Augmenting Public Safety Through AI and Machine Learning

Augmenting Public Safety Through AI and Machine Learning

In this demo day, we took a close look at the tremendous potential AI offers for making communities safer, by helping to reduce, prevent, and respond to crimes. When it comes to public safety, it is often critical to act quickly. AI technologies can supplement the work of people, taking on monotonous and time-consuming tasks that would be impossible for humans to do effectively. Natural language processing can read and analyze public communications and news reports to detect potential problem areas and get-ahead of violence. Of course, this work must be done responsibly and ethically.

Sharing her perspective on the impact that AI can have in keeping people safe was an expert in the field, ElsaMarie D’Silva, the Founder & CEO of the Red Dot Foundation. The Red Dot Foundation’s award-winning platform Safecity crowdsources personal experiences of sexual violence and abuse in public spaces. ElsaMarie is listed as one of BBC Hindi’s 100 Women, and her work has been recognized by numerous UN organizations and the SDG Action Festival.

To go a little deeper into the application of AI for public safety, we shared Omdena projects that took innovative approaches to make communities safer.

 

Case Study 1: Preventing sexual harassment through a safe-path finder algorithm

UN Women states that 1 in 3 women face some kind of sexual assault at least once in their lifetime.”

With the first case study, the Omdena team drew upon Safecity’s crowdsourced data about sexual harassment in public spaces and leveraged open-source data to build heatmaps and calculate safe routes through major cities in India. Part of the solution is a sexual harassment category classifier with 93 percent accuracy and several models that predict places with a high risk of sexual harassment incidents to suggest safe routes.

 

AI Sexual Harassment

 

 

You can learn more about this and related projects here:

 

Case Study 2: Understanding gang violence patterns and actors through Twitter analysis

Our team worked in partnership with Voice 4 Impact, an award-winning NGO whose solution to violence in our communities addresses the questions people worldwide are asking: “How do we keep missing the signs?”

The Omdena team made use of natural language processing techniques — AI techniques that analyze text to understand what is being communicated. Machine learning algorithms were used to understand gang language and AI models built to detect violent messages on Twitter, without profiling. The aim is to predict and ultimately prevent, gang violence.

 

AI Gang Violence

 

You can learn more about this and related projects here:

 

Case Study 3: Analyzing Domestic Violence through Natural Language Processing (NLP)

Finally, we presented Omdena’s work to uncover domestic violence in India hidden due to COVID lockdowns. This work is part of a project with the award-winning Red Dot Foundation and Omdena’s collaborative platform to build solutions to better understand domestic violence and online harassment patterns during COVID-19. The project used natural language processing techniques with social media, government reports, and other text content to create a dataset with which Safecity could mobilize local efforts to protect and support domestic violence victims.

 

 

AI Domestic Violence

 

 

You can learn more about this and related projects here:

 

 

 

 

Host an AI project with us.

 

Finding Answers to the Student Debt Crisis Through Data Science

Finding Answers to the Student Debt Crisis Through Data Science

By Galina Naydenova
 
 

When thinking about student life, we tend to hear the positive sides of the argument — the build-up of knowledge, the thrill of research, the excitement of campus life, and the improved employment opportunities. There is, however, a huge cost in the shape of the student debt crisis, which remains for years after graduation, causes stress, delays important life milestones, and does not necessarily come with improved opportunities and quality of life.

As one of the task leaders, I am going to walk you through the steps we took in solving this challenge. It is less about the data and results, more about what you do with them.

 

Step 1. Setting the scope

Is there anything about student debt that we do not know about?

Omdena’s method of branching out into different task groups allows the freedom to cover many sides of a problem. The content of the task brings a team together — an idea is proposed and then voted into a task, with team members joining in, setting the agenda, and collectively taking decisions. As somebody with years of experience in the Higher Education (HE) sector in the UK, I felt that I am familiar with the subject matter and proposed the scope of our task to be the impact of Student Debt Crisis on different demographic groups (namely: female students, ethnic minority students, and first-generation students) and finding an AI solution for this student debt crisis. I chose this for the following reasons:

  • The gap in participation and attainment between ethnic minority groups and the rest is a well-documented problem, and recent events have made it even more relevant.
  • In the UK and elsewhere, there are questions about the necessity of taking up debt and the value of having a degree.
  • There is evidence that student debt disproportionately affects certain demographic groups, disadvantaging the very groups that higher education was supposed to benefit the most.
  • All of the above, if unaddressed, may lead to a drop in participation of disadvantaged groups in HE, reversing years of progress.

So we set our task goal to explore how student debt affects the different demographic groups, why, and what can be done about it through the multitude of data available.

 

Step 2. Setting the methodology

Keep it simple.

Being a community of ML and AI practitioners, the suggestions that came were regarding modeling, clustering sentiment analysis, web scraping, but there were also voices for the more traditional data analysis for this Student Debt Crisis Challenge. We ended up choosing the latter because:

  • Such an analysis would allow us to compare different groups, which was the aim of our task.
  • When applying machine learning approaches, it is easy to miss out on the detail. The volume of data hides the fact that there are different forces pulling in different directions.
  • Higher Education is a regulated sector and there is a large volume of data from multiple sources.
  • Our data is not coming from a single source, and not all of it is on an individual level, which makes modeling difficult.

 

Step 3. Getting to know the problem

Mapping the Student Debt Crisis journey.

With so many pieces of data, there is a danger of ending up with a collection of separate pieces of data, each one interesting in itself, but disjointed.

Having a framework beforehand would allow us to focus when researching sources, and also overcome one of the problems often found when applying machine learning –the ‘now what’ moment, when we see the data but cannot say what it means.

So, we created a map of the student loan journey, along with the questions whose answers we are aiming to find in data.

  • Loan decision: Why do students have to borrow?
  • Loan dimensions — How much are they borrowing and what are they getting for their loan?
  • Loan and studying — What happens at the different study milestones?
  • Loan and employment — What employment and earnings can be expected?
  • Loan effects — Effect on mental health and HE value perception

To answer all these questions, we sought data for different demographic groups, as per our task goal.

 

Step 4. Getting the data.

Aim for variety.

Luckily, as is the case with other regulated sectors, there was no lack of data. The mixture of statutory, financial, survey, trend, and official stats data, apart from allowing us to cross-validate findings, allowed us to build a rich, multifaceted picture. At one point we had nearly 50 pieces of data in our shortlist to investigate — a challenging task even in itself. Some sources, like the College Scorecard dataset, were massive, and we had to apply a fair amount of manipulation and standardization. Some of the sources came with long definition lists, which took a while to unpick. The agile nature of the project allowed us to sneak in some data on the impact of COVID-19 at the last minute, making it even more relevant.

 

Datasets used for analysis

 

Step 5. The recommendations

The last mile

I am going through these steps together. Often analysis relies on data speaking for itself and falls short of interpreting and giving recommendations. However, having a framework from the beginning, breaking down the big questions into small ones, and linking the questions to data, we can see areas that can be acted upon.

For example, there is no silver bullet for tackling drop-out, but small steps can be taken, to challenge institutions about the completion gap between the different demographic groups, to facilitate non-punishing transfers, to set work/study standards, and to use AI and predictive analytics in order to improve degree completion during this Student Debt Crisis.

 

 

 

What can be done?

 

1. Why do students have to borrow?

Some of the answers:

  • Low income and first-generation students are more likely to come from ethnic minorities
  • Independent students are more likely to be black than dependent students
  • Private for-profit institutions have a higher share of ethnic minorities, female and first-generation students
  • Undergraduates from ethnic minorities tend to have lower financial literacy

In a nutshell: Ethnic minority students borrow more because they cannot rely on their families, they opt for more expensive institutions, and, because of lower financial literacy, may end up borrowing at disadvantageous terms.

What can be done about it?

Increase general financial literacy, encourage planning in advance, establish peer advice network.

 

2. How much are they borrowing and what are they getting their loan for?

Some of the answers:

  • Minority ethnicity students more likely to be in debt, and have higher levels of debt.
  • Minority ethnicity students are more likely to attend two-year institutions (Hispanic students tend to favor associate degrees) and multiple institutions.
  • Private for-profit institutions have a higher share of ethnic minorities, female and first-generation students.
  • Female, black and Hispanic students miss out on STEM subjects.
  • African American households have the highest percentage of unpaid student loans.

In short: Having higher levels of debt does not necessarily translate into studying attractive subjects or landing at well-performing institutions.

What can be done about it?

Monitor and publicize institution data, advise on beneficial loan decisions, tackle the differences in STEM subject take-up.

 

3. What happens at the different study milestones?

Some of the answers:

  • Private for-profit institutions, which tend to attract female, ethnic minority, and first-generation students, have low completion rates.
  • Black students are relatively more likely to drop out or to transfer.
  • One-third of black students are working over 20 hours a week while studying.

Summing up: Ethnic minority students are more likely to fall at the very first hurdle — that of degree completion, and having to work during study further jeopardizes their chances of completion.

What can be done about it?

Challenge institutions over retention rates and/or gap, deploy predictive analytics solutions to reduce dropout, advise cap on work hours, ensure financially fair transfers

 

4. What employment and earnings can be expected?

Some of the answers:

  • Men are over-represented in STEM and business careers.
  • Men’s earning for degree and advanced degree holders rise faster than women.
  • For women with advanced degrees, the wage gap is wider.
  • The share of associate degree holders is larger in the industries highly exposed due to COVID-19.

In short: The gender wage gap persists and is even more pronounced for advanced degree holders, which risks making entire industries short of women leaders. COVID-19 disproportionately affects holders of associate degrees, tend to be favored by Hispanic students.

What can be done about it?

Support women in STEM and be clear on employment reality for advanced degree holders. Use the corona-virus crisis to retrain and up-skill vulnerable groups, including associate degree holders.

 

5. What is the effect on mental health and the value perception of Higher Education?

Some of the answers:

  • The majority of students from for-profit institutions and black students display high or very high levels of stress from education-related debt.
  • Women report higher levels of ill mental health than men in all ethnicities.
  • Degree holders from for-profit institutions are skeptical about their degree worth.

In summary: Students from for-profit institutions (mostly female, ethnic minority, and first-generation students) display higher degrees of stress and are clearly disillusioned by their higher education experience.

What can be done about it?

Make institution performance and loan conditions transparent, invest in debt counseling, emergency support, and relief, promote an unbiased view of the value of higher education and alternatives to a degree.

 

Conclusion

Even with a problem that has been addressed extensively, collective wisdom from an autonomous and diverse group can bring new insights. Contrary to the popular belief, lack of data was not the problem — it was scoping and breaking down the problem in the context of too much data, which is where collaboration and diverse thinking really helped. Yes, a self-organized international community presented challenges in agreeing times for meeting, brainstorming and feedback, but the colorful side of collaboration — the assorted and inconsistent graphs, as many styles as team members, was a welcome difference from the monotonous slide templates from our day jobs. In the end, there was a positive feeling that we have done our small bit to help to understand this big problem better and that it was a time well spent getting to know and learning from each other.

 

 

More about Omdena

Omdena is the collaborative platform to build innovative, ethical, and efficient AI and Data Science solutions to real-world problems. 

 
Four Powerful Tips for Working on an Omdena Real-World AI Project

Four Powerful Tips for Working on an Omdena Real-World AI Project

 

I’m about to finish up my first Omdena AI project Challenge (Mars Omdena), and I am happy to report to everyone that it has been an incredibly positive experience, full of learning, discovery, and wonder. Working on any Omdena AI Projects is a unique experience in of itself, and as such, you can never really be ready for it. Nonetheless, below are some tips I have learned from this great experience I wished I had known before starting:

 

Tip 1: Status Calls Are The Heartbeat of these AI Projects

 

First of all, the number one thing that amazed me the most about the Mars Omdena challenge is how the balance between chaos and order turns out to be crucial for making advances on the problem.

By chaos, I mean that there are 30 individuals all with their own ideas and theories and all these ideas get mixed around serendipitously. The freedom of this unstructured approach allows for creativity and initiative-taking, and it ultimately means that the best ideas win out in the end.

However, life is all about balance, and these AI projects do require some structure for these creative ideas to crystallize. This is where the status calls come into play, where all the teams present their progress to one another.

Whatever you do, do not allow yourself to miss these meetings. They create pressure to deliver results, which turn fanciful theories into concrete progress. Also during the meetings, you will learn other approaches that can help you with your own. The meetings are very focused on question answering as well, so ask as many as you can!

To summarize this point, we can say that the weekly meetings, if attended religiously, will be the driving force into turning your ideas into real-world results. Set a weekly time and, whatever you do, stick to it!

 

Tip 2: Read around the topic

 

Photo by Raj Eiamworakul on Unsplash

Photo by Raj Eiamworakul on Unsplash

 

 

One of the coolest things about Omdena projects is that they deal with all sorts of different fields and topics. Hunger reduction, fighting PTSD, segmenting trees, or discovering life on Mars, just to name a few. However, it is highly likely that when you first work on a project, you will not be an expert in that field.

One key thing that really helped us to tackle our Mars project was using the first couple of weeks just to get accustomed to the jargon and technical vocabulary related to Space Probes and Interplanetary exploration. We needed to understand words like Technosignature and the difference between a landing site and a crash site, as well as becoming familiarized with the industry-specific JP2 file format because our raw data was in that format. Furthermore, we had to brush up on the history and context of space probes to understand the problem better, we had to understand how the previous Mars missions had gone and what the HiRISE satellite actually was and how it worked (because that´s where we would be getting all our data from).

All in all, when you start learning about a new field, there is always specific technical vocabulary that will trip you up at the beginning and that can cause you and your team confusion.

In the first weeks of this AI project(s), I recommend you spent half your time researching, learning, and familiarization yourself with that industry rather than just diving straight into some algorithm optimization. Trust me, this will make life a lot easier later on.

 

Tip 3: Omdena is a do-ocracy

 

Photo by George Pagan III on Unsplash

Photo by George Pagan III on Unsplash

 

Omdena projects like the Mars one we worked on have a strict flat hierarchy. This means that there is little ordering around and nobody is going to tell you what you are supposed to do. People naturally self-organize into groups where they do what they are the best at or what they are the most curious about.

This mode of operating has a name, it´s called a “Do-ocracy” (a play on the world Democracy). During the project, if someone has an idea, we are not going to vote on who should carry out that task or take on that role. The first person who states that they will do the task is entitled to do it. If there are several people, then they should share the role. Simple as.

Responsibilities are attached to people who do the work, rather than elected or selected individuals. For many, this way of working is pretty alien, but you will learn to embrace it and make good use of it. It becomes very empowering very quickly. And it is key to Omdena´s flexibility on which it thrives.

 

Tip 4: Ask for Help

 

Photo by Tim Marshall on Unsplash

Photo by Tim Marshall on Unsplash

 

 

Nobody knows everything. And Omdena prides itself on having a large group of people from a variety of backgrounds. Your specialty might be someone’s weakness. And vice versa.

One of the reasons why I believe Machine Learning projects excite so many people is because they are truly multidisciplinary challenges. If you were doing this challenge on your own, you would need to use high-level mathematics, be able to code proficiently, understand cognitive human behavior, be an expert in data scraping and have some scientific/technical knowledge of the task at hand.

Of course, that´s almost impossible for a mere mortal, but a team of 20–30 people can cover for all those needs. However, the only way that your skills can be complemented by the mass brain is by asking for help.

Without a shadow of a doubt, you will reach a point in the challenge where you ́ll feel overwhelmed and completely unsure of how to progress. This is normal. This how we learn.

In such cases, ask for help! Make a general post to all participants explaining your issue, or directly contact a participant who you know has strong skills in what you need. This is the only way that these AI projects by omdena can progress well. And you will learn so much from the answer of the other members.

 

So never be afraid of calling out for help, it’s what you are expected to do.

 

More About Omdena

Omdena is an innovation platform for building AI solutions to real-world problems through the power of bottom-up collaboration.

 

Stay in touch via our newsletter.

Be notified (a few times a month) about top-notch articles, new real-world projects, and events with our community of changemakers.

Sign up here