Adaptive learning through algorithmic, personalized software is relatively new and it is fair to demand that research and evidence be produced that shows its efficacy in terms of student outcomes.
Northeastern University research
Northeastern University, in Boston, have just published their preliminary findings on the Winter 2014 cohort at Shoreline Community College, who took 1 or 2 courses (Microeconomics and English) online, via an implementation of the CogBooks platform, integrated with the Canvas LMS.
As students progress through the readings and assignments, the adaptive learning software provides supplemental tutorials as needed to individual students, records student time on activities and modules, gathers student assessment data to inform the learning path for each student, and provides reports to instructors. Instructors map course content in the adaptive learning platform using CogBooks to answer questions students pose in the system, monitor student performance, and guide class discussion and emphasis.
Qualitative Findings – English 100
“At this preliminary stage students are illustrating noted improvement over previous sections of English 100 in two areas. First, the quality of their essay revisions shows solid understanding of revision strategies and the ability to apply those strategies to improve their writing. Second, their writing also shows increased awareness of sentence-level issues such as word choice, phrasing, grammar, and mechanics.”
Students’ positive perception of the value of the adaptive learning system and its effectiveness in their learning increased throughout the term. On completion of the final outcome, students were overwhelmingly positive about the value of the adaptive learning integration and its impact on their understanding of the key reading and writing outcomes in the course. One student commented “I have taken lots of online courses and this is the first one that feels like an online course ought to be”.
Intro to Microeconomics – January 2014
Comparing this adaptive Spring 2014 version with the Fall 2013 non-adaptive class; both courses had 30 students. I have seen that this quarter (in the adaptive class), the students are progressing to the next level of content with greater ease…without the same stumbling blocks as in previous quarters. Microeconomics is tough for many students and typically gets really tough just past the mid-point when students have to apply their learning. Analyzing market structures, for example, relies on understanding of concepts from week two to week five. These results and conclusions are tentative – we have tests and test results coming up but students across the board seem to have mastered these concepts better than in previous quarters.
Gates Commissioned Report
In ‘LEARNING TO ADAPT: Understanding the adaptive learning supplier landscape’ the Gates Foundation commissioned EGA to provide some real analysis and insights into the different solutions on offer in the market. Each of the eight suppliers deliver different types of adaptive solutions, so the authors try to apply credible judgment criteria and look at the landscape through three lenses; Approach, Taxonomy, and Maturity.
They start with a general definition of adaptive learning around a personalized learning experience adjusting to the learners needs and wisely point out that this is a radical departure from traditional online learning that needs a “fundamental redesign of the course experience”.
Six pedagogic attributes
This is a very thorough analysis of business and instructional models of eight suppliers and aims to objectify the analysis by applying six pedagogic attributes to all eight of the adaptive learning solutions:
1. Learner Profile is a structured repository of information about the learner used to inform and personalize the learning experience
2. Unit of Adaptivity refers to the structure of the instructional content and the scale at which that content is modified for specific learner needs
3. Instruction Coverage refers to the pedagogical flexibility of a product to deliver an adaptive learning experience and the scope/scale of that experience within the context of a course
4. Assessment is the frequency, format, and conditions under which learners are evaluated
5. Content Model describes the accessibility of the product’s authoring environment to instructors or other users and their ability to add and/or manipulate instructional content in the system
6. Bloom’s Coverage highlights to what extent a product can support the learning objectives within the Cognitive Domain of Bloom’s Taxonomy
We are pleased to see that we, at CogBooks, are included as one of the eight suppliers, and do rather well on the six attributes. We are seen as a cloud-based learning platform that optimizes sequence and speed using “prerequisite sequencing, retention, cognitive load and attention, and level of engagement, among others”. Uniquely we use prerequisites along with repetition to achieve retention. At first the learning activities are constructed manually with a default path and if the student has difficulty a different learning journey is presented. The system is student profile and algorithm-driven to deliver dynamic learning paths. In fact, the learner’s path is remapped after every screen, responding to the learner’s current profile. The system supports content creation within the platform as well as that developed in third-party systems, and open-source resources from the web.
Free copy of the full report ‘LEARNING TO ADAPT: Understanding the adaptive learning supplier landscape’
The Cogbooks platform uses algorithms, based on scientific learning theory, to personalise learning by constantly working out, in real-time, what the student should do next. Ever screen and learning experience that is presented has been judged to be personally appropriate at that exact moment.
The content is not stored as a series of flat and linear screens but as a network of learning. Students vector through this network depending on a whole raft of factors, each student taking different routes. It is like having your own satnav as you progress, constantly working out where you’ve come from, where you need to get to and getting you back on track if you go off track.
These results are tentative, but positive, and show improvements not only in attainment but also in halting drop-out (withdrawals). If these results continue to be corroborated, we will have evidence that adaptive learning, using algorithmic, personalised software may result in huge rises in student attainment. This paves the way for using algorithms and student data to hugely accelerate learning.