Safer with data: protecting Pakistan's schools with predictive analytics

A new technology partnership backed by UNICEF aims to help make 1,000 schools safer in Pakistan. Spearheaded by predictive analytics firm Predictify.Me, the Pakistan Safe Schools initiative will introduce simulation software to assess the level of risk preparedness of schools and generate recommendations for school and community safety plans. Predictify.Me CEO Rob Burns explains more.


Pakistan school

"Everyone has the right to education." So states Article 26 of the Universal Declaration of Human Rights, recognising the fundamental role of schools in societal development and the individual's pursuit of prosperity.

Upholding this right is a continual challenge for a variety of reasons. Ensuring children's safety at school is foremost among these challenges, especially in countries suffering from a high incidence of terrorist attacks. In Pakistan, the horrifying Peshawar school massacre in December 2014, during which Taliban-affiliated gunmen killed 145 students and teachers before being neutralised by special forces, has been the deadliest recent example of the dangers faced by children simply exercising their right to learn.

Although the Peshawar attack stands out for its shockingly high death toll, it is indicative of the dangers faced by schoolchildren in Pakistan, which since the 1970s has experienced more attacks on education facilities than any other country in the world. In reaction to these dangers, the Safe Schools Initiative was launched earlier this year, supported by Pakistani Prime Minister Nawaz Sharif, UN special envoy for global education Gordon Brown and the Global Business Coalition for Education.

As the initiative works to secure funding to improve the security of schools, US-based predictive analytics start-up Predictify.Me has offered to provide the tech for an alternative, data-driven approach to safeguarding schools in Pakistan against the risk of explosions. The technology being offered by the company was developed over eight years by chief data scientist Dr Zeeshan-ul-Hassan Usmani, who was born and raised in Pakistan and has become an expert in the simulation of blast waves in open and confined spaces.

We spoke to Predictify.Me's CEO Rob Burns to discuss how the company's predictive tech could help make schools in Pakistan safer, and the search for partners to carry the project forward.

Chris Lo: Has the Pakistan project been a focus for the company from the beginning?

Rob Burns: Well, we have a lot of focuses. We've grown from three people to about 48 people right now. Our belief is that governments around the world, and foundations and non-profits and NGOs, are all putting massive amounts of data online, and in accessible formats. And we think some of the critical issues in business and society can now start to be answered in a far better way using the massive amounts of open data available, and very sophisticated algorithms and techniques that have been refined over eight years by Zeeshan and his team in Pakistan.

Almost all of the [predictive analytics] tools that have been developed in the defence industry are about protecting soldiers. When we look at the existing tools that are out there from a Northrop Grumman or a Lockheed Martin, they are built around 20-35 year-old fit males in protective gear. They're not built for an eight-year-old girl. Therefore the solutions that come out of those tools are fundamentally flawed for a civilian deployment. This is a fundamentally different approach that comes at it from the civilian side instead of soldiers in-theatre.

CL: Was it a complex process to get involved with the UN on this project?

RB: It wasn't really, to be honest with you. Because the tool had been available, the way it worked was one of our advisors had introduced us to Gordon Brown's team. We then went to DC and met with Gordon Brown's team and reviewed the programme. They then reached out to the prime minister's office in Pakistan to validate whether this was something that was worth moving forward with. There was general consensus immediately that it was, so it really wasn't more than two or three meetings to put the concept together.



n the fight against terrorism, defence and security agencies are turning to behaviour prediction software.


CL: Given that your chief data scientist Dr Zeeshan-ul-Hassan Usmani was born and raised in Pakistan, is there a personal component to this project?

RB: It's personal for him and it's personal for me. This is how we started talking to each other, and we talk every single week about the impact we want to have on society. We're going to build commercial products and we will be a profit-generating enterprise, but at the same time both Zeeshan and I have a very deep commitment to social causes. So I would love to see this tool ported into a lot of different applications for the protection of children. We've looked at whether you could use public data to solve for drug dealers, for child abusers, and really start to understand patterns and how to disrupt those patterns.

CL: Could you describe how your SecureSim and Soothsayer software identifies risk factors for schools?

RB: There are two aspects of what the software does. What Soothsayer does is it takes 170-plus geosocial indicators - so events that are occurring; it could be weather, it could be time of the year, availability of raw materials, patterns of violence in other cities and other countries - those are formulated into a risk score for a particular location. That risk score is to identify with a reasonably high degree of probability what the likelihood is of an attack occurring in an area. So that is the preventative measure.

The second side is what happens when an explosion occurs. So let's say you've tried to prevent it, but for whatever reason the attack still occurs. How do you minimise the impact of an explosion? How do you take the physical layout and adapt that to minimise the impact, and then when you understand the impact, how do you triage the event? You have deaths that occur from the initial explosion, but you then have a set of people who just bleed out because there aren't enough ambulances, there aren't enough nurses and doctors that are around the scene and prepared to handle it.

Eliminating secondary deaths is a big deal. We believe that, from the work we've done, with 72% accuracy we can predict the area where an attack is going to occur. We're refining that down to the school level. Once you get to the injury patterns, they are predicted with over 90% accuracy.

CL: The tool focuses on the threat of explosions; are other threats - like an attack by gunmen - going to be included?

RB: Well, we have talked about refining the tool to include active shooter scenarios, so it's in discussion. To be quite honest with you, we would probably need a partner because this is not going to be a part of the business that we build out directly. If there are companies like G4S that operate in this space every day, we can develop the tool but ultimately you have to deploy it. And that means you have to be able to operate and sometimes secure infrastructure. That's just not something, as a small company, that we're set up to do.

CL: What progress has been made on the project since the announcement in March this year?

RB: We've gone out and started modelling schools on our own. That's not what we want to do long-term; we would much rather find an in-country partner that can take control of it. We're working with the United Nations and Gordon Brown's office to identify the ideal partner. As you can imagine, that's complicated. We have requirements on our side before we're going to transfer technology to anybody, and they have requirements on their side of the type of partner that they need.

"We've gone out and started modelling schools on our own. That's not what we want to do long-term; we would much rather find an in-country partner that can take control of it."

So the process has started, but we are not at the speed which we would like to be. We do need to find the appropriate in-country partners, and that process is ongoing. If there are people who believe they can be good in-country partners, they should absolutely reach out to us. If they're in that business, it can be, in the long-run, very lucrative for them. Our goal is more humanitarian. This is not supposed to be a gold mine of money for us, but the reality is the global risk situation probably warrants one of the big players getting far more active in this.

CL: What kinds of new security measures might your recommendations prompt in at-risk schools?

RB: I'll give a few small examples. One is we believe you could put, in some cases, a T-shaped brick wall at certain points that would block a car. It's a fairly inexpensive wall and if it's positioned in the right place could significantly reduce casualties.

The second thing that we've seen is that some of the larger schools have security metal detectors at the entrance, which means you get this huge consolidation of kids outside of the protective barrier. So you have a specific time of day that you know all the kids are going to concentrate outside of the security perimeter. That's a problem.

Setting up your response teams is another critical factor. We can model based on past explosions - not only the likelihood of an attack but the likelihood of a specific type of attack - almost down to a triage list of what the injury patterns are going to be and what supplies are going to be required. That requires emergency response personnel to do a significant amount of training and preparation.

CL: Are you confident that there will be enough support from the Pakistani government and international agencies to ensure that these recommendations are acted upon?

RB: I think it's an open question, but I do think there is more than sufficient momentum, and I think the question is, can we pull all the pieces together in an efficient manner? I do think there's momentum, I do think there's sufficient interest to move it forward, and I do think there's sufficient funding to put behind a partner. It's just a matter of putting the puzzle pieces together.

CL: Is there an opportunity to expand this programme to other schools around the world?

RB: Oh, I'd love to take this thing all the way, to active shooter in the US, drug dealers, exactly where the company started. We may get there in a completely circuitous route, it may not be in the way we exactly envisioned it, but if we could use this tool to protect as many children in the world in as many ways as possible, we would be ecstatic with that.