TRANSCRIPT
00:02
Welcome everyone join tonight. I’m so excited that you have all decided to join. And as my as Delilah from my local radio station. You’ve done your part and enjoy the show. Celebrate each and every one of you is a celebration for all of us, and it cannot be possible without all of your hard work. So far, together, we have a new power, so I’m going to get started. Let’s get this party started. And a lot of people have joined us, and more are in the waiting room than they are joining us. So, let them join in on the fun. So, welcome everyone. Good evening.
01:08
Good evening. Good evening. All right, hello. Oh, there’s so many people coming so I need someone to take care of design and maybe I’ll make you a moderator. Is that okay. Yeah, sure.
01:48
All right, cool. So we are going to start tonight by you know, awarding and honoring our mentors, and they have been the backbone of this hackathon like being there with all the teams for the 48 hours 70 hours we get paid extra time to all the teams, and without their commitment and dedication, he couldn’t be here. So we want to celebrate them, and he has some very fun awards for them, but it’s going to be based on what you think. So we have personality of our mentors and is going to be based on your who you think deserves the best award. Okay, so there’s a fun gentler gamification, years, and hands ready, fingers, and who gets to work for so everyone gets to work, but we have a fixed time. So here we are going to start with the first one. The great insulator award. And this award is given to the mentor, who has been the greatest and creator, during the hackathon. And I know that you will only know the mentors that you’ve worked with, and that’s okay, that’s fair. So feel free to start polling, I’m going to start polling in a second. But then you need to say what you think and who you think deserves this award. So the first one. Which man or SME deserves a great and courageous Award, and the poll is like, you have 15 seconds or fewer seconds. Okay. You can see the poll at the bottom. Is everyone able to see the poll? And no, unfortunately. Okay. As soon as it launches I’m sharing the screen so it should show you. Because we have technical difficulty, I’m going to wait. Or you can go to the polls below, and then click on that
04:36
Isn’t technology fun. All right, we’re going to end this poll here. For now, we have many, many more boys so please make sure that you get this sorted out. I’m going to share the results with you. No, I don’t think we can see your screen. I can, I can see I’m sharing third country of poll.
05:36
Yeah, I’m not sure, let’s see. So are you able to see. Not everyone was able to see the whole person reach out again.
05:51
For everyone able to see them all I can relaunch it. There you go. Are you able to see it now. Yes, that’s just killing me here, versus, like,
06:27
yes, you can still win again. And you have nine times nine special category awards for every member is a winner, each vendor will go home tonight. How’s that, that’s all right. All right, so we are going to end this. And now I’m going to share the results. Some time. I wasn’t able to vote for both of them. So, somebody should just be the person. Moving on.
07:53
Because no zoom has been made it so difficult. So, the second is, who has been the greatest listener for being the greatest listener ever goes to the next slide.
08:27
First listen to me complain all the time. name is Amir. All right. Looks like everyone has passed their votes, and the poll. Oh, Bruce 30%
09:32
Have to compile these results because obviously, some of them are valuable so then she was the best one for. All right. Awesome, everyone is enjoying little. But the next motivator. Who inspired you the most, the greatest motivating LED lights. Realizing Suzanne is not in this for mentors. We consider ourselves. Because you guys. Alright, that’s it for now. Oh my goodness. That’s embarrassing. Oh, This. All right. You can guess what the next one is a steam. Oh that’s coming. These are just some great ideas. So as someone who has abandoned or was there to help you. All his life mentors who have tried to get other mentors. Yeah, I see the dilemma I can see the. If you click on the polls you should be able to see. Yeah like that should be exempt. Everyone see the poll. That’s right, the seven people voted and 29 people are participating so that is out me and Susanna, so I guess everybody’s 27 knots, she says. So, awesome that’s just at the bottom of your screen you should show up. Some ideas are trying to see what’s the next one, a great helper for being the greatest helper ever as a mentor. Yes, yes. is life. I go on camera. Oh, terrible. But there’s more seconds. Yeah. Yeah. Yeah, they have to think. OSHA regulations. And based on what categories you will avoid for this one. The next one is for rain Collaborative for being the greatest collaborator ever, And the man who has helped us the most. Big ideas in life. Welcome passion. All right, everyone. I think people are still thinking 30 seconds from now on we need to. Just think of what you want to do this and show results purchase through our dye. Duck purchase regulations hustler. According to you, as the hustler as a mentor or as an SME,
17:46
who has keep on track. Follow the timelines, made sure you submitted this one. You didn’t do so bad.
18:06
Probably to possibly be me, when the only person who made sure that every single group was training in everything. And, basically, they can come out unless they didn’t get on board, which was Suzanne put herself on there. I wasn’t gonna put her up for like so many things but she’s like I literally was going to submit a proposal because of her emails I
18:34
got scared. Just grabbing a lot of hours this evening. I mean you’re only going to get while everyone go home. All right, who is the greatest morale builder. All right, we need to wrap this quickly so that we can go over to Team presentation. Faster,
20:10
Finish on time. Poor Joe is like us on Eastern Time and spending time with us and is waiting for long. Right. Okay, I think they finished 29% at a time sheet. And I got my dog off that was in the video place with me some of you might have seen their
20:59
Houses too. All right, I think this is the last one. Great Communicator. You’re ready. Ready. Thinking. Sorry, I missed. It’s still going on, you can still prevent results by the Liga rules for people in Thai communicators, bizarre NF dos. Congratulations to each and every one. Thank you, mentors, and subject matter experts. So next, we have presentations, and they’re going to start with AI in criminal justice teams that presented as team status code, or you guys are here and ready to present.
23:28
I think only myself and leave with equal join too early. First of all thanks for giving us this opportunity. As you can see, our aim is status quo. We have worked on a very important problem in our society is good friends before I dig deep into it I would like to give us a shout by
25:02
Myself working in this industry for the last few years, trying to find interesting datasets in this project I work in a coordinator, kind of role. I also helped with a team dashboard which we created, help our team to come up with different mutations which we don’t have many people that we have limiting here so
25:41
Yeah. I myself am from Chennai, India. Currently I’m working as a software developer in the automobile industry and NLP enthusiasts. In this episode I contributed to my way of building chatbot a conversational AI solution to the problem that we’re so happy to join you.
26:07
I am Australia and people from different parts of the world study Linden, not join see mainly work from the visualization individuation part, but he also had this 10 point solution. He helped us with the final review of the art defense which is
26:29
Suited to help us with the national government, the last few days on the chalkboard solution, Give some important insights on how
26:48
He did his data analysis on the data which we got the data he asked, he helped us find some key insights or solutions. Raghavendra worked on the content data in the system. Pay things which we highlight was how social media is impacting creating the injury claims. We don’t have much of a data to showcase a solution, but he was part of the solutions, but mainly
27:23
A page which will cater introductions. As I said, we are very important. Our society in tribes as a collection of word which is moving towards a phase where it looks like he has that slide, engage our computers recognize. But yes, we can make one finding what are the impacts that have caused the high level impacts which is happening, being one of the main ones, clients, the plants have been most important, Because it. Basically, in terms of a graphical view of the summation dashboard. Because often you’ll see that the numbers are going to be who I have a visualization which gives insights of what’s happening. So BSD students that has worked on this ambition to end on this problem. Why do we think this is raising problems facing social media age, no face? When you’re face to face, you might not say or spread those kind of fake pages, which we which people do with your social media platform, everyone now has a mobile phone. It is so easy to spread fake news, political agenda. So, the problems we’re solving with this, we are mainly solving two problems. One is what’s happening in terms of hate crimes. We had us data so all our work was needed to us. But I think this can be extended throughout the whole. He also worked, which was the main part of our solution was empathize with people who have suffered bias, all the time. So, those two problems. I lost a liberty to discuss and to showcase more of how our solution can enhance future resolution.
31:16
You all know that 10 minutes maximum to present right. So please, let’s keep it other than that, thank you so much. So I’d like to elaborate on the magic solution that we have built for this hackathon. So our solution consists of a dynamic dashboard, a power chatbot and an awareness page. The dashboard and built using Tableau, as I mentioned, so I should primarily focus on case analysis. In the US, basically, information access ratings difference per year, and total incidence and victims for each US state and hedge funds classification, basic offenders. One more thing is that the most common hate crimes that occur is also visualized so that’s the second solution that we are providing over chatbot, which was built using Google’s data to NLP engine, and we have named it as Justice buddy, so it’s so straightforward. It will cost people about eight brands, and its nature. What does it grant, what can you can and avoid crime? Only these basic questions. It helps them. And secondly, it helps them to safely report ahead friends without any hesitation or fear, because many people many prefer to put up a hate crime, compliance, and suddenly it connects them with normal focus groups, undergone similar sufferings, similar experience similar models, okay. So the third part of the solution that we have heard of the BAP is an advanced web app which is a one stop solution happening around them. And it’s always an interface to reporting. So these are the things that we have. As you can see that this is the program that you purchase. During the study abroad, it has a hackathon, it can be enhanced. The things that we would like to propose as a future prospect of isolation is that can be made more dynamic, just keeps track of all the headlines and updates regularly, thus providing one solution as I mentioned earlier, and the Chatbot can be integrated with a various range of platforms with Slack, without Telegram, and it can also be integrated with government websites in your organization’s website, and also sites, and the web app can be enhanced in terms of providing more insights regarding payments. So these are things that we like to take as a part of our solution. Thanks.
33:59
Thanks, thanks. I would just like to conclude the presentation on this problem and come up with a place to sand amendments, or people who think that there aren’t many events happening, is just a one off case. There’s not much happening. So it’s, it’s an eye opener for those people. Also, most importantly, we thought, and worked on a solution for the people who are suffering from this. The main part of our solution is to empathize with those kinds of people. We know that there are a number of people working in organizations who are working to clean up the social media platforms and they want to stop the spreading of hate, fake news and all those things are happening. Of course, as we speak, a lot of solutions which I think, to improve it. But there is not much. We are we are very far from seeing that we will do. So I would like to end this with two sentences, is we cannot kill the heaters. Otherwise, he said, and pray for him I will blind the whole world, but we can always empathize with the people who are suffering. Because, as you have been said, Fear of disease kills more people than the disease itself presentation. Before that, I would like to thank our mentors committees Fatimah and devise an awesome helped us, guide us. And this would not have been possible without you guys. Finally, thanks to the organizing committees, we’ll be looking forward.
35:56
Thank you so much. Status quo team this is awesome job presenting partner solution. And they also interview mentioned how you want to continue working on it and we have some ideas on those solutions. And now it’s time for us if you could stop sharing. Alright, so the next team that is going to present is Ai COVID contact tracing team two. Are you guys ready? Again, I want to
36:37
Get to connect but I guess my teammates. So I really want to help them. I’d be happy to help you present perfect, if I’m saying should be like I get the point. Because other are other team members here, but I think Audrey and I can just kind of sit back and forth.
37:00
Yeah, so I get caught up in what we were doing and then you can kind of, it’s it should be.
37:21
You let me know if you can see my screen, you can see that. My name is Mark and I have so much team on the call. We want to give such things were mentioned apart. Yeah, so I’m going to start off with just. So, think about COVID-19 content. We really have to take into account the different populations that are affected by COVID-19, and how COVID-19 may not really address the population. So for example you can think about, you know people who are people of color, people don’t have health insurance, and how we think about our contact rating application to address, and be able to capture. So for example, there’s a lot of different equities, And a lot of companies like Google and Apple for example and coming up with new content creating an application is something we really have to consider is how do you take into account any type of competition. We’re not going to come up with a contract with. So it really matters because we want to make sure that you know all of the populations are represented right, and you can have a lot of bias, if you’re doing, and behind this group, you know, for example, people who are older, may be more susceptible to COVID-19 Definitely, they are, but they’re probably less likely to use smartphones or any technology. So how do you, you know, kind of make sure that there’s inflammation integrated into it. So, we have to we have to design something that’s segmented and kind of take into account all of these populations. So Brett’s gonna share that individualization. So we looked at over, I think, 3540 searches, and we consolidated our features together to identify what are some of the advisors that we could see in the data that we’ve had for COVID-19 And what does that mean tell us about how do you design, patience, and quality. In order to Trump, 30 different user at different segments of the population. So we looked at, we basically segmented the different population, but not race, ethnicity, ownership and employment and more. So I did want to start off by talking about that.
39:59
Yeah. We checked out the United States healthcare data. As we can see here, in all states. Most people have a private insurance. However, we can also identify that there is still a sizable percentage of people without health care like they’re uninsured, that’s in there. And while the amount of people who are publicly and privately insured do outnumber the amount of people who are uninsured. That’s still a lot of people and that means that there are still these people in particular are very vulnerable to Coronavirus, and we need to keep that in mind when creating these contract tracing apps,
40:42
Too. I like to have. For example, people who are probably uninsured may not even want to declare, they have COVID-19 As you might have heard from kind of half the population with COVID-19. Very handsome example. So people are uninsured, or we don’t want to do that because if they’re taken to the hospital right, or they may be running into all these hypotheses, which they may not have.
41:11
As we can see here this is political leaning by state, and as you can see there’s a mix of conservative and liberal people and COVID-19 people stances on subjects such as masks and healthcare are affected by their political meaning, meaning that some people may not take precautions that they need to prevent to getting COVID-19, or may be more likely to get COVID-19 in areas where people do not take these prevention preventative measures into their political beliefs, and that this will cause differences in the demographics in the United States, which means that we need to change how the apps, track it, depending on where we’re focusing on. As we can see here in the United States. These are different phone ownerships by different demographics. This is important because if someone doesn’t have a cell phone, then it means doesn’t have a smartphone but it has a cell phone that means they will be unable to use these apps, which means that it makes it basically inaccessible. For example, the elderly range of 65 Plus, most of almost half of people who are on phones, on a cell phone, but not a smartphone, and they can’t use the apps that are needed to track and so we have to reach these demos demographics in different ways than just that. As you can see here these are the amount of COVID-19 jobs in the United States. Typically, the older you go, the more likely you are to COVID-19. Men have a higher death rate than women in almost all the demographics, with the exception of 85 years and older, because the percentage of female gaps is actually higher. I mean the number does not percentage. And now, this is our statistics on unemployment versus age. This is very important because as we can see there is a amount of people who are unemployed have lower income, meaning that they may not have accessibility to COVID. Maintain treatment and health care. So, perhaps, massage, 16 and 17 year olds are there 16 to 19 year olds have the highest amount of unemployment, however they are the youngest demographic and that makes sense, considering both of them are high school students in that age range, or early college student. So, most people aren’t
44:30
Really open towards hiring these younger demographics, especially if they’re under 18 and due to legal issues.
44:38
But the older demographics, they all have lower unemployment rates, that is also possible that unemployment rates change depending on changed between 2019 and 2020. If you compare like the bars like you’ll notice there’s little bars and there’s a big bar, that the employment, employment rate increased by an amount. The older you go the unemployment rate tends to drop. However, it’s also noticeable that for people of color placement rate is significantly higher than for other demographics.
45:22
And then in the next few slides. We tried to show here between technologies we’re thinking about. COVID-19 is a data normalized by the publisher, so it’s not like the data and you know how that got there but you would actually be that white or Caucasian tend to have a high COVID-19 Hi everybody, normally I can find the information. That tends to be interesting to come in, depending on how you interpret and evaluate to kind of see the same thing across other GA to Georgia, actually, if we had not normalized it again kind of white tend to have more Americans and so what we don’t know and that we thought we should be public population. And then in the afternoon is that a lot of them minorities. Hispanic population as well as Native Americans tend to be like the highest that we even know when you look at the overall population if you will not normalizing this
46:49
Decided and tended to have the highest interest in it much, but in terms of trends that we noticed that the US. So we had about earnings. And then, or did you want to conclude.
47:11
Yes. So, taking, taking the future steps on improving on COVID-19 tracking, we’re gonna need to continue investigating the data and from the United States and other variables. This is very important because, there definitely needs to be more research in this area and identifying the bias. We need to create a model to predict which methods will work that’s where, since this is very location sensitive. We want to increase contract tracing apps will increase accessibility, for example visuals instead of words, due to the fact some users may not be very well versed in English, or may be less technology literate, as leader more accessible for younger generations than older. And typically, so we want to counter that problem, and we want to develop plans and continue collecting data for people without the smartphone app, for example through calling and mailing, because not everyone has a smartphone.
48:19
Impact on black and indigenous people, community is that in the United States and the UK they suffer the most from COVID-19 That’s, I think it’s important to reduce communicate communities to share information, and increase awareness and improved the collected data. Our goal with this is to improve COVID-19 tracking for them. And, and especially for their communities, and to reduce the spread.
48:56
Think about you know where devices in our data, and how we think about more effective operations and COVID-19. And you know, different way that we can target them. Yeah, thank you so much for your time. Thank you. Thank you so much everyone. Maybe insights on the job. Right. Next ratio. Right. I’m going to share my screen. I know remind all the teams. When the time goes to seven. So,
49:51
Okay, great to see this. Great. So hi everyone I’m Willie Castellano, I’m here with my teammate Jasmine for our other teammates sorry I couldn’t be here tonight, And she says, regrets. We’re here to share with you our, our submission and our web page that we call racial bonds. It’s a statistical analysis of pretrial detention and bail bond practices in the state of Connecticut. And this webpage as you’ll see, takes the reader through our problem. And through our magic solution. So what we’re going to do in this presentation is is walk you through this website and tell you what we did. I’ll say at the start that Jasmine. Did our concept, a lot of the grading that you’ll see myself in Daria, we did the data wrangling statistical analysis data visualizations that you’ll see. And we’ll start off by. I’m going to hand it over to Jasmine and take it
50:57
Away. Thanks. Going to start with our issue, the broad context for setting up the HR team explored rest and the implications of incarceration in the criminal justice system, we acknowledge mass incarceration in the federal and state prison systems but focusing more specifically on a large proportion of unconvicted individuals, more than 70% of which are people of color being detained across the country as they await trial. Past and overpopulate, resulting from pre trial related racial disparities continued to grow due to many systemic factors we’ve focused on the increase in adoption and use of algorithms, specifically pretrial risk assessment instruments as an interviewing factor. These assessments intend to predict the risk of a defendant reinventing or failing to report. Additionally, these algorithms are used to determine the terms of attention such as setting thought and the ethical problem presented by this is that the data that drives these algorithms may be bias. This may contribute to the racial disparities that you’re seeing in pre trial detention populations due to the use of systemic and economic inequities indicators of risk goal of our webpage to explore the impact of risk assessment algorithms through racial disparities and a free trial in a population. We also hope that serves the purpose of raising awareness of this growing and facilitating conversations around the use of these algorithms in court systems in the United States.
52:34
Thanks to take it over from here and talk about our specific case study, because when we decided to do for this webpage was to look at one particular state. This is Connecticut, and we chose Connecticut, because first they’ve been using risk assessment algorithms since 2003 and their algorithm was recently validated in 2015. And the second reason is that the state of Connecticut makes publicly available data on all of its pre trial inmates, and they do this nightly. They reveal information on who is being held in pretrial detention. Now we don’t actually know everything about these inmates that the risk assessment algorithm knew about them when they were determining the conditions of their detention. But we do know several things about them. We know the amount of their bond, the number of days they’ve been detained the offense for which they were arrested. And we also know, their race. So we wanted to study the impacts of these pre trial risk assessment algorithms by, by studying, you know what the effects of the pre trial in the population were what effects that had on the pre trial risk assessment, pre trial, pre trial inmate population. So the first thing we looked at was just the overall racial dynamics, demographics, and as you can see here, the blue line at the top is the black population. And the vast majority of inmates are always from the years 2016 to today have been black. Oh, and secondly by white to me it’s a classy especially in recent years by. Now, you might think that perhaps these pretrial inmate population demographics are just reflective of the state’s demographics but that couldn’t be farther from the truth. So the vast majority of Connecticut’s population is white, black, next, people make smaller and smaller minority. And yes, the numbers are completely reversed in the pre trial inmate population. But the disparities don’t appear, it’s not just that we see differences in a number of different reasons but in particular. Notice, vast difference in the meaning and bond amount that’s been assigned to white inmates versus inmates of color, so you can see clearly here. If we just look at the median bond for students of color, it’s, it’s over $10,000 higher than even the highest estimate immediate for white men, you still might be thinking well perhaps this because this is because the beads of color tend to be arrested for different offenses which tend to carry higher on demands. And we’ve looked into this, and we found that even if you control for the type of offense. And still, you see these racial disparities. So we’ve highlighted seven. Seven particular defenses here in this chart, and you can see in each case, the median bar for innovative color is substantially higher than white, whether it’s violation of probation or breach of peace, even when the offense is the same, we still see these patterns. Furthermore, we showed that this is just because it needs to color tend to repeat offenders, either. Even if you control for that, we still see these disparities. We also looked into gender differences, to see if that played any role, and we did find that generally speaking, the situation looks a little bit better for women, there’s a higher amount of racial equality in terms of bond demands, when you’re just looking at female inmates but still there are a few offenses, like the ones listed here were female inmates of color do still have the kleinbard white female inmates. We looked a little bit into what might explain these differences, and one of the interesting things we saw between the male and the female population is that well as we saw before the male population like, like the general population is majority black female in the population, is it has always been in the four years of our data, majority white.
57:15
So, from these findings. The question we’re left with is okay, what is, Connecticut. Now, what is the state’s pretrial risk assessment that we don’t, because as I said at the start, the data that we have from Connecticut, on their profile in a population isn’t all the data that these risk assessment. Now, when they’re setting the terms, or the detention. Now we know, according to reports from the state, that these risk assessment algorithms take into account the marital status, their employment, their education, their substance abuse history and our mental health history. And so the question that we want to leave readers with is okay if these are the additional data points that are being fed into risk assessment. Is this really justice, or is this is perpetuating societies existing in Africans. So I’ll hand it back over to Jasmine now.
58:19
So as I mentioned, we explored Connecticut’s use of retail risk assessment instruments, but the implementation of these tools is not exclusive to Connecticut. This visualization, courtesy of mapping retail and justice, gives us a better understanding of utilization of these instruments in their various forms, all over the country. We hope that having tools like these featured on the web page and support viewers and exploring this issue in the context of their specific state and country or county. This is definitely an overwhelming issue with many moving parts, and despite the challenges, there has been modeled success, the District of Columbia’s pretrial services they can see properly uses these instruments in combination with minimally restrictive supervision practices, and social interventions to sustain large proportions of reach on defendants that are released. Do not reinvent and do show up to court, compared to approximately 60% of individuals and local jails around the country awaiting trial, and 12% of those in DC jails are pretrial that
59:25
We have only a cross section of this national issue in order to contribute to the conversation around ethical use of data in criminal justice contexts. There’s much more to learn, and action to take. Mapping racial injustice the algorithmic justice, and data for black lives are great starting points for continuing to learn about this and we encourage viewers to explore their resources.
59:47
So thanks again, that brings us to the end of our web page, our presentation. Moving forward, we really want to expand this webpage to kind of include more statistical analysis to include more impactful data visualizations to look at data from dates and bring that in, and ultimately also develop a interactive model, where users, adjust certain parameters and see how on demand might change. But thank you, thank you very much for listening, and best everyone. Thank you.
1:00:25
Thank you. Great job. Thanks. Good night everyone. Good evening, good morning everyone. We are Team gaveled. Our Team J J M Jonas will be accompanying in our presentation. My name is Chuck wake and try to present our solution to the challenge in the criminal justice, justice issue, develop the hybrid solution, takes the form of web application, gamble.
1:01:31
Closely examine a widely popular commercial AI risk assessment tool, which uses predictive compass is used by judges for scoring. Facebook activity because it’s biased results. This bias not only emulates sentiments of a racist judge, but has also already done, its damage by villainizing certain racial groups in the system, as well as failing to do proper response for those who may be an actual society. Unfortunately, this algorithm exasperates the inequities within the criminal justice system, and further criminalizes dmpc communities be identified. For first and foremost, compass, first and foremost the compass, risk assessment tool is there, as well as showing during our presentation. The compass is 2.14 times more likely to school, African Americans as future syllabus, one competitor.
1:02:45
Second with current concern about machine learning algorithms. They operate as black boxes, make it possible to identify how and why
1:02:56
Each particular decisions recommendations or predictions. Yet judges rely on these awkward increasingly accepted decisions on questioning. This poses a problem, especially when it comes to verifying legal decisions as results. Lastly, this lack of transparency and accountability. Compass shows the judges. Final risk, but people’s lives should not be blindly trusted, as the decision maker, but must remain the role of the decision support human machine. So we asked ourselves, what do judges need to see to change their minds concerning our algorithm is designed for the principles of Explainable AI and showing the discrepancy between different races, or solution offers the judges, presenting on racial bias. In comparison, treated differently. With our solution to channels that make by AI and raise awareness about the judges to use these algorithms responsibly, and as a tool of justice should never be the sole responsibility,
1:04:27
Decision making, must remain the judges. Now my teammate Jonas we’re working through the tuition. So I’m going to actually show you the demo. So it will sell. So as you can see here is amongst the Trans races so perspective rate percentage person being wrongly credited as in this case. So, as an African American, you have much more chance to be credited as future miscellaneous consequences. So well. First, we had to replicate proxying them but some, some performance. And we also removed the variable race and when we found that it had the same discrimination. So, discrimination. So that could be one of the main one is that your friends rate, different races, is kind of different in African Americans. First like this guy’s raise a lot of questions. Deep rooted injustice in such disparity in terms of understanding. Actually biased. Because as much more strict with African American. In terms of slides, what do you think? So, here you have the kind of sweat. Find the proxy we had binary
1:07:03
Classification. So, like you have two receivers. So, x axis, and C. What is it that you choose? So, again, is your rich personality traits, decrease, and it’s different. That’s a consequence. So when compass. The compass was to assign the same kind of for everyone. So same treatment to everyone, but here. These are the consequences. Once you decided to choose this. What’s up everyone, the present and then as
1:08:34
The different races, and that implies that, having different treatments for different races so that’s the choice impacted. So now I’m going to change the application.
1:09:07
Number one, anyone hearing me. Everyone knew probably. Thank you. Thank you join us. My name is Dave Johnson, and I’m proud to be a part of the team dabble 2.2 my responsibilities include working on data visualization, collecting statistics for and brainstorming the web app, and writing the article. I will now take you through a demo of a web application gobbled users presented the case study, the case ID is entered in the search bar. That’s under case 555 taken to a page with information on case five by five. On the left we are presented with a dependence profile where we see the individuals name, their prior offenses and their subsequent two parameters, if they would like to see more information about the defendant, we can click more info. The functionality of the session hasn’t been built yet so let’s see this progress, please check back later. But in future. This page will include historical and current information about the defendants in question, which will give the judge a sense of who this person is and what their criminal record is like. So going back to the main page, we see two scores on the right side. This score labeled as combat algorithm score tells the judge what scores range from the complex algorithm. The second score algorithm score presents a normalized score, produced by all algorithm. As we can observe in case five by five, or algorithm has produced a lower score. Our hope is to ensure that the judge is cognizant of the current biases and errors associated with the algorithms, and to gain more insight into the case studies presented. So, while. As mentioned earlier, the combat algorithm is known for producing test scores by race and voting present the judge with an opportunity to view the arguments directly according to race on the left side we have six options, choosing one of these options generates corresponding error rates produced by the company. Because this is supposed to judge to the assumption. On the right, with almost a whopping 10% difference in error rates, which is substantially lower and the need the scores, the judge can choose to learn more by viewing a preview of a factsheet or go back to the initial page and see another case study by going back to the case ID screen. If the judge chooses to view the Batchi, he or she will be in two sections. On the left we provide a brief description of the aims of novelty pointing to an explanation of the IRS we observed a comparison of the accuracy between the algorithms and a final statement on the right side of the judgment and general statistics regarding algorithms used for predicting recidivism. If we are afforded the opportunity to work on this in future, sometimes more information and data visualization to this function. So all in all, our goals are to provide the judge with supplementary information so that a more important decision can be made to provide support and help bring answers to a UC LGBTQ, and other marginalized communities, as well as individuals who are being treated unfairly. Three direct accountability back to the judge, through the use of Explainable AI. We hope that this demo can serve as a beacon to be amazing web application gavel to point to pique your interest into the potential impact that can bring to our justice system. And you’re also having your time and patients are grateful for the opportunity to present. Thank you. Thank you team 2.0 this is fascinating job. I’m actually really impressed. All of this product solutions that they’ve come up with. It’s not easy to like when the whole world is coming out with, you know, making it better and the solutions and criminal justice predictive policing contact is not easy to come up with those, just for you know, thinking of how we can do a job everyone. Just a little bit running out of time so please. Now we only have five minutes for being, please just get an hour out somebody, everyone can look at the details of your article, and all your work later on. This is just the time of everyone’s money and also I would like to invite to Criminal Justice. Are you ready? Nice. I guess.
1:14:00
Okay. That works. Okay. Hi everyone, I’m interested in criminal justice team one name. Today we’re going to be tackling some of the intersection of data, ethics, in particular, we’ll be focusing on challenges of data and criminal justice system and what that might mean for policy. So everyone is here today, which is amazing because everyone should be a great person and introduction I’m Kristin Lee, California and I helped coordinate and manage the project as well as help with the policy. High School with research. I also attended. I helped with the policy
1:14:55
Agenda I worked at Morgan Stanley, and I helped with the presentation.
1:15:03
So our team came together, joined forces that got to thinking about why we chose this topic of criminal justice in the first place, and something that we came to realize was that in the words of James Baldwin color is not a human or a personal reality it’s a political reality. Color is a political reality, which means that society operates a political system that judges you based on the color of your skin, not the human or person behind it. So to illustrate a statistic. Black Americans are in turn, incarcerated at six times the rate of white Americans and use longer sentences for crimes.
1:15:44
Okay, so I think that we’re
1:15:47
Probably shouldn’t. It shouldn’t be so extreme that we have, like, a dimension of time.
1:15:53
So what we decided to do is to try to assess the presence of biasing us criminal justice system. In contrast is exporting to other countries. Now we decided to contract with other countries because it cannot be assessed, without any point of reference.
1:16:13
This comparison may celebrate interesting results if we compare also to compare the analysis of the society and ecosystem to see how we can decrease the amount of bias.
1:16:29
We know our sources from the ICPSR, the UCR ended at the corporate level of analysis we decided to examine the amount of the US versus UK overall, as we said we were able to see, about 3% of the US population with the rest, which 2% the UK, which is 68% difference in the amount of people who are arrested for the purpose of uncovering potential racial bias people get our data into four variables, the black population the white population Asian population, and American population, and doing so, we see that black people are more likely to have the recipe, any other race checker IDs rates of arrest by crime. And by prime time, after doing a breakout of people arrested by phrase. You see that this is consistent across the board, even when we’re looking at specific crime type 60 Bytes as we see here. So we should break out our pies to examine out, which states which aid level are invaded by race across the globe with the rest, we see that race a factor. So, when we then begin to bring out and as to every splash we see in our comparative results. But unfortunately, you’re still in the UK, but people are bracing UK commuters.
1:17:55
So after looking at the results from these datasets we decided to create a policy solution that’s centered on international crime Distribution Analysis. Specifically, our solution is to create a policy tool that supports detecting racial bias through the creation of a readily accessible open source crime data available to the public. So it means both governments and everyday people can perform their own analysis on non-identifying crime data. And the reason why we want this policy to emphasize making this information, open source, and including an international plan is because as I mentioned earlier, you can’t really detect bias, if you have nothing to compare the bias to, so maybe like everybody’s equally biased, but it’s hard to tell what qualifies as bias in the first place without that sort of internet perspective. So hopefully this solution will shine a light on bias and also encourage more equitable policing by facilitating the work of comparative lawyers and promoting an independent branch of knowledge that serves as a check and balances against possible abuses of the criminal justice system, or solution. Dictation is our presence on the state level
1:18:59
In purchasing society, legislation, and it can be done pretty much right now for states
1:19:12
To the international scale. It could become a resource for the competitive for the competitive players, so they can see like the consistence would be interesting for comparison results as compared to the US. And potentially, it’s good. Integrated twins’ policy checker. In order to see how employee sees changes biasness system. And later on become a machine learning model, which could even predict what’s
1:19:50
In justices. So we know that our policy is not perfect for one we weren’t able to get the exact information. And due to the constraints in the publicly available data we have to focus on eroticism 60 Not exactly representative of how many people were actually thinking. The data is biased in itself, you know the factors such as sex, age and income, and other socio economic factors, and data reports themselves. We also can control the different ways across states and countries that they were using to record their fight online. And we know that you may not be able to find it online, especially if you’re working internationally. However, we do know that this is pretty strong first step towards a transparent and equitable future
1:20:45
Future interest rates. So in other words, what justice of color to be our reality. Thank you for listening and a special shout out to our mentors and precious star Samantha Wigglesworth shout out to her she was. And of course she’ll be and Susanna for the entire team in criminal justice. And now for our last two present data farmer.
1:21:33
Opportunities bank’s vice president, which we have come up with is a kind of country solution. So, surely. Yes. Yeah. So, basically here policies which we have supporting these realizations we have been using these policies. So, this is the introduction. And he’s studying in university as a professional law student out of California, and myself. So, coming to problems. We have come up with theories
1:22:59
Which we have, we have come up with the risk assessment tools. What is the usage of to reduce the cost and increase the fitness? So when we go through the actual presentation,
1:23:13
Other graphs. We understand the basic idea was not, because if you see the graph. Cash. Cash bonds to get paid. So, we have come up with these policies. Out of these five policies,
1:23:42
Lack of information, heavily on angels. They know exactly what the concept is. So when is lack of understanding and helping Donald? So what we have proposed as a policy is like, first of all first of all, it is should be perfectly suited NSA. It should be available publicly so that people will understand. Because Bunnicula is getting better understanding of defending any case, as well as it will be helpful for three days. Supporting become a state by country device, the different tools, corresponding companies are using, we have come up with a split second policy we come up with basically even though summary, it is available. There is lack of
1:25:11
Understanding. It is not open source. So, only some question content so but because it is publicly very famous. So, this is the standard by the people, have some consultant so that PPT is having an update available that can be easily used for medical. So, even so easily NGOs, They can focus entirely on this system.
1:26:11
So that transparency is the issue even third one is lack of contestable data. So, most of the states, which is during the Super Bowl, we have mentioned, is what’s the issue is not comprehensive enough to understand because various different companies have issues. Different states have different policies, you should be good to have one generic to make available to come up with. My name is Leticia is here. We should have different educational assessment. And this policy issue for injury. Today, consideration of PVS health education household is so many vendors out there. So these are the kind of data, so whenever there is the
1:27:43
Issue is having so far that we should be keeping practice. So what we suggested and how to eliminate these kinds of proposals, how to understand the central policy indicators. So that that’s one. And the last one is solutions so people really
1:28:38
Pressing issue to people. So what we have said is, use the. The least we can, we have sent that text messages to people attending group secrets and calling things. So we have started responding, have to go to states, dance, as we have different special dance courses. Last week, how we can get things and our
1:30:01
Team better farmers job everyone. Guys have really blown it out of the water. And now
1:30:16
You’re reading patiently and eagerly. Here our judges. So let me introduce our judges. We had three judges with diverse skills and backgrounds. Please welcome Joe Morocco, of corporate compliance and ethics, data protection officer, and also advisor. Phoebe Bajwa, she’s a deputy attorney general at California Department of Justice at Patrick, who is the director of ethics that’s emerging. At California Polytechnic State University and Joe PC Patrick wouldn’t be here tonight, and I want to welcome Joe.
1:31:03
Thanks for all of us really appreciate everyone’s submissions jumped by difficult, as we went through various sessions, a couple of times last week, ourselves. It’s quite difficult to come together to make a final decision for the awards. Most congratulations simply work really interesting to watch today. The reason we were concerned at the time but to see everyone’s level of commitment. The heart that was behind all of these projects is really nice to see.
1:31:40
Thank you so much for what Joe said it was, it was quite difficult we spent quite a bit of time doing everything, and I do appreciate presentations and gave logical perspective on what he’s trying to accomplish, and it was very interesting solutions that he came up with across background. Thanks, chief for that. We’re sitting in cities office. What is that?
1:32:16
Yeah, it’s from Dropbox, it’s a JPEG if anybody wants to find it. I have no idea gala background Zoom meeting. We’re gonna say Europe last year and some flowers. Shower decorations. New York City.
1:32:52
All right. So what I wanted to say is thank you all again for our participants and certificates for participating, as well as the chance to participate in this program for our winning team. We have five tracks, and that was visualization model that was an AI model, they could come up with an AI model an application or a hybrid for you we have had some great games, some of these categories, not all of the gadget categories got an entry, and I would now like to YPB to share their thoughts on the overall submissions, as well as specific feedback for each. Yeah I think so I’ll go first. And I’ll turn it over to Joe. Okay, so there was six teams and the three of us we divided up their three. So the first one that I’m going to discuss is the eight criminal justice policy. Overall, this was a very interesting area and comparative molecule to do that. We really liked that, this particular topic was trying to hold people accountable, regulators, accountable, and give employers, a tool for how they can invest the data and come to come to salute and publicly accountable I don’t come solutions that would stop the abuses of the system. Without the international crime dispersion tool that they come up with was quite was quite novel. And it was just a great concept to educate hold accountable, clear data sources and good references. The next few that I was gonna get a little bit of feedback on was the five that’s gavel 2.0. What we really liked was really clear. The judges, and there was a really great definition of the problem. And I think that that was what, that’s what some of the issues come up with is there’s so the topic is so big, what you’re trying to get to. And this is very clear, we are giving this particular tool to new judges, and this is how we are going to help you. And I really like the fact that the stress the importance of using AI as a tool for judges to review and make a decision, versus how I do the job for them but I like to get that perfect accountability to the judges, which is you can’t just take a number that’s that. Look at what’s all the factors and the comparison with. It really helps when judges are going through the 100 pieces a day, and having to make quick decisions. And then the last I looked at was a crown justice, seven of the Reformers reporting really highlighted the problem of lack of transparency and the risk assessment tools, that’s what really boils down to is that there’s no transparency, there’s going to look at what evidence is being used to get the clients right, and we really like the policy of moving towards an open source code, because we are dealing with, even people with people’s liberties here right so we should not hide behind regulations to keep it separate. We thought it was a very nice data visualization, and we’d like to have focused on education transparency and how these tools work.
1:36:41
I’ll turn it over to Joe. Again, pretty much. Three more. I think it’s important to present themselves here today. It’s actually a lot more necessarily present. So, starting maybe with with racial bonds. Again, you said on their website. It was quite compelling really crisp videos that went with that as well, and they were in the visualization trying to have that to your message. Presentation stay focused. And just focused on stands behind. And I really recommend that video if you haven’t seen it, really crisp message. Targeting I’m kind of making this area mentioned COVID contract pricing. I think you’re still here, but again this is a video based presentation today, obviously, we saw some of the background video that went with that as far as a huge focus on what I was actually quite surprised at some of the statistics that they came up with, there are more reasons that it was valuable that they do find them and present them. Again this is just reference memory. This was the need for alternative communication for communities that are underrepresented particularly in terms of phones and on smartphones etc some really interesting projects. The last one I mentioned very briefly is the status quo project just presented briefly on the screen and one other web based on the chat, chat bot actually worked it wasn’t just
1:38:29
this. It was really good, and also was supplemental video presentation criteria we were interested in are generally interested in the applicability of the selection impact potential that we could have, and also presentation. So, again, we did see a five to 10 minute version of each of these. Definitely returns anyone. Thank you judges for taking the time to not only review but also providing feedback, going forward, this is similar to being able to submit. We’re not able to submit. So thank you. And now I would like to invite the judges to announce a winner should be announced, announcing Buddhism the first three categories. So in the policy track. The winner was a team. One, they came up with a really great, great idea, great thoughts on how to solve this issue with discriminatory practices. We really like the idea of using comparative law to be that decision. So, the second award is the hype in the hybrid category. The winner in that category is the animal justice Team Five gavel 2.0 Get the bump the last the last visualization, a couple of entries there. The Windows API COVID contact tracing team. I think everyone agrees that they had amazing visual graphics, they really put a lot of work into that, and it was really, it was eye opening actually presented in that way. So congratulations. And now I’m going to turn it over to Shilpi and Joe Bergman and announce the overall winner, grand prize winner. So I want to quickly show you what is the what is the grant choice. So that is given to one grand prize winner will be awarded books. What am I doing sponsors. And as before, I want to jump to analysis. The other one is a time prize winner. The judges Choice Award is also sponsored by one of our community so that, for any bamboo solution directly has members from the impacted minority or marginalized communities. This could be but not limited to VIP USC LGBTQ. So Joe, I invite you to announce the winner. This was particularly difficult. There were specific
1:42:11
Targets deliverable areas that people chose but that’s trying to choose from very different types of entries. The entries were very much software based code and in terms of the temporary and others were policy based. Obviously it’s difficult for us to try to look at them, and obviously come up with an overall winner began trying to come back with our general matrix of potential applicability. And obviously the presentation aspect as well. So we do have two today one grant winner overall and the other results. The judges Choice. Choice in particular, it was, It was very difficult in fact we disagreed. So just be very honest. We’re at three different teams that are listed, and then ultimately we have to make a final decision at the end, so she’ll be any, any preference between choice, the grand winner of this year’s Hackathon was AI criminal justice gala 2.0
1:43:28
to join us again, obviously, you know, that was a wedding that actually works, concepts, but the concept of working concept compares complex results with a biased reality. The working prototype presentation of likeability potential, it’s ready to launch massive impacts were released in the right way so congratulations to the team. Thank you for the judge’s choice,
1:44:13
as I said we were three shortlisted were reformers racial team status quo, and reformers and racial bonds in a similar topic area, both really interesting budgets, ultimately team status quo has been chosen as the judges choice for this year,
1:44:42
Incredible projects that again. It’s already working. It also came with a video presentation focused on hate crime reporting and analysis some really interesting research.
1:44:58
For the other two that were shortlisted I promise it was, it was a hair’s breadth, so it’s a really great job as well. And unfortunately we had to choose one way cool. All right, so now I’m going to thank you Joe. Thank you. Thank you Patrick. Decision making decision making process. And we’ve all seen the
1:45:34
Solution at least a startup something difficult to choose from. Now I’d like to invite, talking about the books that she wants to give. As for the.
1:45:57
Thank you. I think you can put up instead of down seeing my face. Actually, I have a lot of difficulty trying to take like good collection of books for people to choose from. I am biased towards psychology. I come from a humanities major, that’s not my view is more important than just you know, giving you advice books on AI and how to do machine learning, deep learning. There’s plenty of them around, and some of them are even pretty so I think I skipped all of them. And when these three categories. I actually believe you should start over, artificial intelligence. So these are my singularity is the sentience. Human compatible and designing artificial intelligence by super jumping. Also here in the audience. More than that, I feel like if you understand the impact that you’re going to have before you do anything that is that would be a great introduction of AI ethics in the middle. This is very hard to pick up on. But I have a great introduction. And finally, human cognition overview is something new and whenever I hear deep neural net networks are designed after the brain and some of them are. So overexcited advanced it is and how so many things. But then there are others also who are not impressed with it. I just think you guys need to understand how the human brain actually works, and how we are, and how our decisions and how we live our life are based on the society and conditions. So these three books give you a good overview. I believe on the team members’ case to book from these three categories congratulations to all the judges. Choice Award has been awarded by the judge who’s also in our yard tonight. I know she said, to not invite me to never do that like to say a few words. This is amazing. This is amazing. I love the creativity of any hackathon and this is mind blowing because the clarity of the insights and the passion of the teams and how they work together. And every one of them. Outcome of how they’re going to change the world. This is not just another product is just going so I feel so blessed. I was surprised Susanna was asking me about what are the AI books that I like and I
1:49:29
Didn’t know what she was even doing with that. So I’m surprised to see my book show up. So I’m happy to give that to every single participant. Oh wow.
1:49:42
I think it’s just unfair sometimes, you know, like the judges said it’s, it’s very hard and I want everybody, I’m happy to do that. Thank you. Alright, so I want to also tell everyone. Great sessions and I want to add to that, that you guys can continue this activity. But we also have the accelerator solutions, please reach out to us, the winning teams as well as all the teams participated invited to the ethics on the search funding that we are looking into and I’m talking to other people, to see how we can do it. This is a big announcement and it’s not good. Only for your ears and eyes. This is like a preview of what is coming. We are going to make the first open source package. And so what that means is we will be bringing community projects, and community based research, social benefit organizations to partner with
1:51:06
people who are diverse in like human oversight, with expertise, not just in technology and the psychology, and all these different areas marketing and UX AI so we are going to take all of these and we are going to start building solutions that will really change the world. Join me in this journey and your project partner with us. Solutions that I want to leave you. Someone part of the year joining us for this event support. This was a foolish undertaking. I had never to get foolish, smiling. As we had 250 62 actually submitted solutions, positive energy. All right, Thank you, everyone, was awesome.