see what the impact would be by doing that. And I think that's, this is one of those things that isn't commonly done because the data just doesn't exist. And so hopefully this will be kind of the start of analyzing this big data to be able to give policymakers and legislators a little bit more information on the impact of their decision.
And I'll also say that, you know, from there was a number of what we call kind of lifetime convictions that would prevent any cases from being eligible or any convictions for being eligible for sealing. It was if you had a certain number of felonies or if you had a certain number of level one misdemeanors on your record, it would prevent you from getting cases sealed automatically. And so we also looked into those to determine, well, if we change that rule, how many cases are currently being blocked by that restriction. There's also a handful of particular statutes or charges that if you were ever convicted of that charge, nothing else really mattered. You couldn't get anything sealed. And so we looked at those as well and said, well, if you remove, how many of those are actually relevant? And for, you know, it's only like six charges.
And for most of those charges, we couldn't find anybody who had been charged with or convicted of those crimes, which was kind of surprising. So, you know, our legislators spending a lot of time making sure a handful of these things get implemented in a bill when it's really not impacting anybody.
And I will say also in terms of kind of the difficulty of making some of these algorithms and determinations. One issue we had is because we only looked at data from four counties, if we're looking at things like subsequent convictions or these lifetime conviction rules where you were ever convicted of a particular charge, we could only look at those four counties. And so if you were convicted outside in a separate county, we wouldn't be able to see that data. We might be, you know, determining that case is eligible when it isn't. And so, you know, we make that clear in the paper. And if we had data from all the counties, we might be able to do a more robust determination. But this is hopefully seen as kind of a first step towards making some of these decisions. And hopefully other researchers will be able to build on this with more data and kind of clean it up and, you know, bring their own analysis and algorithms to it.
[00:20:02] Speaker D: Absolutely.
[00:20:03] Speaker B: And in Pennsylvania's clean slate law refers to automatic clearing. Am I right in saying that?
[00:20:11] Speaker D: Yes, that's right.
[00:20:12] Speaker B: Okay. And do they have also petition based record clearing?
[00:20:16] Speaker D: They do also have petition based record clearing. So that's a great distinction to make. So the beloved term, the Clean Slate Initiative, is really about automatic record clearing. And so all of the criteria that Matthew has already described for how we determined what was eligible is really the criteria dictated in the legislation or at the time of, at the time we wrote this paper into this analysis, that legislation has throughout the years been expanded in snippets here and there.
So that refers to automatic record clearing. I think Pennsylvania really was out front, out in the front of this initiative. And so they. And that was really because of their unified case management system where they had all of their records already in one electronic place where they could then create a system that could do this automatic sealing of records. They do have petition based sealing, and they do that predates their automatic record clearing, but also now works in parallel to automatic record clearing. So petition based sealing was the precursor to their automatic clearing. So they first did legislation that allowed for petition based sealing, and no one was doing it because it's petitions in and of themselves are arduous to complete, and that also presupposes that you're eligible. And in the paper we talk about some of the very nuanced and complicated eligibility requirements and that.
So we thought that was nuanced and complicated. After looking at it for months and months, you can only imagine how an individual who is just like, thinking about it on their own, with no law training might struggle with that.
But now petition based sealing is really used as a companion to automatic sealing for those cases that have these weird dispositions or somehow are missed by the automatic ceiling algorithm, but the individual believes they are eligible, they can petition the court and ask the court to take a second look.
That comes with all of the same barriers that petition based signaling had at the start. But it is an attempt and I'm not sure what other attempt there could be other than in addition to auditing and cleaning up the data to allow for mistakes in the automatic sealing process and still make. Make sure that folks who are eligible have.
Can take advantage of that benefit.
[00:23:11] Speaker B: So condition based clearing brings its own access to justice issues, but it seems like it's the best we've got as an alternative to automatic clearing at the moment.
[00:23:21] Speaker D: Yeah, and we have a different paper that I'm sure will be. That will be a different subject of a. Of another podcast episode from the Kansas portion of this evaluation which only has petition based clearing.
And talking in that article exclusively discusses the barriers that exist in a petition based clearing, which are somewhat problematic for anything that requires an initial petition or application.
[00:23:55] Speaker B: Well, this North Dakota Law Review article brings up Kansas a little bit near the end. Is that something we should discuss a little bit? You know, it goes into the petition based clearing and some of the obstacles there. Is that something you can expand on now or do we want to wait for a future podcast?
[00:24:12] Speaker D: Oh, I think we can expand on it as it relates to what's in the paper then Matthew, if you want.
[00:24:18] Speaker B: To share anything you came across as far as the, the data analysis and the limitations there, particularly in comparison to Pennsylvania.
[00:24:27] Speaker C: Yeah. So you know, we, we looked at Kansas as well and you know, as Rene just said, we, Kansas doesn't have automatic ceilings. So this analysis looked more at how many charges were eligible with petition based expungement. So if people actually went out and filed the form, how many cases you know, were. Were actually eligible for expungement?
So we scraped all this data in about 2020. We scraped it from Johnson county, which is the biggest county in Kansas, and a handful of other counties that are had already moved onto the statewide kind of uniform court case lookup website. Not all the counties had moved on, unfortunately, also including the second largest county in Kansas which had not moved on to this portal at that point. So we couldn't get all of the data from Kansas like we would have wanted. But we got the largest county and then all of the counties that we could from the statewide portal. So in total we pulled about 255,000 cases from all the various counties.
And the results were we found about 50,000 convictions that were eligible for expungement in Johnson county and about 90,000 that were eligible in all the various non Johnson county cases out there. So there was a large number of cases that were eligible for expungement in Kansas that we were able to identify.
We also ran into kind of data quality issues.
One really interesting one was the Johnson county court system had their own website that we had to scrape. And then there was also this state aggregated portal for some of the other counties, but they didn't necessarily have the exact same data on both portals. And so for instance, in Johnson county, we couldn't determine when the the sentence that the defendant received. And under expungement law in Johnson county, you have to start counting the waiting period based on the end of the sentence.
And so we couldn't really do that in Johnson County. So we had to use the disposition date right when they were given a conviction as that starting period. And in order to make kind of a rough estimate as to whether when the waiting period would be over. And we tried to account for that and what kind of the issues surrounding that by looking at what the average sentence given was and all of that. But it's just one of those data quality issues that you work with the data you have, not the data you want kind of situations.
We also ran into situations where we had to match, try to find other cases that a person might have on their record in order to determine if it was eligible or not.
That's always fuzzy at best. Trying to do kind of match people. Unless you have some kind of specific like Social Security number or state ID number, which we didn't have in this case. And so we had to do this fuzzy matching based on name and date of birth, which is good if you have a unique name, but if you have a John Smith or something like that, there's chances are there are other people named that born on your birthday as well. And in the non Johnson county cases on the statewide portal, we only had the year of birth. So it gets kind of even fuzzier.
So there were some issues we had to work around. So, you know, as is the case in many kind of data research projects, we did the best with the data we had.
And I will say all of the algorithms, everything we did to analyze this is all publicly available on GitHub. The link is in the paper. And I encourage anybody who's interested in kind of our algorithm or data analysis to go take a look if you find A way that we could have analyzed it better. Let us know and maybe we could reanalyze the data.
[00:28:03] Speaker D: Excellent.
[00:28:04] Speaker B: Maybe we can bring this discussion to more of a, more of an overview with policy. Is there anything that you are gleaning from your results in terms of recommendations for policies or you're talking about expanding and reducing eligibility? What is the goal there from a legislature point of view?
[00:28:27] Speaker D: I think we wanted to just give a flavor of how you could use the existing data to predict what might what to predict the effects of changes in legislation or policymaking.
So, you know, one of the things that I thought was pretty captivating that's in this paper is Matthew mentioned earlier the requirement that in Pennsylvania that someone be crime free for 10 years.
And we analyze well, what if we change that to five years, how many people would be eligible? And it was, it was an improvement, right? It was, it, it made, it would make a huge impact.
And when you couple that with the research that's still evolving and we are clear about that in the paper. But when you couple that with the research that's still evolving about, you know, propensity to commit crime and to, to be involved in the justice system, it kind of makes sense to make that reduction. And it doesn't seem like it is, oh, you know, a political burden. And so the other thing we did in the paper was say you, there are things you could do and these are the impacts they would have. But there are other considerations for policymakers like how big of a kind of political issue would this be to make that change. So Matthew mentioned those six or so offenses that nobody was actually charged with. We never found them in the data, but they are a complete and total ban of any record clearing. They are fairly egregious sounding offenses at least. And so while it makes sense to make, to, to maybe make the legislation silent on those or change that provision, I think it'd be a hard sell for a politician to say this is my platform going to say we should change legislation about expungement for this, these particular crimes. So I mean, I think that this was a different area of concern for us. We're typically not, we're typically like doing the research, giving the evidence, not as, not, not always singularly thinking through policy. I think our goal is to help our policymaking friends think through policy with research. This was just a different, a different way to do that. And, and you know, we look forward to seeing others doing that too.
Yeah.
[00:30:58] Speaker C: And I'd like to just expand on that is another kind of purpose I think is when legislatures are writing these rules, they can take into account that the court ultimately is going to have to write some kind of algorithm to implement whatever rules they make. And so legislatures can keep that in the back of their mind that if they add some particular provision, it might make the process 10 times more difficult for the court to actually implement. If they require things like subsequent convictions within a period, keep in mind that then requires that the court has to basically analyze the person's full record. If there's, you know, issues with state IDs not being accurate or identifying if this case belongs to the same person, it suddenly introduces a lot of issues. And so those subsequent conviction rules might be warranted. But just keep in mind that implementing things like that might cause significant delays in the actual technical backend process. And the same goes for odd dispositions. And so, you know, we had a lot of these like none dispositions. So just we were missing the data. And so what could happen is if we can show that there's actually a large amount of cases that have this, the legislature from the beginning can say, okay, if a case has a non disposition or some odd disposition that, that we can't make heads or tails of, count it as this and automatically seal it, or don't automatically seal it or make a third category.
But try to kind of proactively think about those situations as well early on so that the court might have an easier time kind of coding for those situations.
[00:32:38] Speaker B: These are great points. Definitely not black and white formulas there.
[00:32:42] Speaker D: Just one more, one additional thing, I mean, to Matthew's point about, you know, streamlining the legislative drafting process in Kansas, you can see the legislation was drafted piecemeal. I mean it's, you can just intuit that as you're reading it.
And even, even for.
So in Kansas, we don't yet have automatic record clearing, although the move to a fully online docket maybe makes them that state a good next, you know, iteration of that. But you can see that through the legislation that in certain periods of time, certain offenses were eligible and then they weren't, and then they may be in part were. And so what that means is even for practitioners in this area, for lawyers who are experienced in record clearing, they still also need to navigate was this offense charged in this period of time, does that make this eligible or ineligible? And so that is difficult logic flow for a practitioner who is doing this. And so it would also be difficult programming.
So just to consider like as we're thinking about making changes to legislation, data informed changes, of course, obviously we would only do that thinking about it more holistic, the piece of legislation more holistically, and making sure we're not just doing a piecemeal modification that in the long run is going to make it more difficult to actually realize the benefit that we're attempting to give to the community.
[00:34:33] Speaker B: Good summary.
[00:34:35] Speaker D: Is there anything else? You know, actually, I apologize for this.
[00:34:39] Speaker B: Offhand question I should have asked at the beginning, but how did you end up working with Kansas and Pennsylvania initially? How were those two particular states in those particular counties decided on that.
[00:34:55] Speaker D: Is simply through relationship building. So a professor at the University of Pittsburgh School of Law did a visiting professorship. I don't think that's the technical term, but at Harvard.
And Professor Patchou and she and Jim Greiner, our faculty director, began kind of exploring this idea in Pennsylvania with the legal. And Professor Chu had a lot of connections with the legal aid organizations in Pennsylvania, so we were able to begin working with them. And then our colleague April Faith Slaker really had relationships with the Kansas folks. And Marilyn Harp at the Kansas Legal Services, who was the executive director at the time, was very excited and interested. And we, she and I have been working on this evaluation for a while now for Pennsylvania, the clean slate project or the like legislation was implementing as we were getting started on the study and making the.
Changing the dynamic of the evaluation. So now that record clearing is automatic, it made less need for Legal Services to really focus on that. So we ended up sunsetting the Pennsylvania site for the record clearing study that we talked about in the beginning and strictly worked with Kansas, but that didn't mean we didn't already know, like, how robust their data was. And so Matthew already had experience scraping websites and it just seemed like a good place to try this type of work as well. In addition to Kansas.
[00:36:36] Speaker B: Makes sense. Would either of you like to add anything that we haven't discussed about your findings in the paper or broader policy discussions?
[00:36:48] Speaker D: I think this is an exciting piece of work that we've done. It's different from anything else we've done. I really like the idea of the forward focus, the predictive modeling for what could, what could be if we as, you know, a society decided these were. These were the values that we wanted to project in our legislature, then did that as the elected body that we get with the power that we gave them. As the elected body.
[00:37:17] Speaker C: Yeah. I will just second that. You know, I, I think this was a really interesting study that we did here. And, you know, we, we were limited in part just because, you know, we just a limit on person power as well as time and resources. And so I'd love to see studies like this in other states that could potentially expand and look at the state in the whole. I will also say that coming up with the algorithm to determine if something is eligible or not is kind of extremely difficult.
And so having more and more lawyers that have some kind of tech background or interest in tech, I think is really useful because then there could be more eyes on this algorithm to determine if something should be tweaked to make it more accurate.
[00:37:59] Speaker B: Sounds like a unicorn field there that I mentioned. There's not too many, but yeah, Matthew.
[00:38:05] Speaker D: Is of the vanguard of this field, though.
This is the best guess we could have had.
[00:38:12] Speaker B: Well, we're lucky to have had you. Matthew and Renee, thank you both for joining us. We look forward to discussing the.
Excuse me. The interim results of the expungement study coming up. And, and we'll, as I said before, we will share the North Dakota Law Review article in the show Notes of this podcast.
[00:38:32] Speaker D: Great.
[00:38:32] Speaker B: Thank you.
[00:38:33] Speaker D: Thank you.
[00:38:33] Speaker C: Thank you for having me.
[00:38:35] Speaker A: Proof Over Precedent is a production of the Access to Justice Lab at Harvard Law School.
Views expressed in student podcasts are not necessarily those of the A J Lab.
Thanks for listening. If we piqued your interest, please subscribe wherever you get your podcasts. Even better, leave us a rating or share an episode with a friend or on social media.
Here's a sneak preview of what we'll bring you next week.
[00:39:01] Speaker E: And so, you know, just to provide a really quick background, what we're going to be talking about in this podcast, we're going to be first moving on for, like, a background on cash bail generally and what the system means for people going through it.
Then to the California referendum and what it actually did and would have impacted. And finally, the rhetoric and the ways we sort of start thinking about bail reform and ending cash bail and how it might give us some lessons in the future about, like, how can people who want to end cash bail learn from its failure in the California electoral process?
