Non-Cognitive Signals and Systems: The Science and Technology of Connectedness

Introduction

Andrea is currently working two jobs in an effort to afford college. She found a school which provides both online and in-person, night classes, plus they transferred some credits from her first attempt at the university experience. This time has to be different. Everywhere she has gone, Andrea has run into the wall of not having a degree. But while she does not have to pay the exorbitant prices of living on campus, the cost of living for her and her boyfriend are still just out of reach. Considered “food insecure” by national surveys, Andrea goes without meals around 35% of the week, since being let go from the restaurant and having to find work at a gas station. This also means that Andrea is quite lonely, not even able to see her boyfriend on a regular basis. The little interaction she gets with fellow students is almost exclusively during in-class, group work, which is not a social experience in any way. Andrea has yet to find a class or a professor who really challengers her to think beyond what she is capable of, but she also regularly misses assignments and even quizzes on occasion, as work must take priority. As a result, her instructors do not spend much time focusing on her, the other students do not know her, and Andrea is likely a paycheck away from having to drop out again, this time likely forever.

Background

Much has been written about “cognitive” learning analytics as a risk signal. Siemens (2011) has provided an often-cited definition for learning analytics as, “the measurement, collection, analysis, and reporting of data about learners and their contexts, for purposes of understanding and optimizing learning and the environments in which it occurs.” Yet ‘contexts’ in this definition, while ambiguous, has proved to be largely myopic. Many companies such as Civitas Learning, HelioCampus, EAB, Starfish, and more specialize in retention strategies based largely on cognitive (or at least proxies for cognitive) markers, juxtaposed against demographic factors which are then weighed against cohort averages. From the Civitas Learning website: “Our Student Success Intelligence Platform (SSIP) ingests student data to inform student support systems. With a direct connection to their students, institutions learn from each interaction, creating an iterative learning cycle (https://www.civitaslearning.com/).”

But again, the term “student data” is misleading. The phrasing should likely be “student coursework data” or “student academic behaviors” as students are impacted by and see learning context far outside of the classroom, yet none of these “non-cognitive” factors are included in most learning analytics initiatives. According to Vetter, Schreiner, MacIntosh, & Learning (2019), “Despite decades of research on student involvement, few studies have examined how co-curricular experiences promote holistic student success outcomes. Fewer still have differentiated the characteristics of co-curricular involvement to determine practices most likely to predict student success” (p. 39). This can be seen in the strategies of top “Student Success” platforms today: