A New Zealand university is enlisting the aid of an AI student concierge to increase recruitment. Known as NINA, the bot appears around Auckland University of Technology’s (AUT) mega-screens, its presence expanded through scan-coded advertising outside the campus.
But NINA is only one facet of the university’s tech-driven Business Intelligence solutions that enhance student experience and make better use of university resources.
Representing AUT at the 2018 Business Analytics Tech Fest were Tim Davidson, AUT Manager of Business and Intelligence, and Megan Skinner, Manager of Exploration and Analysis. Both played key roles in NINA’s development and the ongoing quest for the ultimate AI student concierge.
Moving into the Emerging Technology Space
The bulk of Davidson’s and Skinner’s work is supporting Business Intelligence across the university administration. But over the past year, says Skinner, the university has been moving into the emerging technology space.
“Five years ago our VC visited Deakin University in Australia and saw the Watson engagement advisor. Watson is a language processing tool that Deakin uses to address general student enquiries.”
Davidson and Skinner were asked to develop a business case and work out the possible cost benefit analysis. Financial resources were limited, so they looked at an existing Microsoft product called Project Oxford. This offered AI functionality and services accessed through APIs.
Enter NINA, the Student Concierge Bot
They came up with a tool, an AI bot, that answers questions like when’s my next exam, when’s my next class, what happens if I don’t hand my assignment in on time?
The bot was named NINA, and six months after her debut, student services approached the AI team. They wanted to use NINA to reduce pressure around graduation.
NINA went live and received about five thousand questions.
“It did a reasonable job,” says Davidson. “We set up a big screen that gave a real-time feed of the questions and answers. So if it was giving strange answers we could hop in and fix them up!”
NINA won her development team the Overall Innovation Award at the 2016 Microsoft Tertiary ICT Innovation Awards for an AI bot that guides student learning.
Using Messenger as the Interface
Delighted with their success, and looking for a way for NINA to further engage students, the team turned to Messenger.
Advertisements were posted in physical spaces and potential students could scan a phone code into Messenger to interact with the bot. If the student was interested in a course, they could apply on the spot. A basic application process began, and if students met criteria, they got an instant provisional offer of a place.
“The benefit for us is that it helps encourage conversion rates from applicants to enrolments, and remove some of that fear factor that students have around bureaucratic or administrative processes,” says Skinner.
Apart from engaging online, NINA is moving into physical spaces too. The university’s new Engineering and Technology building has a screen two stories high featuring NINA. She gives information about the university, and poses brain teasers to entertain students.
Robotic Process Automation
Using NINA to provisionally enrol students is an example of meeting a business need and creating efficiencies.
But although that worked at the start of the application process, AUT’s management system had no integrated API. Students connecting through Messenger had to be manually entered into the student management system.
The answer to this flood of new activity was a Robotic Process Automation. This goes into Messenger to gather the student application data, connects to the student management system and processes the application. The robot then tells the applicant via Messenger if the provisional offer has been confirmed.
Analytics Helping Student Learning
Apart from streamlining administration, AUT is using AI to support student learning. In 2015, the team developed a predictive learning analytics model that looks at the risk factors for students not coping academically.
Skinner describes this as feeding student data – the schools they’ve attended, how they performed, their ethnic background – into a predictive model that gave reasonably accurate forecasts on academic achievement.
“But we had some concerns about it as well, because those were things students couldn’t change about themselves. Even though a particular ethnic group might not do as well as another ethnic group in studying, that’s not true for all students in that group.”
Tracking Students on Campus
Faced with this concern, the team began thinking about other engagement indicators. They developed a student scheduling footprint as an attendance proxy.
“We looked at events generated when a student was physically on campus,” explains Davidson.
“It could be their phone connecting to wi-fi, logging in to an AUT computer or borrowing a book. Then we looked at the time stamps on those events and compared them to their scheduled classes. If those times occurred in close proximity to a class time, we marked them in attendance. Testing showed this method was fairly accurate.”
By week four of semester the team had a fairly good idea of which students were at risk. The data helped direct student services to students who needed help.
Another data development is the Student Digital Workspace, a calendar that lets students see intense periods of activity. It shows assignment due dates, the weightings on each and how much time they should spend on them.
Further Developments in AI
The team is now trying new ways of getting their AI bot into the student population. Large screens showing AUT Now TV exhibit key facts about the university. The vision is to create a single AI personality for the university, one that can engage with a present or past student, an academic or a parent.
A virtual classroom is also in the works. AUT spent $400 million on new fully connected buildings recently, and the team is working with the AUT hub in streaming analytics to collect all the information.
One example is monitoring room CO2 levels to tell if it is occupied. This can be useful for space planning. For instance, low CO2 levels would alert scheduling that a booked classroom was actually not in use.
Going beyond Admin to Academic Help
The key to the success of AUT’s AI team is its agility and its emphasis on free-thinking and innovation.
The team has realised the importance of designing AI functions to address specific business needs. For example, more accessible onboarding of students means less pressure on administrative staff. Automation of application processes has a direct impact on staffing levels.
AI is also making inroads into how students study, and even helping to recognise situations where students may be struggling.
What started as a way of streamlining administrative processes has developed into an AI driven interface having an impact on academic outcomes. AUT’s award-winning bot is leading the way for a new generation of New Zealand students.
About the Speakers
Tim Davison and Megan Skinner began working together in AUT’s Strategy & Planning team six years ago. Tim is responsible for Business Intelligence at AUT and has won multiple accolades for his work in emerging technologies, including the 2016 Microsoft Innovation Supreme Award. Megan leads a team exploring what’s next in higher education and manages a number of cross-university strategic projects. Tim and Megan are also co-Chairs of AUT’s all-staff network, Kin, which aims to build connections and collaborations between staff on campus.