If you start with the reasonable assumption that students learn better when instruction is personalized to their needs, and commit to realizing that goal, you embrace a series of instructional challenges. Many of these challenges result from trying to personalize within the context of traditional school structures that standardize the curriculum, the assessments, the grouping, and the instructional time.
If you start again with the assumption that these traditional structures are not going away, then you are faced with a genuine problem: how to achieve the tremendous academic gains that are possible through personalized instructional methods within the constraints of a traditional classroom.
Teachers who personalize instruction face many challenges, chief among them is managing the varied progress of 30 (or 150) students through what is still generally a standard scope and sequence leading to a standard end-of-year assessment.
Blended learning models attempt to solve this problem by dividing instructional responsibilities between teacher and computer. Intelligent learning systems personalize pace and pathway and provide corrective feedback based on developing mastery, enabling teachers to focus on deeper learning goals with smaller, differentiated learning groups. Rotation and flex learning models are specifically designed to coordinate this individualized and group instruction. With good data and a spreadsheet (or a mastery-based grade book), teachers can flexibly group students according to mastery of standards and differentiate their instruction accordingly.
Foundations of Personalization
Effective teachers personalize: they interpret what a student knows and is capable of doing in order to build on that knowledge and advance the student toward new learning objectives. But differentiating instruction at this scale (e.g., 30 students X daily differentiated instruction X multiple weekly learning objectives) would be impossible without technology and intelligent data mining and analytics. Two key concepts comprise the foundations for this work: knowledge maps and student models. Educational data mining and learning analytics puts these two in conversation to discern what conditions for learning will be most effective for what students within what contexts, and uses that information to personalize the learning experience for optimal results:
Knowledge maps
Formalizing a learning map--sequences of connected concepts and skills that define how one masters a domain, such as beginning Algebra--and mapping student mastery on the map, enables intelligent learning systems to recommend the next concept or skill to be learned, propose aligned instructional content, and present appropriate questions and tasks to assess mastery.
Student models
Other variables beyond a student’s current position on the knowledge map affect his or her progress in mastering a domain. Prior experience and interests, past academic performance, goals, motivations, self-concept, self-efficacy, and behaviors all can be predictive of student outcomes.
Learning analytics combines data from student models with data on learning behaviors, knowledge maps, and learning outcomes, and mines these data sets to identify patterns that associate student attributes and behaviors with successful outcomes. As systems become empirically smarter, observing correlations between behaviors and outcomes, they refine their recommendations, further personalizing them to learner profiles.
This type of work is far different from the formal hypothesis testing of most education research. Learning analytics provides a suite of analytical techniques that are better suited to address the frequency and quantity of data available in a digital age. The challenge for learning analytics (and its applications) is not in having large data sets to analyze, but to identify the right data to capture and analyze--in other words, to separate the signal from the noise.
New Challenges
As personalized learning expands in the coming years, so will the demand for learning analytics to improve the quality and efficacy of personalized learning designs. Schools are purchasing truckloads of tablets and advancing quickly toward 1-to-1 device implementations. The number of students today with access to Internet-connected personal devices is exploding, and schools are opening their networks to support them. Such trends indicate that we are just scratching the surface of tech-enabled personalized learning.
But the potential of analytics is not without certain obstacles. Certainly, student data privacy is an important issue that has received extensive attention recently, but there are others.
One key challenge is expanding educators’ perspectives on what is possible through analytics-driven personalized learning. Learning analytics marks a significant departure from traditional data-driven instructional strategies. That’s because so much more data is available to mine, make sense of, and use.
Where schools previously relied primarily on classroom, benchmark, and summative assessment data to differentiate, it is now possible to integrate a variety of other relevant data sets, including attendance, behavioral data, learner satisfaction, and the mass of data produced in computer-mediated environments. Using these expanded data sets, learning analytics can find patterns and develop predictive models that take into account a more nuanced student learning profile, helping educators refine their understandings of what works for what students in what context.
This brings us to a second challenge: to actively engage with practicing educators on how we design and integrate these tools into classroom environments. Researchers and designers of analytics-driven learning experiences must maintain close contact with the challenges of today’s classroom.It is not enough to design cutting edge analytics to shape educational decision making if we do not understand how teachers can apply them to optimize student learning outcomes. The value of a tool always lies in its use.
In the end, we feel that education has the potential to follow other fields’ journey toward personalized practices (such as medicine); however, it is important we are realistic about the nascent character of the field. Computer-mediated instruction is not new, but the proliferation of technology and data is. The quality and quantity of data available opens up new opportunities to deliver effective personalized learning experiences, but with it, certainly challenges. However, we have seen first-hand the benefit these technologies can have on millions of students, and so believe it is a journey worth making.
