COVID-19 has provoked a lot of conversations around data privacy framed as, “what if we erode privacy rights in exchange for the greater good of protecting human life and reopening the economy?,” says Todd Mostak, CEO of OmniSci in an exclusive conversation with Geospatial World.
How can location intelligence help optimize resource allocation at a time of pandemic when virtually all supply lines are disrupted?
Supply lines were disrupted with the advent of COVID-19. First and foremost being able to quickly examine, explore and interrogate data provides information about which factories are operational and which of their suppliers are online. As we get into the local market, that data can also provide insight into how things are being distributed in real time because there is a greater need for manual intervention simply because things could be disrupted at a moment’s notice. Due to COVID-19, many areas were shut down and anticipated deliveries were disrupted, so that requires a lot more human intervention. That is where our system excels by allowing human analysts, subject matter experts and decision makers to get the context they need to make decisions under changing conditions during black swan events like COVID-19.
How aggregated mobile location data can prove tracking and tracing of the virus?
We work with a number of partners in this space like Skyhook, SafeGraph and X-Mode to give us the ability to look at human movement patterns and correlate it with the spread of the virus. The first thing that we can do is build a predictive model of how fast we expect the virus to spread in a certain community to help plan for access to ventilators, hospital bed capacity, testing and reopening policies. Using a very coarse population model one can only reach a certain level of accuracy, but what we’ve found is that by incorporating more granular metrics of number of miles traveled per day, number of buildings visited and the aggregate population density of places people are visiting we can certainly correlate it with the spread of the virus. You can even bring together other factors like correlation to weather or demographics to inform the predictive model. The second thing we can do is get deeper into use cases of hotspot analysis and contact tracing. Hotspot analysis is particularly good for this sort of location data. If you are looking at country, state or municipality level data you won’t have 100 percent coverage, but if you know certain people are getting sick or visiting emergency rooms etc., you can basically back track and find if there are some points of interest or particular locations that stand out from the background. If certain restaurants, shopping malls or grocery stores are visited by an inordinate number of people who ultimately got sick, you can see that there may be some sort of unsafe condition happening there or perhaps it’s being visited by someone who is a “super spreader.” Lastly, in regard to contact tracing and patient zeroes, there’s been a lot of talk about how perhaps folks who were in New York City traveled to Florida or other places in the country which led to the spread of the virus there. You can use this data to help determine aggregate population movements between hotspots or infected areas and places that aren’t hotspots yet to predict flareups in these secondary locations.
How can use of artificial intelligence tools on location data help health organizations and local authorities for fighting the disease, and further disease modeling for prediction and prevention?
When we say AI, it can be as simple as a statistical model but at scale. One of the things that we have done a great amount of work on is integrating data science, or converging data science workloads and machine learning workloads, into more traditional analytics workloads. First off you may want to visually analyze all this data to make sure that you understand what is going on and have some context, because maybe some of the data is anomalous. When you are getting reported cases of COVID, you might see spikes or aberrations that pertain to how the data is being reported. Somebody who is a subject matter expert can quickly spot those and ideally correct for those, but the challenge is then being able to seamlessly transition into something as simple as a regression, a more complicated XGBoost or something even more complex like a neural network. Being able to do that at scale on the same platform has been a boon for our customers — particularly for the predictive aspect when it comes to being able to forecast the number of ICU beds or testing kits in a supply chain, for example — because everything is a moving target right now.
Pandemic has also seen the value and benefits of location data and intelligence tools going up like never before, what would be the impact of COVID-19 on location and AI industry?
It is causing people to think more spatially, and not only about problems related to COVID-19, but in relation to more general business practices. As people have seen the power of geospatial analytics in general to provide intelligence that they wouldn’t necessarily get from more traditional BI tools, they start thinking of other use cases for our technology. We have actually looked at a supply chain problem with a federal government that quickly translated into a broader discussion about ensuring that there will be enough food, ventilators and other essential items in any disaster moving forward. I think people immediately grasp the correlations. Even in the retail or CPG sectors, by understanding aggregate movement patterns in a way that protects privacy you can get a greater sense of demand for your product and the best channels or stores to sell a given good. Even within a store you can start to think more spatially about shelf space, like where a certain brand of deodorant will sell best based on people’s pattern of movement through a store, and by doing so you’ll learn how to maximize presence and contact with a customer. I think this a conversation starter and once people see the power of analytics with COVID-19 they will start thinking about other corollary use cases outside the context of the outbreak.
Among your user industries where do you see the business falling and which areas do you see the demands going up?
We have seen a lot of interest from the Telco sector, which is already one of our key verticals. COVID-19 is changing everything in terms of demand for data and bandwidth. Maybe there is less demand because people are at home and are on Wi-Fi, or maybe there is more demand because people are surfing or consuming content in ways they wouldn’t be if they were at the office. So, everything is changing, and we see all these Telco companies throwing away their old models and trying to navigate how to not only survive, but actually deliver relevant benefits to their customers. For example, because people are consuming more content, how do companies then deliver discounts and gain market share? How do they ensure that, because people are in different places than they normally are, that they have the network capacity to deliver good service? Telco is one big sector where we have seen a lot of interest in either COVID-19 use cases or adjacent things pertaining to the more general impact on the economy and people’s behavior.
Use of location data for contact tracing for COVID response have been a fresh debate on data privacy, what could be done to balance data privacy along with effectiveness and how do you see data privacy landscape evolving going forward?
COVID-19 has provoked a lot of conversations around data privacy framed as, “what if we erode privacy rights in exchange for the greater good of protecting human life and reopening the economy?” I think that’s a little bit of a false dichotomy or false choice. One of the problems right now is that even if location data is anonymized it still provides contextual information. If you have enough data points and know where someone lives or works, you can probably determine who someone is based on their location data. That understandably concerns people. One of the main issues right now is that people will give out this raw data, and that presents a lot of risks because anyone could look into the data for good or nefarious purposes. OmniSci is working on ways of protecting that data. Because OmniSci can process data server side, we essentially use differential privacy to roll up results based on individual, aggregated data. That way you get the full power of all the demographic data or patterns of movement you might have, but the end user, consumer or decision-maker can never see or get access to individual location data. Everything is rolled up into groups of 20+ people but it still provides the granularity and power needed to make better decisions. This is something I think will be become increasingly important as people are trying to navigate policy decisions.
Also Read: Using spatial data to protect ecosystem
