Tuesday, November 26, 2013

What does it take for students under the classroom model of group instruction to learn as effectively as students under good tutoring conditions?


video

In this seven minute video I share some ideas about the effectiveness of one-on-one tutoring and the path to "2-Sigma" learning at scale.

Wednesday, November 20, 2013

Scaffolding and Feedback

I'm learning more about the power of scaffolding and feedback for optimized student learning.  I've learned about it from the learning sciences perspective by studying research papers on the subject, but I've recently gained some new practical and tangible insights as I tutor one of my sons.

Effective tutoring uses both scaffolding and feedback to help students bridge gaps in understanding and develop new skills.  The techniques are conceptually similar but different in that scaffolding is more proactive and feedback is more reactive.  An example of scaffolding is when a tutor is helping a student learn how to solve a multi-step mathematics problem and prompts the student, "What do you do first?"

Scaffolding provides a safe structure within which the learner continues to build on and reinforce existing knowledge and skills.  It provides guidance and motivation in the process of learning while keeping the learner at the center of the learning experience.  It does not let the learner off-the-hook by providing answers.  If the student says, "I don't know."   The tutor might rephrase the question as a multiple choice or ask a more leading question such as "What kind of problem is this?" or "What is the problem asking you to figure out?"  Scaffolding is not the same as instruction. A good tutor resists the urge to say, "The first thing you need to do is…"   Only when it is clear that the learner has a knowledge gap does the tutor step in…and that brings us to the role of feedback.

Formative feedback in tutoring is the tutor's response to a knowledge gap, misunderstanding, or area for improvement aimed at correcting/improving the cognitive imperfection.  Feedback is reactive in that it's driven by formative assessment or observation data.  Formative feedback is descriptive and specific, shedding light on the nature of the misunderstanding or procedural error and guiding the learner to correct it.  Feedback can be instructive like, "No, you skipped a step; the first thing you need to do is…,"  or it can be scaffolding-like, helping the learner self-correct, such as, "Check your signs."

Research shows that scaffolding and feedback are most effective when addressed at at-least "step-based" granularity.  For example, on a multi-step mathematics problem, effective feedback goes beyond telling the learner whether or not the answer to a problem is right or wrong.  Expert mathematicians know how to solve for x, because they have mastered the process-skills, but also because they have deep understanding of the concepts that make the process work.  On competencies that don't involve process steps the scaffolding might involve questions that prompt metacognition (i.e. when the learner thinks about what they are thinking) and explore new connections leading toward deeper understanding.  For example, building connections between historical events.

Scaffolding and feedback are some of the most powerful catalysts for learning.  I know this from the research and from practical experience.  Learning takes place much more effectively and efficiently under good tutoring conditions that include scaffolding, feedback, high expectations, and competency-based advancement.

Friday, October 4, 2013

Digital Infrastructure for Personalized Learning at Scale

One of the ideas I'm working with for my upcoming presentation at the iNACOL Blended and Online Learning Symposium is the role that various data initiatives can play in the infrastructure to support personalized learning at scale. This graphic shows some of the key players and concepts in that infrastructure:


Note the learner is in the center, and a continuous measurement-feedback loop is central. Technology is not a complete solution, only an enabler. The data infrastructure is especially important for providing the right kind of immediate feedback to the learner and with educators adjusting learning experiences to learner needs. We know from learning sciences research what works, we just can't do it to scale without the help of technology.

A complete scaled solution must also include innovations in professional roles, practices, organizing principles, and fiscal models. I plan to share thoughts on these other innovation in my session at iNACOL and later in this blog.

I'm also leading a workshop in which participants will design a competency-based "the-learner-is-the-school" model using some resources I've developed.  Here is the session information:

Blending Learning Science, Learning Standards, Process Optimization, and Data Standards for Competency-Based Learning at Scale
Monday, October 28 from 3:30-4:30

Workshop: Design a Competency-Based System/Organization/Culture to Maximize Student Learning Wednesday, October 30 from 10:00-12:15

Thursday, August 8, 2013

"Learning Styles" vs. Personally Optimized Learning Experiences


I've been following a debate on a LinkedIn group on the value considering "learning styles" when designing online learning experiences.  The Wikipedia definition of learning styles is: "an individual's natural or habitual pattern of acquiring and processing information in learning situations."  One of the most commonly used learning style categories are visual, auditory, and kinesthetic.  The following is an adaptation/expansion of a comment I posted to the thread...

Like, Howard Gardner's theory of multiple intelligences "learning styles" are ways of trying to classify or make sense of the idea that each human being is unique, and therefore the factors to optimize learning for each person, within each learning experience, are unique.  The theories are attempts at moving away from one-size-fits-all "factory" models of education to more individualized approaches.  Many of the comments on the LinkedIn thread express the opinion that these theories  fall short, have not been proven, and can sometimes be a distraction from the more important factors that affect learning. 




When considering validity of a learning sciences theory, the context and application are important.  Under the constraints of teacher-centric, resource constrained, classroom models, there is no evidence that trying to deliver instruction to each learner's perceived "learning style" is a good approach.  However, design of online/interactive content can include multiple paths, experiences,  and modes of delivery that the learner can self-select. This seems to be a more valid application of the concept. It doesn't require a teacher or artificial intelligence engine to make a judgement about segregating learners by "styles." Instead it provides the learner with the ability to try multiple 'styles' of information delivery and/or different kinds of learning experience until the learning goal is reached.   (I don't know if there is research that compares the effect size of online content that provides multiple delivery "styles" vs. single delivery "styles," but I suspect there is value in providing the learner with alternative delivery options.)

In my opinion, the role of the learning sciences is to better understand those factors that may impact an individual's learning, to test methods designed to optimize factors in delivery of a learning experience, and to determine if those methods impact learner outcomes, generally or for specific populations/conditions. 

Today we think of Vygotsky's ZDP theory as "just good teaching practice," (quote from the LinkedIn discussion), but there was a time when the concept was not considered, and unfortunately is still widely ignored in the practice of group-centric fixed-pace instructional models.  This is why we can't rest on what we know (and do).  What the education and training communities have to offer is not working well enough for many learners.  There is much more research to be done, more learning innovations to be discovered, and a great need for improvement of methodologies toward personally-optimized learning experiences.

Tuesday, June 4, 2013

Competencies for Competency Education

iNACOL's Chris Sturgis recently blogged about the organization's efforts to define a framework of competencies for implementers of competency-based education and a system of training and "badges".  This is important work at a time when state and local policy makers are re-examining "seat-time" requirements and opening doors to competency-based alternatives, when schools are piloting new competency-based models, and when new post-secondary delivery models are emerging, such as that of Southern New Hampshire University's College for America.  

Education leaders face significant challenges in transforming the culture and work processes within existing organizations, and in collaboration with external entities, to successfully implement competency-based models.  iNACOL is asking what competencies education leaders need to successfully transition their organization, and then to successfully manage the competency-based delivery model.

For educators, competency-based delivery requires professional practices that are different from what has worked in seat-time-oriented instructional models.  What competencies are needed by educators to facilitate competency education, and what training is needed?

My suggested approach is to start with an understanding of the process models for competency-based education and the work functions within those models.  How do schools that have successfully implemented competency-based education do it?  What does the process look like?  What are the inputs and outputs of the process? What are the critical work functions and process steps within the model?

One reason why I think it is important to break down the work elements as functions within a process is  because emerging blended and online learning models distribute the work of teaching and learning in new ways.  In the past, we could define the set of teacher competencies needed to teach within a subject area or grade range.  The assumption was that a classroom teacher would be for the most part an independent practitioner within the classroom, having full responsibility for the many teaching functions. Emerging models allow for and require more collaboration  and specialization, instead of one role, the professional roles vary by model and implementation specifics. Greater professional specialization and differentiation of educator roles is not only a key to success for some competency-based instructional models, but is also providing new opportunities for educators that can result in higher levels of job satisfaction and compensation.  Information systems also change the nature of the work, e.g. reducing the burden of manually tracking individual learner progress.  The competencies required by educators may also vary based on the information systems used.

The process for delivering competency-based education has some work functions that are common regardless of the model or implementation environment, such as advance-upon-mastery-decision-making, diagnosing misunderstandings or skill deficiencies, making prescriptive recommendations, and delivering remedial instruction.

A good starting place for defining the "competencies for competency education" is to discover those work elements that are common across delivery models.  A next step is to discover the skills, knowledge, and  "habits of practice" needed to perform the work elements effectively.  The resulting set of competencies, grouped by work elements, would provide a flexible framework.   Each implementation may assign work elements differently to specialized jobs, taking into account the role of technology and implementation-specific factors, and then be able to reference the needed competencies for the person filling each job.



Thursday, April 25, 2013

Shifts in Educator Roles, Professional Practices, and Organizational Models Driven by eLearning Disruptive Forces


Michael Sandel is an international superstar.  His 2013 tour filled 14,000 person stadiums in places like Seoul Korea.  He was broadcast on national television and became so popular that he was asked to take a place of honor at a national sporting event.  In China his Internet video has been watched over one million times.  If you have not heard of Michael Sandel you might be surprised to hear that he is not the lead singer in a rock band, or a Hollywood movie star.  Michael Sandel is a Harvard philosopher and lecturer.  Sandel’s “show” is a series of lectures on justice and political philosophy, not exactly a topic regarded in the popular mainstream.

Andrew Ng is both a “rock star” and an “edupreneur.”  In October 2011 he offered his Machine Learning course at Stanford for free to anyone in the world.  Over 100,000 people enrolled in that first iteration. His work subsequently led to the founding of business start-up Coursera in 2012.  He is quoted as commenting that to reach that many students within the traditional model he would have to teach for 250 years.

An ecosystem empowered by global telecommunications and Internet technologies is creating the opportunities for some teachers to have a positive impact on significantly more lives and reach unprecedented levels of fame and fortune. Disruptive forces driven by eLearning technologies, “big data”, new human capital strategies of educational institutions, market demands, economic conditions, learning science discoveries, and innovations in professional practice are driving toward new education delivery models.  Some of these models give good lecturers a “bigger stage” without significantly changing the big lecture hall experience for learners. Other models leverage the specialties of a team of education professionals to personalize and optimize student learning.  Market demand for these new delivery models has already begun to impact traditional institutions.

New models of delivery require changes in professional practices. Emerging online learning models, for example, shift the focus from hour-long lectures and textbooks to more interactive shorter cycle learning experiences in which individual learner understanding is checked continuously.  While much of the hype around massively open online courses has been about the number of students watching video lectures, a more important metric is the higher quality that can be achieved with economies of scale and the massive amounts of data that can be mined to optimize student learning.  This data is powerful as a feedback loop for optimization of individual learning experiences and as a means to optimize learning experiences across populations.  This “big data” also supports new research methodologies driving down the cost and time that it takes to do gold standard studies and with greater applicability.

“You can turn the study of human learning from a hypothesis-driven mode to the data driven mode.  You can ask questions like what are some of the misconceptions that are more common and how can we help students fix them.”
(Koller, Daphne; http://www.ted.com/talks/daphne_koller_what_we_re_learning_from_online_education.html, June 2012, downloaded 21 April 2013)

Part of what is giving this "big data" the greatest impact is the adoption of common data vocabularies, such as those defined in the Common Education Data Standards (CEDS), an initiative led by the U.S. Department of education.  CEDS doesn't collect data, it provides common definitions used across organizations and sectors so that data about student learning experiences from one institution may be compared to data from another institution.

Professor Neil Heffernan teaches artificial intelligence at WPI.  Unlike Andrew Ng, however, Heffernan doesn’t lecture to hundreds of thousands of students through a massively open online course.  He is achieving a different kind of fame as an inventor of next generation learning technologies. Heffernan developed intelligent tutoring technology that is bridging research and practice.  He is taking on Benjamin Bloom’s “2-sigma problem.” Bloom found that one-on-one tutoring is two standard deviations more effective than a group lecture, and the problem is that one-on-one tutoring is cost prohibitive at scale…or at least it has been cost prohibitive.  Heffernan’s free-to-the-public ASSISTments platform offers “scaffolding” similar that provided by a tutor for students learning mathematics.  The system can be configured to identify common mistakes/misconceptions and give descriptive feedback to the student just at the right time, but it is not designed to replace teachers, rather it supports a teacher-directed blended learning model.  As a research platform ASSISTments can identify common misconceptions and perform short-cycle studies on the effectiveness of possible interventions in alleviating those misconceptions.

So far we’ve identified three emerging roles in which college professors have gained greater impact, not to mention fame and fortune.  I’ll call them the “rock star”, the “edupreneur”, and the “innovator”.  There are many other roles critical to the success of the emerging eLearning models.  Other roles include the learning experience designer, the producer, the cinematographer, the game-based-learning developer, to name a few.   Online and blended learning models could bring about a renaissance for the education profession with new opportunities and opportunities for greater rewards. 



Disclosure: My current work includes facilitating the Common Education Data Standards (CEDS) an initiative of the U.S. Department of Education, and I’m on an advisory board for WPI’s ASSISTments project.

Thursday, March 14, 2013

Thursday, February 28, 2013

Online learning less effective?

Recent studies suggest that online post-secondary courses may have a negative effect, especially for some learners within some subjects, when compared to on-campus courses.  However, decision-makers should not interpret the findings to indicate that online delivery is generally less effective, only that the predominant models for online delivery currently used by the institutions in the study are less effective than currently used on-campus delivery models.

These studies may only confirm, for example, that video lectures 'broadcasted' online can be less effective than in-person lectures in which the lecturer can see the audience and gauge engagement/understanding.  The studies generally analyze data sets about students enrolled in online classes without differentiating online delivery models.  The online delivery model, especially for MOOC-style courses, is too often a less interactive substitute for a lecture series with homework and non-formative assessments.  There is one advantage of a streaming video lecture over the physical lecture hall experience, the learner can pause and review the material again.  However, the streaming video by itself does not have a feedback mechanism to gauge the learner’s understanding and adjust instruction accordingly.  Effective online learning processes, like effective tutors, track individual learner competency and continuously optimize the learning experience.  

The recent studies tend to compare the “place” in which learning takes place, i.e. online vs. in-class, rather than the models of teaching and learning used in the online and physical environments.  The findings raise legitimate concerns about the currently used online models and the implementation of those models at a time when societal and financial pressures are pushing universities more toward online delivery.  So rather than categorically discount online delivery, we should now ask “why” the models are less effective, and “what” are the characteristics of online models that are more effective. 

Decision-makers and practitioners need new research that focuses on the learning model rather than the delivery mode.  Online technology allows for interactive models of online learning that continuously monitor understanding and skills, provide timely feedback, continuously adjust the learning experience within the zone of proximal development, facilitate relationships for learning, and address motivational aspects of learning.  I suspect that the less effective online models are the ones that overlook the individualized feedback, relationships, and motivational aspects of learning.  I also suspect that, in many cases, the online models that address these aspects and leverage technology to overcome time/space/pace constraints will prove to be more effective than traditional fixed place/time course delivery models. 

Friday, February 8, 2013

Next Generation Learning Roles and Human Behavior


Moving to a learner-centric, competency-based, system of learning requires changed attitudes and behaviors. It is not enough to layer innovative models or tools on top of existing behaviors, people must change what they do. One of the most interesting challenges we face in the next few years has to do with the adoption of new roles, professional practices, and learner behaviors required by next generation learning models. We are at the beginning of the learning curve on figuring out what new roles and practices are needed.  The greater challenge will be scaled adoption of these new roles and practices.

This is a good time of year to ponder the topic behavioral change as many of us are trying to better our lives by keeping New Year’s resolutions. Knowing what to do is not enough, habits of practice are difficult to change.  Some observations about the knowing-doing gap:
1.       A change in behavior often requires a change in belief.  Habits of behavior are captive to habits of belief. A new behavior is constrained by what people believe about the problem, themselves, and their ability to overcome the problem. 
A first step is to provide a path and conditions leading to “aha moments” in which people to learn, often by experience, that:
  • It is in my best interest to change what I’m doing.
  • I have the ability to change what I’m doing.
  • If I do this differently, my life will be better, more meaningful, and/or I will better fulfill my calling. 
2.       A change in behavior often requires multiple points of motivation.  It usually takes more than one “aha moment” to adopt new habits of belief and behavior.  It takes a change in priorities over time.  Something needs to become more important and/or less important.  Each person may have a different set of motivators that drive the behavioral change.

Motivational factors include:
  •  Purpose, Meaning, Calling
  • Accomplishment
  • Ownership & Possession
  • Feedback / Course Corrections
  • Social Pressure
  • Scarcity 
  • Impatience
  • Curiosity
  • Avoidance of Loss
Simulations and online game based learning experiences may be designed to include all of these motivational factors.  However, some motivations may be more effective if sources from the real world, e.g. social pressure leveraging a person’s real social network may be more effective than from a simulated peer group. 

3.       Feedback loops are critical for learning and adopting new behaviors.  It is easy to slip back into old habits without long-term, ongoing supports.  Even painful habits are “comfortable” because they are familiar and people will slip back into a destructive habit without continuous feedback loops.  The most effective feedback will be just-in-time, provide the right level of challenge (within the zone of proximal development), and key into the individual’s interests and motivational “hot buttons.”  For example, an automated reminder to exercise may not be as effective as a friend waiting for you at the gym or even sending a private message via Facebook to ask if you exercised today.  Even more effective may be a social network of 20 people working on a goal of 1000 hours of exercise due to the peer pressure for each participant to contribute to the collective goal.

Bringing next generation learning models to scale can be supported by scaled infrastructure, the right economic conditions, technology, and policy enablers.  However, more is needed to help educators and learners to change what they do.  A path to new roles and practices is paved in part with redesigned professional training and development programs for specialized educator roles that develop proficiencies and practices optimized for new learning models.  Effective programs will likely take advantage of digital learning and social networking technologies, and designed to incorporate individualized motivational feedback loops.

Monday, January 28, 2013

Professional Specialization for Educators Emerging from Next Generation Models

Online and blended programs are changing the learner experience by flipping, flexing, and rotating the classroom experience around online/individualized learning. Educators in these programs are adapting professional practices to support the new models, sometimes figuring it out as they go. Some of the more innovative school models are abandoning typical features of schools, such as schedules, grade levels, and classrooms, in order to optimize a learner-centric design. Some models also recognize the need for new professional roles to support the different uses of time, physical space, and modes of learning.

One example is the model used by Cornerstone Charter Health High School in Detroit. The school, a Next Generation Learning Challenges grantee, recognized the need for professional specialization to support individual student learning absent of grade levels and class schedules. Instead of one-size-fits-all classroom teachers, Cornerstone has specialists:

  • Relationship Managers ensure students set and meet their daily, monthly, and yearly goals. Similar to a traditional guidance counselor, relationship managers follow a student from enrollment to graduation, helping students craft their individual learning plans and use student data and feedback to ensure students stay on track toward their goals. Relationship managers are the primary contact for parents and guardians. 
  • Relevance Managers provide direct instruction and support students in the design and evaluation of real-world projects and internships. 
  • Rigor Managers oversee online coursework, providing support and setting standards for mastery. 
  • Success Coaches work to help students make the transition to college and career, providing practical advice as students consider life after graduation. 


It is too early to tell whether this is an effective model of professional specialization for the student-centric model, since the school just started in 2011. Other organizations are piloting other models in hopes of optimizing the use of human resources for optimized student learning. Some universities offering scaled online courses have teams of instructional designers, licensing managers, counselors, staff trainers, adjunct professors, and process managers to ensure fidelity of implementation when different students will take the same course with different professors at different times.

This shift to learner-centric models and new educator roles raises questions about educator preparation and certification for these emerging roles.