e-Learning Ecologies MOOC’s Updates

Teacher/Learner Agency and “Customized” Learning

[approximately 2000 words]

There is some confusion when we talk about different forms of “customized learning” (adaptive learning, individualized learning, differentiated learning, personalized learning). In the end, through all of the various and sundry definitions, there are four or five specific aspects that distinguish these different forms. The major dimensions to pay attention to are: learner agency, focus on learners’ needs, teacher agency, technology and learning content. That’s it. I will go through these, ending with a particular vision of personalized learning because I believe it corresponds best to the vision that Bill Cope and Mary Kalantzis have laid out in the e-Learning Ecologies MOOC.

Below is a video from EDUCAUSE, a U.S.-based higher education technology association that focuses a lot of attention on the intersection of pedagogy and techology. This video offers a bit of an overview, mostly by talking about “personalized learning,” but also evoking dimensions that are significant for most of what is to follow (learner agency, focus on learners’ needs, technology). Then I’ll offer some comments on the various kinds of customized learning.

Media embedded July 13, 2017

Let's explore the terms. First, “Adaptive Learning”  has a multiple meanings. I’ll comment on only two here.

Generally speaking, it refers to an educational approach employing interactive computer technology that manages or guides learning processes, or that controls and orders access to learning materials or learning modules, with the intention of optimizing the learning experience; algorithms process data based on measurements of learner behavior, allowing the computer to select and present appropriate lessons, modules, activities or questions to the learner. For one recent article on such a system, see Drissi & Amariat 2016. A commonly used implementation is diagnostic computer-adaptive testing, where the system will pose a question of a certain level of complexity or difficulty or propose a task. If the learner responds incorrectly or inappropriately, the system may ask an easier or simpler question or give a simpler task. If the learner responds correctly, the system may pose a more difficult or more complicated question or give a more complex task. Over time, the system determines the approximate level of mastery or competence of the student. Such systems can be part of e-learning (see Drissi & Amirat 2016) and are sometimes employed in professional education and training (e.g. the setup discussed in Kleinhans & Schuma 2015). In an adaptive-learning module, the system might use diagnostic or pre-learning testing to determine the learner's level of knowledge or competence in a specific domain. Based on the diagnostic results, the computer might then present information, artifacts and/and activities for the learner to work through, verifying the learner’s progress and understanding via periodic checks. The algorithms will determine what is presumably an optimized learning pathway through a corpus of material, adjusting to the learner’s needs, or speed of processing or level of competence in artifact creation. An adaptive learning system could even guide a learner through an exercise in self-reflection about her or his learning by offering a series of interactive algorithm-guided prompts or questions. “Adaptive tutoring” is another implementation of adaptive learning, as are “MOOClets,” which can help teachers customize learning pathways for students in Massive Open Online Courses. To be clear, adaptive learning systems can have a variety of forms and functions. What they have in common is the use of interactive computer technology to customize the learning experience for individuals. Note that learner agency is low, although learner needs are attended to. Teacher agency is relatively high, in the sense that teachers help shape the algorithms and content. However, teachers are not necessarily monitoring learning. Technology is the functional agent in adaptive learning.

Note that “adaptive learning” can also refer to processes in machine learning or computational biology, where an artificial intelligence agent or a network of artificial neurons employs an analogous algorithmically-guided set of feedback mechanisms to manage the improvement of its ability to act competently or appropriately or to increase its mastery of a particular domain of knowledge or expertise. Such instances of machine learning may be either theoretical or applied. Examples of applied adaptive learning systems include artificial-intelligence-guided automated systems, such as a bipedal robot whose walking is controlled by a neural network that learns to walk in different terrains (Manoonpong et al. 2007) (click here for a two minute video) or computer-managed spacecraft maneuvering systems (Wu et al. 2015). Theoretical instances tend to be laboratory- or mathematics-based initiatives exploring such phenomena as human head-and-eye-movement coordination (Saeb, Weber & Triesch 2011) or dimensions of neural plasticity in adaptive learning (Khorsand & Soltani 2017). This latter example points toward the convergence of these two distinct kinds of “adaptive learning.” That is, machine learning approaches can be used not only to manage the adaptation of the learning paths to match learners’ needs (as measured and manage by feedback mechanisms), but also to enable learning systems’ acquiring (learning, if you will) greater adaptive competency in the guiding of learners. One recent example of this convergence may be found in an application of “swarm intelligence” machine learning as a way to improve the way in which tutoring systems are modeled, for greater efficacy (Rastegarmoghadam & Ziariti 2017). In a word, it is likely that in the future, adaptive learning systems may themselves become good adaptive learners and offer increasingly effective guidance to learners.

“Individualized Learning” generally refers to learning where a certain measure of control is in the hands of the learners themselves. In some implementations, there may be few significant differences in the learning content or in the kinds of activities or tasks that different learners must do; the "individualization" lies in the individual learner having control over the pace of her or his learning. In other instances of this approach, students may have some measure of control over the specific learning content and tasks, as well as the pacing. The key to the "individualized" track is a measure of "learner control," especially learner control of pacing. Note, however, that in individualized learning, teacher agency is high. Learner agency is present, but relatively low (students control pacing and have input into sequencing of content, although teacher agency and teacher-controlled content are fundamental). Teachers ostensibly focus on learners' needs, while offering little real differentiation in learning content or learning trajectories. Technology is not a necessary element, although it may be used in individualized learning.

“Differentiated learning” refers to an approach where the teacher crafts a customized set of learning materials, activities, goals and timetables that are keyed to learners’ needs. The teacher controls the learning content, pathway and pacing, while taking into consideration the particularities of a learner’s (or group of learners’) abilities, propensities, background and preferences. This way of customizing learning is often used in special education. Again, note that teacher agency and teacher control of content are high, as is attention to learners' needs. Learner agency is present but relatively low. (The personalized learning plan is forged with some input from the learner, but the teacher is the ultimate arbiter.) Finally, in differentiated learning, use of technology is possible but not necessary.

The last form that I will discuss is “Personalized Learning,” which is the form that focuses most on learner agency and the individual learner’s needs. In this educational approach, the learner, not the teacher, will make most decisions about the specific content, activities, forms of practice and modalities of inquiry and communication within their learning pathway. Learners will be able to focus on their interests, even though they will be required to function within appropriate learning parameters. It is the least rigid or regimented of the approaches named here. The teacher is not without agency in the process, serving as a guide and support in the process, without being fully in control. Teachers facilitate learning by helping to organize and scaffold the learner’s experience. Teachers serve as expert guides and advisors (being the "guide on the side"). Some teachers are wary of such an approach because they believe that it diminishes their agency and their authority as experts. Indeed, a robust regime of personalized learning is radically different from the typical teaching practices of last century (the teacher as the "sage on the stage"). In truth, many teachers are uncomfortable with the enormity of the shift.

Despite teacher resistance, personalized learning has a number of distinct strengths. By making the learner an active participant in her or his learning, it increases the impact of noncognitive factors (enthusiasm for pursuing projects and subjects of interest, increased metacognitive engagement as the learner plans, reflects and self-assesses) and it promotes cognitive engagement (since the learner is almost exclusively responsible for inquiry, research, analysis, communication of results and managing collaboration with others). It gives learners a stake in the creation of new knowledge and allows a greater range of learner-organized collaborations. Through a “wisdom of crowds” effect, it has the potential for developing new lines of research, novel modes of critical inquiry and new veins of knowledge and channels of communication that may turn out to be productive and meaningful for society at large. Finally, note that "personalized" pathway learners may choose to use some of the other approaches that I've mentioned (adaptive learning tutoring systems, approaching canonical learning content in traditional ways, varying only the pacing, or seeking out teacher-constructed learning content or activities). To be frank, there are risks to this approach and it needs to be managed deftly, with enough teacher monitoring to assure that learners acquire minimally required knowledge and skills, but not so much that it undermines learner agency. Teacher education and training will probably be needed to bring educators on board in a way that will be productive and helpful as they develop and further this approach -- and, of course, assess its effectiveness. However, it seems clear that this educational approach, supported by digital communications and connected archives of information and media, fully enabled by effective support from teachers who serve as learning experts, domain specialists and advisors, holds the most promise for robust, creative and effective learning in the twenty-first century (Zmuda, Curtis & Ullman 2015).

Certainly, "personalized learning" is the approach that best aligns with the affordances of e-Learning Ecologies (ubiquitous learning, active knowledge-making, multimodality, recursive feedback, collective intelligence, metacognition and differentiation in learning). To recapitulate, personalized learning emphasizes learner agency and leaners' needs most of all, with support from teacher agency. Teachers do not, however, control all of the learning content (or even most of it). Technology can be used (and could prove very helpful in forms like an artificial-intelligence-guided tutoring system that can learn to improve its performance at figuring out what works best for various kinds of learners and learning challenges). However, technology is not necessary, even though some edtech companies claim that their products or services provide perfect calibration of personalized learning or are essential tools for such an approach. (Such claims are patently untrue.) Autonomous learners, supported by teachers and administrators who prioritize good learning outcomes and learner agency can, without using any particular techology, promote good self-directed ("personalized") learning. See the video below, If Students Designed Their Own Schools..., about the "Independent Project" program at Monument Mountain Regional High School in Barrington, MA. Note the emphasis on learner agency and the relative absence of technology.

Media embedded July 14, 2017

As a post-scriptum to "If Students Designed Their Own Schools..." I note that the principal high-school-student instigator and designer of the "Independent Project" alternative school program, Sam Levin, recently co-authored and published a book with his mother, Susan Engel, a developmental psychologist. A School of Our Own focuses on the Independent Project as a model for learning in high school (and beyond) in the twenty-first century (Levin & Engel 2016). You may also watch Sam Levin's TED talk about the experience by clicking here.


Basye, Dale. (2016). Personalized vs. differentiated vs. individualized learning [web article dated 23 October 2016]. Retrieved 13 July 2017 from https://www.iste.org/explore/articledetail?articleid=124.

Drissi, Samia; & Amirat, Abdelkrim. (2016). An adaptive e-learning system based on student's learning styles: An empirical study. International Journal of Distance Education Technologies, 14(3), 34-51. https://doi.org/10.4018/IJDET.2016070103

EDUCAUSE. (2016). How do you define personalized learning? [video published 7 March 2016]. Retrieved 13 July 2017 from https://www.youtube.com/watch?v=CJqZrV-Xsgg.

Khorsand, Peyman; & Soltani, Alireza. (2017). Optimal structure of metaplasticity for adaptive learning. PLOS Computational Biology. https://doi.org/10.1371/journal.pcbi.1005630

Kleinhans, Janne; & Schuma, Matthias. (2015). Increase in testing efficiency through the development of an IT-based adaptive testing tool for competency measurement applied to a health worker training test case. International Conference on e-Learning, 2015, p. 42-50. Retrieved 13 July 2017 from http://files.eric.ed.gov/fulltext/ED562462.pdf

Levin, Samuel. (2013). Why high schoolers should be in charge: Sam Levin at TEDxOxford [video of a TED talk published 28 April 2013]. Retrieved 14 July 2017. https://www.youtube.com/watch?v=Nlql1ESio5w

Levin, Samuel; & Engel, Susan. (2016). A School of our Own: The Story of the First Student-Run High School and a New Vision for American Education. New York, NY: The New Press.

Manoonpong, Poramate; Geng, Tao; Kulvicius, Tomas; Porr, Bernd; & Wörgötter, Florentin. (2007). Adaptive, Fast Walking in a Biped Robot under Neuronal Control and Learning. PLOS Computational Biology. https://doi.org/10.1371/journal.pcbi.0030134

Rastegarmoghadam, Mahin; & Ziarati, Koorush. (2017). Improved modeling of intelligent tutoring systems using ant colony optimization. Education and Information Technologies, 22 (3), 1067-1087. https://doi.org/10.1007/s10639-016-9472-2

Saeb, Sohrab; Weber, Cornelius; & Triesch, Jochen. (2011). Learning the Optimal Control of Coordinated Eye and Head Movements. PLOS Computational Biology. https://doi.org/10.1371/journal.pcbi.1002253

Tsai, Charles. (2013). If students designed their own schools.... [video news report published 13 February 2013]. Retrieved 14 July 2017 from https://www.youtube.com/watch?v=RElUmGI5gLc.

Williams, Joseph Jay. (n.d.). [web page on MOOClets]. Retrieved 13 July 2017 from http://www.josephjaywilliams.com/mooclet.

Wu, Shunan; Wen, Shenghui; Liu, Yuliang; & Zhang, Kaiming. (2015). Robust adaptive learning control for spacecraft autonomous proximity maneuver. International Journal of Pattern Recognition & Artificial Intelligence, 31 (5), 1-15. https://doi.org/10.1142/S0218001417590078.

Zmuda, Allison; Curtis, Greg; & Ullman, Diane. (2015). Learning Personalized: The Evolution of the Contemporary Classroom. San Francisco, CA: Jossey-Bass.