Skip to content Skip to navigation

Predictive Analytics Discussion Notes - North Regional Meeting

Predictive Analytics was one of three shared areas of interest for participants at the North Georgia CCG Regional Meeting held at the University of North Georgia on March 26, 2016. Brief notes from the breakout conversation on this issue at the meeting follow:

Predictive Analytics discussion North Georgia Regional Meeting

  • One Concern: Establishing a predictive analytics model that is effective and ethical

Key Concepts:

  • Predictive analytics, gone too far, can derail students by predict them away from their dreams or goals.
  • Success with Predictive Analytics is about relationships with students and not their graduation rates
  • It's one thing to identify the students, it is quite another to do appropriate interventions.

One Big Idea

  • It's a tool, and one of many tools to help students progress. Campuses have simple tools they may have in hand to do some of this work.  
  • One thing we can start doing but start sharing the factors that show up as predictive.  Build more shared understandings of the strengths of each institution, then one of the things we can talk to students about when they have trouble at one institution is where they can benefit.  

Discussion Notes:

  • Too much data, but there is an absence of consistency
  • Need to look at an abstraction of the D2L data to be useful.
  • Working with faculty to put learning outcomes into their courses so they can be mapped onto students  for early alerts.  
  • Analytics is moving toward an individualized model--much like Amazon will use your previous purchases to determine what you see.  The curriculum, the learning style can be matched to students to maximize their success.  
  • How do we start building toward accessing data (the System's data) that don't have student behavior but have student academic data to do analytic work to do some analysis.   
  • Civitas was able to work without a long-period of data.
  • There was concern that comes up with who has used proprietary models and what have their successes been.  You want to be able to follow along the initiatives at other institutions to realize specific gains.  P/A allows for testing hypothesis but the evidence on success with strategies implemented is lacking.
  • There are opportunities to apply the information from Predictive Analytics on both the academic and student affairs side.  The models work unless they are used with other units. 
  • PA models may need to be adapted and adjusted for the type of institution you are.  
  • Proprietary systems are not cheap and it may be better to invest in a more local solution with people who know us better.
  • Predictive Analytics gives an opportunity to come to a diagnosis quicker.  Often in higher education we are hoping that something works, but we need to have a focus on high-impact strategies, but we need the analytics to make the decisions 
  • Instructional design for predictive analytics and early alerts remain a challenge.  Early alerts require some information early on.  Not all of them are driven on grades, some involve student engagement question, which may be an argument for proprietary vendor software since they have the validated model and process.  Initially GHC looked at participation at the 25% point (logging into the course LMS, speaking up in class, showing up, turning in assignments, etc.) 
  • Once you have the early alert information, what do you do with it.  Once the list is generated, the intervention is where the rubber meets the road.  One process is calling or texting the student when they have not engaged.  EAB has a cascade of emails or messages for students.  
  • When you start leaning heavily on analytics, you can get a lot of information on students, which opens up privacy and ethical information.  What you can and cannot use is an open ethical conversation.  There is a difference between predictive and prescriptive analytics.  Prescriptive analytics are the sort of thing that scare students (we have dictated your path).  When you get really good models edges very close to prescriptive.  Predictive analytics provides students with information and advice.  
  • Metamajor effort helps students identify the right fit for them without being prescriptive for students.  
  • Predictive analytics can help to structure conversations with students about their likelihood for success.  
  • It is part of the obligation to the student to frame our advice to students in the context of their interests.  Students benefit from having opportunity to explore and determine what adjustments are necessary
  • Using a range of information of student background to courses with different interventions that may help them be successful.  Suggest courses for students based on intake data and provide suggestions.
  • Concern with the conversation the student hears with the data.  Based on hope alone, having a student want to do something, but they need information to help guide them on the way.  
  • The human interaction is what makes the system works; Predictive analytics is a tool that is only as good as the person who wields it.  Human interventions take resources.  It's an investment but the return on the investment doesn't appear immediately.  
  • How can our colleges be prepared for the coming college-going population--more FGCs, more diversity, less prepared.  Many of the same interventions still work.  
  • Who has had success with collaborating with faculty in using Predictive Analytics? As the demographics change, what do you need to do to adjust to this new culture.
  • Promoting the scholarship of teaching and learning helps to improve instruction.  Makes the work of becoming invested in new teaching styles have currency among faculty with real value within their disciplines. These become ways to grow as scholars in their disciplines and to become better teachers.
  • Work with faculty with respect to fit for the program and institution?  This may be something for more research, especially with helping to diversify the disciplines and programs.  
  • Student fit of the instructor.  Need to focus on the factors in your control, but there is an issue of teaching/learning fit. The GAME initiative provides this opportunity to make those connections.