Consumers in a world that is increasingly blended with online services and metrics are accustomed to companies and services gathering personal and preference information. From Amazon to Facebook, Google, and Netflix, our movements leave a trail behind us. This wealth of information has transformed the business world at every level, sometimes in advance of the ethical and privacy frameworks needed to guide construction of the new world. Notwithstanding global scandals involving the abuse of private information, as seen with the Cambridge Analytica incident, analytics and predictive modeling have the potential to deliver more personal experiences across everything from business to education. ^[1]

Educational institutions are famously slow to change, often with good reason. Indeed, large portions of today’s educational model, including the lecture hall experience, remain nearly indistinguishable from a model that dates back centuries. However, education is also not immune to the social and technological forces shaping the rest of the world. In the last half century, education has become more accessible and students have become more diverse and technically literate. The students that do much of their business from a mobile device are also completing digital assignments and interacting with technology across campus. Recent student technology studies show that the average student brings 3-5 internet capable devices to campus with many bringing upwards of 7 devices.^[2]

While we have not reached a future in which artificial intelligence is used to support just-in-time, adaptive, personalized learning at scale, technology is already having an impact in the classroom. If technology has a potential for transformative change in education, it is one that is not lost on businesses: Forbes notes that “in 2017, across every market involved in edtech, international funding reached a new record of $9.52 billion.”^[3]

Through all of these interactions with campus systems, students are leaving a digital trail that can be leveraged to support persistence, retention, and outcomes. However, it must be managed with respect for the privacy and security of students. The possibilities have caught the attention of institutional research offices, student support offices, academic technologists, faculty, and startup businesses. The NMC Horizon Report for 2018 lists analytics technologies as a key development to be adopted in the immediate horizon, a prediction that was included in the Horizon Report as early as 2011. As campus leaders weigh the options and return on investment for learning analytics solutions, there are several key challenges that must be overcome. ^[4]

The first is that relevant data is often disbursed across many campus systems and managed by disparate groups. Campus swipe card systems, library reservations systems, the student information system, advising systems, learning management systems, and even WiFi access and dining card data have potentially useful information for predicting student outcomes and engaging students when needed. Consolidating that data into a central warehouse that can run meaningful reports is often a sizable project in itself. Governance, campus politics, and data stewardship in these cases often overshadow the technical tasks needed to consolidate data.

The second issue is privacy and access role definition. The Spiderman refrain of “with great power comes great responsibility” must apply to learning analytics initiatives. As data is consolidated and linked, it is possible for campus administrators to know significant details about students’ comings and goings, and to make predictions about student outcomes based on this information. This power comes with large ethical questions that should be addressed. Who on campus should have access to this kind of information, what level of granularity should faculty and support staff be able to see, and how should information and predictions be shared? It is one thing to use information to help a struggling student. It is another to predispose faculty and support staff opinions toward the likely failure of a student. False matches, limiting predictions, and removal of student autonomy and agency through prescriptive systems are all potentially dangerous outcomes. ^[5]

This fact begs college administrators to carefully examine which questions to ask, and what the meaning of the data is before launching analytics projects. Through careful inspection and acknowledgement of bias, project teams can focus on positive impacts and ameliorate negative side effects in a learning analytics solution. The important thing to remember is that the math and algorithms in learning analytics solutions are not inherently neutral.^[6]

When in doubt, start with a concrete question that the college is trying to answer. As exciting as learning analytics opportunities are, it is important that the questions being asked through these initiatives are ones that have value in answering. The bottom line is that technology and information may inform policy and process, but it shouldn’t drive them—this is especially true when considering a very human activity like education. Just as importantly, the institution needs to be prepared to both understand the answers and act appropriately when analytics queries return results.^[7]

In this regard, campus culture and acceptance, as well as student support mechanisms are critical considerations for any learning analytics initiative. There is little value in developing a learning analytics initiative if student and faculty culture is not sufficiently accepting of the project, or if student support mechanisms are not prepared to effectively respond when potential students at risk are identified. Initiatives that answer specific faculty or support staff questions help keep learning analytics grounded in operational results. Starting simply also avoids costly investment in “what if” queries that may identify students in need of assistance but has no corresponding plan to deliver it. It’s always easier to expand a model that is working well than it is to fix a large initiative gone awry. There is undoubtedly much opportunity in learning analytics, but getting the most out of these initiatives and avoiding unintended consequences, requires planning and engagement of the stakeholders. ^[8]