Wednesday, August 20, 2014

GetSet Uses Natural Language Processing To Reduce College Drop-out Rates

GetSet Uses Natural Language Processing To Reduce College Drop-out Rates

Next Story

Startupbootcamp Launches Barcelona-Based IoT Accelerator


GetSet, a new stealthy US edtech startup that’s aiming to reduce the high college drop-out rate is uncloaking today and revealing its first rollout at Arizona State University, with its 10,000+ freshmen.
First up, in case you’re feeling a spot of deja vu, last week TechCrunch covered a UK startup called Wambiz that’s taking aim at the same problem. Yes, yes, you wait ages for college drop-out reduction startups and then two come along at once. So it goes.
That said, they’re not identical. Wambiz is building an engagement platform cum social network as a better way to reach/engage with students, rather than sending comms via more traditional channels like email and SMS.
While GetSet is taking an algorithmic approach to the drop-out problem, building a natural language processing (NLP) engine that asks students to feed it with data about their college aims and problems which it uses to match students to others who have similar goals/backgrounds or who had the same sort of issues previously and overcame them.
Although the GetSet front end does also include a social network layer where students have profiles and can share content with each other and generally participate in a digital community that’s specifically tied to their college, so there is some overlap in the approach. But engaging young people digitally is inevitably going to involve something social and connected.
Fixing the problem of student orientation in a new and potentially alienating environment is key to the college drop-out problem, argues GetSet CEO and founder Karan Goel, because “social factors” play the biggest role in high US college drop-out rates. He says research has shown that more than half (54%) of college drop-outs are driven by social factors, such as students not feeling like they fit in or not making friends, vs around a third (30%) leaving for financial reasons, and even fewer (16%) for academic reasons.
Goel argues that traditional support channels for students — such as face to face counseling — aren’t working well any more because students are no longer comfortable utilizing these forms of support. They want something faster and more accessible via the channels they are used to: aka their digital devices. “The challenge has changed a lot in the last 50 years,” he says. “Traditionally students would go see their counsellor when they ran into an issue. In today’s age students just don’t go. They don’t reach out to the counsellor. They want something instant.”
The GetSet system gives freshers a social platform that connects them to similar peers — based on things like their shared goals — to help them make friends when they first arrive. And even before day one at college. “We use a community of peers to create instant support,” is how Goel puts it. He stresses that it is instant — it’s not a forum style system where you post a question and have to wait an indeterminate amount of time for a response; the matching is done immediately.
“Whatever we’ve learned about the student, we show them someone just like them who’s run into that same challenge and overcome it,” he says. “We call this vicarious success. Showing you examples of good behaviour and how to overcome challenges or solve problems. And it’s instant. It’s not like you post something on the school network and you wait for people to respond. You just tell us what you want to accomplish, or what your question is and we instantly find a match for you.”
The platform also provides an ongoing support role for students after they have initially settled in by giving them a quick way to find peers who can help them when they are having specific difficulties, beyond the challenge of arriving at a new school. It does this by real-time matching an individual with a problem to someone else at the college who was able to resolve the same sort of issue — again using NLP to achieve real-time matches.
GetSet’s NLP and matching engine is called PeerWisdom. “It’s very powerful because, as you can imagine, the average 18-year-old is much more likely to listen to something from a 19-year-old who’s relevant to them than from an expert who’s much older — even though the expert might have the best answer, they’re more likely to listen to the peer,” Goel adds.
Obviously, the more data this sort of system has the better it gets — so there’s initially likely to be the equivalent of a learning curve as it accrues data from the students that will ultimately be reflected back to them to provide community support for their very specific problems.
“It just gets better over time. All the information the school has to give us is the name, email, cell phone number if they have it… It basically learns about them over time. Let’s say they first question comes up — what’s your inspiration [for going to college] — so you’d answer it and that would be the only thing we would really know about you at that point. And we would match you with somebody at the same university with the same inspiration. And then you answer one more thing and then we keep building a richer and richer profile of you over time,” says Goel.
The startup is leaning on psychology as its underlying basis when it comes to matching criteria. “Our chief scientific advisor, his name is Dr Robert Feldman, he’s the deputy chancellor of the University of Massachusetts Amherst. He’s spent the last 30 years — he’s the leading researcher in student success, [asking] how do you get students to graduate? He’s helped us develop these questions in a way that we get students to open up,” says Goel.
“GetSet facilitates the rapid development of meaningful relationships and sense of connection with other students at the very start of college.  In turn, this significantly raises the likelihood of future college success,” adds Feldman.
Goel’s prior edtech startup, PrepMe, sold to the owner of Blackboard in 2011. He’s been bootstrapping GetMe since founding the startup in October 2012, running a series of experiments and trials since early last year to refine the technology.
On the funding front, GetSet is in the process of closing a seed round of funding — it’s taken in some of this financing already (but not announced it before today) and is continuing to expand the round. Goel says the aim is to close $2.5 million in total.
Investors will include Social+Capital, founder of Braintree, founder of Fieldglass, Chicago Ventures, serial edtech entrepreneur Paul Freedman, and serial entrepreneur and prior investor in PrepMe Howard Tullman, according to Goel, plus some additional unnamed investors.
How exactly does GetSet work? It initially asks the students a handful of questions, such as what they hope to get out of college or their reasons for attending, and uses the initial shared data to power matching — with a view to helping them find others with similar backgrounds or aims who they might be likely to make friends with. As time goes on, students can set more goals or ask the system for help with specific problems.
They can also choose to share the information they submit publicly to the GetSet social network if they want to, but there’s no requirement to share to be sent matches. That means students can ask for help with a specific issue privately, i.e. without going public about it, and still be matched with a relevant peer who may be able to help them.
Matches are presented as one main match and a few  secondary matches. It’s then up to the student to contact the suggested peer if they choose. Students are incentivised to help each other via the system — which gives positive reinforcement in the form of thanks when students help others. “That’s why we deliberately use the term PeerWisdom, because I think a lot of students don’t think of themselves as wise — so it’s this nice surprise to feel like ‘hey I have wisdom, I know something that could help someone else’,” adds Goel.
From its initial trials of the tech, GetSet usage skews towards new students wanting to orient themselves in the environment — falling back to more of a support role after that. “Our current usage shows that students use it pretty heavily when they’re starting, so in the first few weeks, but then after that they’re coming back maybe 15 minutes, or 20 minutes, just to put in a new question or a new challenge, or something they want to do that they don’t know how to do. So the usage, they’re coming back, but they’re not using it as intensely — and that’s fine, because that’s really the point,” he says, adding: “We’ve got them embedded in the community, they feel like they belong, they’ve taken some of those relationships and are now engaging with those people offline or through Facebook or through Twitter — but we helped them get introduced to those people. And really what the school is measuring us on is did more students graduate, did more students stick around.”
Goel says the early indications for GetSet’s ability to improve drop-out rates look good, with the results from three months of trials involving a few hundred students — who self-selected to try out the product — showing users being 5% to 10% more likely to “stick around”, as he puts it. That’s pretty early data so it will be interesting to see what kind of success rates GetSet can achieve with far more substantial usage as it rolls out across universities — today’s ASU rollout being its first sizable deployment.
“We’ve got another big college in Southern California that’s launching the week after [ASU], so that’s our next big launch,” adds Goel. “And then we’ve got a whole set of schools. We’re trying to get it right — we’re only launching one school a week or every two weeks right now. We’re not going to try to sign up lots of schools but we have a very deep pipeline of traditional universities, online colleges.
“The drop-out problem is really across the board. I think people traditionally think of it as something that only happens at lower tier universities or online universities but really with the exception of maybe the top 50 or 100 universities in the US, everyone else has a big drop-out problem.”
In addition to targeted help to cut drop-out rates, Goel says the platform can help universities to quickly identify large-scale problems that are affecting the student body — giving them a chance to intervene early — such as, for example, a Chicago university that the startup has been working with being able to identify a parking problem that was making it tough for students to get to their classes via comments made on its platform.
“We provide some pretty deep business intelligence to the school. So the university has a word cloud so they can see what are emerging issues that are occurring on campus. This is really important for them because traditionally universities will only know if something is wrong if the student comes in to see a counsellor. Which today almost never happens,” adds Goel.
The GetSet business model is equivalent to a SaaS one, with the universities paying the startup so they can offer the platform as a free service to their students.

Total Pageviews

GOOGLE ANALYTICS

Blog Archive