Monitoring the “dark web” for chatter to predict terrorist activity is expensive and largely the purview of law enforcement. Travel risk management startups, however, are developing technology to help clients more quickly respond to threatening incidents. It’s possible, for example, to issue tsunami alerts following an earthquake. Good intel can highlight areas with a higher likelihood of violence, preparing organizations in case it happens.
The U.S. federal government has determined that enhancing intel gathering and communications in a race against time is an endeavor worthy of taxpayer assistance. TRM provider Stabilitas has won a $225,000 Small Business Innovation Research grant for research and development that could lead to an additional $750,000.
The funding will accelerate existing efforts. “We think we can take the world, in baby steps, from a pre-travel report to something that makes them more situationally aware every day,” said Stabilitas co-founder and CEO Greg Adams. “If people are more informed about what’s going on, they can avoid emergencies. In certain cases, there are warnings out there where we can get people word to take shelter, move to high ground, etc. We want to get as predictive as possible. You can’t predict a lone wolf attack, but you can predict protests.”
Stabilitas takes in data from traditional and social media to look for threats. It’s seeking to apply algorithms and big data concepts to analyze even more data, more quickly and with more accuracy. The information helps clients in security or other corporate departments take action.
Stabilitas is one of a relatively new crop of providers aiming to take TRM to another level through technology. Many have a military background.
The overall goal is improved responsiveness. Adding to the “lone wolf” metaphor, Adams also noted that care must be taken to avoid crying wolf. While highlighting areas of higher risk can be useful, it’s not always. “You can make people numb,” he said.
This is part of the reason why simply pulling in, boiling down and exporting information is not enough.
“A lot of providers call their capabilities artificial intelligence or machine learning or intelligence when really they are just automating the inflow of data,” according to Toby Houchens, CEO and founder of competitor Travel Recon. “What matters is the analytical processes built that can automate and aid analysis of relevant and granular data and therefore intelligence. You can’t avoid the need for human factors, nor should you.”
Adams agreed. “When it comes to machine learning, you need humans in the loop somewhere,” he said. During the Global Business Travel Association convention last week in Denver, Stabilitas chief of data Sean Maday said analysts identify points of interest, skewing the algorithms to focus on particular areas. “Every person on the platform, to some extent, is an analyst,” said Adams. He suggested that a community approach in which clients’ analysis is shared among them is a worthwhile goal.
Houchens emphasized the value of “true ‘crowdsourced’ data which is geocoded out of the box.” He also highlighted the need for “a user experience that stimulates interest and buy-in for intel dissemination.” Travel Recon just released its GoRecon app, using gamification concepts to draw user-submitted security intel. “We are seeing steady growth in our user base” of travelers and locals, according to Houchens.
Stabilitas also is looking to develop crowdsourcing capabilities.
“A version of that is in place,” said Adams. “We can crowdsource right now, but this is some of what we’re researching.”
Also speaking at the GBTA event, Stabilitas co-founder and COO Chris Hurst illustrated the potential for intel from the ground. When he managed risk for a non-profit operating in the Congo, he said, a female colleague transiting from Rwanda was sexually harassed by a border guard. “It was frustrating because we found out that the ex-pats working [there] all knew to avoid this border guard,” said Hurst. “There was a pattern of harm there. Why didn’t someone tell me? Why didn’t my employee know? Now I know, so how would I tell someone else? No analyst would have caught that. It never would have made its way to social media. So patterns of harm are not captured in the old model.”
Maday outlined four kinds of artificial intelligence that he said were “ripe” for use in TRM.
Sentiment analysis: Computationally identifying and categorizing opinions as they are expressed — looking at a tweet, an email, news reports
Usefulness scoring: Training a computer system to read and understand human language, and recognize patterns of speech that have relevancy to safety and security
Semantic topic modeling: Grouping a set of objects with similar objects, for example computers can start to understand that Beijing semantically is related to China, which is different from Japan
Geo-parsing: Identifying and disambiguating places as they are mentioned in text, so computers understand when a location is referenced and bring that location data back to be plotted on a map
Stabilitas offers a tiered licensing model, with a “bulk plan” also available, Adams said. New clients include AXA and Torchstone.
He described the federal grant as a “vote of confidence.”
Additional info: The Small Business Innovation Research grant program is designed to support research and development that both benefits society and can be commercialized. It funds about one-quarter of “all federally supported basic research conducted by America’s colleges and universities.” The Stabilitas grant came from the National Science Foundation, one of 11 federal agencies involved in the SBIR program.
Adams declined to share internal development milestones, but said as part of the grant the company must submit a progress report to the government within six to 12 months. In the meantime, the company is permitted to commercialize any resulting products or services. Grant beneficiaries retain intellectual property ownership.