Eitel Lauria, Director of Graduate Programs at the School of Computer Science and Mathematics, Marist College
Could you tell us a little bit about yourself and how your career has been so far?
I hold a 6-year Electrical Engineering degree from University of Buenos Aires, an MBA from Universidad del Salvador, and a PhD in Information Science from SUNY Albany.
I am a Professor of Data Science and Information Systems and the Director of Graduate Programs at the School of Computer Science and Mathematics, Marist College, a liberal arts institution in Poughkeepsie, New York.
My area of expertise comprises both theoretical as well as applied data science.
As my role suggests, I have been involved in a lot of projects related to data science and analytics, machine learning, data mining and predictive modeling. You could say that I have been a data scientist long before the term was even coined.
Leveraging my knowledge and experience over the years, I wanted to explore the possibilities of learning analytics to mathematically create models that could help improve the chances of student success. My work startedback in 2011 with the Open Academic Analytics Initiative (OAAI), a project funded through EDUCAUSE’s Next Generation Learning Challenges (NGLC) and the Bill & Melinda Gates Foundation, aimed at developing an early detection prototype system of college students at academic risk, using machine learning models trained with student data.. The OAAI was the first early detection prototype developed on an open-source platform and as such received considerable amount of attention and recognition, including several prestigious international awards. Pilots of the predictive modeling framework were tested at two community colleges and two HBCUs, and were subsequently implemented at North Carolina State University and at several universities in the UK. A revamped version of the system, called MUSE (Marist Universal Student Experience), was implemented at Marist College in 2018. It is currently part of ilearn, our learning management system. Using this platform, we are able to detect at least 87% of the students at academic risk 6 weeks into a 15-week semester.
What are the factors to keep in mind for educational institutions when leveraging analytics for student success?
Many institutions have used analytics as descriptive statistics rather than prescriptive modeling. Nonetheless, I would always suggest educational institutions to have a culture of data collection and feed the data warehouse to develop prescriptive models. The central aspect of data prediction is to have consistent data for several years, without which one cannot make a good prediction.
What has been the impact of COVID19in the use of predictive modeling for early detection?
COVID 19 has brought a huge disruption in all of our lives, and it has certainly affected the use of AI-driven predictive models. These models learn from experience and use data from the past to make predictions for the future, so the outlier created by the pandemic disrupted our ability to make accurate predictions for the time being. When things get back to normal, we will probably have to skip the pandemic years, or factor them in, if they have introduced new behavioral patterns in student learning. We have to test this, time will tell.
How do you envision the data analytics landscape, especially in the education sector, in the next five years?
I think the future is bright for data analytics to track student performance. The pandemic has been an eye-opener for everyone, including the education sector, to understand the relevance of digital technologies. So, with a shared vision and a shared strategy tied to this idea of evolving toward digitalization, the education sector is ripe for adopting data analytics for tracking the behavioral traits of students.
“So, my advice would be to put together a leadership team within the organization capable of developing and evolving data collection, data sharing, and if possible, data analysis, data modeling, and predictive modeling methods in-house, with the goal of improving the chances of student success"
With that, I would also add something from a faculty member’s perspective on adopting digitalization in a learning environment. Undergraduate and graduate students look at online learning differently. While graduate students live and breathe in online learning platforms, undergraduate students prefer campus life. So, when strategizing, we need to bring the perfect harmony between the two with a hybrid learning model. Online classrooms are helpful in several ways but do not replace classroom activities. On the other hand, many graduate students who are also working prefer online classes with flexible schedules. For undergraduates, again, online learning should not create a sense of disconnection as they are not receiving education face-to-face. All these thoughts should be kept in mind when any educational institution is planning its digitalization endeavor.
How do you want to inspire leaders of educational institutions to undertake analytics initiatives?
Different institutions follow different approaches depending on their analytics strategies and their resources availabilty. Marist might be a small liberal arts college, but its technological mindset is equivalent to any large-scale research institution, because of its long-standing vision of using technology to support teaching, learning, and scholarship. I should also mention its close relationship with IBM over many years. As a result, we have the resources to do any research in-house. If institutions don’t have the resources in-house, they can definitely partner with relevant providers; but they still need to collect raw data.
So, my advice would be to put together a leadership team within the organization capable of developing and evolving data collection, data sharing, and if possible, data analysis, data modeling, and predictive modeling methods in-house, with the goal ofimproving the chances of student success.
Institutions must embrace digitalization that allows data administrators to collect information, whether perceptual or transactional, from any given system: information management systems or student information systems.
It would help the institutions understand the reasons for student dropout from one of the three perspectives: financial, academic, and behavioral. The cost of college keeps rising regularly, and some students cannot keep up with the rising costs, eventually being forced to drop out. The academic aspect, on the other hand, may or may not be related to the financial factor because a student can be good financially, but they may still struggle academically. And academic struggle could have behavioral roots. I want to stress this, as it is certainly the most difficult dimension of the problem to gauge from the data we collect. Behavioral data is out of bounds due to privacy regulations. I believe that the students behavioral dimension deserves a higher level of analysis and exploration.We need to identify surrogate data without being intrusive. This can certainly improve our predictive models and actually benefit a student before it is too late.
Jonathan Daitch, Associate Provost for Online Education, Western University of Health Sciences and Jonathan Labovitz, DPM, FACFAS, CHCQM, Associate Dean, Clinical Education and Graduate Placement Professor, College of Podiatric Medicine at Western University of Health Sciences