Feedback: The Broken Loop in Higher Education – and How to Fix It
Learning science and teaching practice agree on the power of feedback to enable and enhance learning. So why is feedback so spotty and ill-timed in higher education?
Feedback is a proven enabler of learning. Yet in higher education today, it is spotty, ill-timed or utterly missing. Here’s what to do about it.
WHEN WE WANT TOLEARN a new skill in life, we try out a new behaviour. Some behaviours fulfill their intended purpose, while others do not. Feedback — signals from the environment that tell us whether or not the behaviour we produced had the intended effect — is essential to changing, adapting or modifying human behaviour. This is what learning is all about.
Whether or not a skill can be reproduced by an algorithm, learning any skill requires feedback and is essential to the measurement of learning. In the skills environment of the Fourth Industrial Revolution — wherein information is free and complex, interpersonal skills whose development requires textured, precise, timely personalized feedback have become the highest value contributions to human capital. The result: Feedback has become the critical missing link in higher education.
Learning Science and Practice: A Picture Emerges
The science of learning and teaching offers abundant evidence of the critical link between feedback and learning. In recent years, it has come to focus on identifying the right kinds of feedback for different people and learning environments: Whether you are learning a foreign language or a computer language; learning to supress impulses; or learning to communicate coherently, empathically and responsively — each requires specific types and sequences of feedback.
Timeliness, precision, intelligibility, actionability, repetition — all represent features of learning-enhancing and enabling feedback across different domains of knowledge, skill and expertise. The discipline of machine learning has made rapid advances in the last 10 years precisely because of its use
of fast mechanisms that allow algorithms to ‘learn’ from their own performance via feedback that tracks their successes and failures in replicating or predicting the data sets they are trained to compress and replicate (or ‘understand’).
Despite the momentous advances in understanding the role that feedback plays in learning, professional and higher education are lagging dangerously behind what is now both possible and desirable. The ‘lecture-homework-quiz-exam’ routines that pervade higher education — whereby feedback is provided en masse — lag student performance by a long time and are not adaptive or personalized to the learner or to her task. As such, current teaching practice — and the learning environment it produces — lives in self-sufficient isolation from the findings of learning science, deep learning science and engineering and the neuroscience of learning regarding the impact of feedback on skill and competence development.
Today’s feedback practices resemble those in effect 50 and even 100 years ago — an inertia driven partly by the economics of higher learning and partly by the cultural imperviousness of pedagogical practice to learning science and technology. But the opportunity costs of this ‘knowing-doing gap’ are very high and rising quickly: This gap represents both a significant drag on the learning curve of students and an important opportunity for disrupting the $2 trillion (2016 dollars) higher education industry. It is one that some organizations, as we will see, have already laid the groundwork for.
Missing and Counterproductive Feedback Patterns
To understand how the current feedback landscape of higher education fails learners by falling short of state-of-the-art learning science, let’s return for a moment to the century-old lecture-homework-quiz-exam routine that is the central model for learning today.
Lectures present concepts, models, methods, heuristics, along with their derivations thereof and applications. Homework problems, quizzes and exams often test for the normative or correct application of a skill or method to an unfamiliar problem. Feedback on the exercise of the skill by the learner is, for the most part, given by teaching assistants on problem sets, quizzes and exams turned in by learners — in batches, and days or weeks following the completion of the work. This is the exact opposite of what learning science tells us about feedback that maximizes skill development. Specifically, feedback in our current model is.
• OO LATE: Graders typically take days or weeks to deliver feedback to learners, in sharp contrast to the results of studies that indicate the importance of immediate feedback in the development of skill.
• TOO RARE: Feedback is infrequent relative to both the weight it should receive vis à vis other learning activities (such as listening and taking notes) — given its importance to learning, and especially to the learning of complex skills. An artificial neural network can ‘shatter’ — or, learn to classify — a large data set containing lots of non-linear relationships only if it receives profuse feedback about its performance as it ‘learns’. Why would a real neural network be any different?
• TOO IMPERSONAL OR GENERAL: The feedback the learner receives is not adaptive to her specific patterns of thought or behaviour. Because of the ‘economics of feedback’ in higher education, there is little time or scope to adapt the feedback to the learner’s specific goals and stock of existing skills, which significantly decreases the actionability of the feedback for the learner.
• TOO IMPRECISE: Learners usually receive ‘0 or 1’- type feedback (i.e. correct/not correct) relating to the degree to which they answer a question or solve a problem as a whole — but not on the specific pitfalls of the thinking or reasoning underlying an incorrect or partially-correct answer. This makes the feedback signal difficult to interpret as an action-guiding and behaviour-correcting input.
• TOO NOISY: Much of the feedback learners receive is heavily dependent on the rapidly-changing and idiosyncratic biases, moods, dispositions and physiological states of the graders. Different graders can disagree sharply on the quality of
Machine learning has made rapid advances in the last 10 years due to algorithms that ‘learn’ from their own performance, via feedback.
a particular piece of work, and a single grader’s feedback can be more or less favourable, precise and cogent at different times of the day, before or after meals, and before or after sleep.
• MIS-CONSTRUED AND MIS-CONSTRUCTED: Most of the feedback in higher education is given — and interpreted as having been given — for purposes of evaluation, filtering and selection, as opposed to being oriented to learning and behavioural change. It is evaluative and selection-oriented — something that educational research has steadfastly shown to undermine the effectiveness of feedback as an enabler and facilitator of learning, which developmental feedback encourages and facilitates.
The Right Feedback at the Right Time
We currently have the means at our disposal to fix this ‘broken feedback loop’: Converging models and evidence from cognitive science, deep learning theory and practice, and the neuroscience of learning (together making up ‘feedback science’) document the qualities of feedback that is maximally conducive to learning for most skill sets. We can — and should — turn this knowledge into a set of principles for the design of feedback protocols that fix the broken loop.
Not all feedback is equally useful or good; and some is actually counterproductive, uninformative and useless. What kind of feedback is most useful to learning? Some of the answers are intuitive, others less so. Learning-enabling feedback is:
• TIMELY: It follows promptly in the footsteps of the learner’s behaviour. Feedback given in a week is far inferior to feedback the next hour or the next day. In fact, neuroscientists have found that for cognitive tasks—like learning the grammar of a moderately complex language—instantaneous feedback trumps feedback that is given even a few seconds later;
• SPECIFIC: Feedback that enables learning is not general or fuzzy. It does not evince the cluelessness of currently common grading practices, in which the grader struggles for something meaningful to say to justify a letter or number grade arrived at on account of causes that have nothing to do with the reasons given for the grade. It is specific to the following:
• To behaviour or output — to the details of the learner’s written answer or verbal and non-verbal behaviour, and to the components of the output that can be usefully modified.
• To the context in which the written answer or verbal or non-verbal behaviour is embedded. Good feedback points out, for instance, ways in which the learner misconstrued the situation or the question.
• To timing — to the order or sequence in which the learner’s answer or verbal or non-verbal behaviour occurs. Good feedback singles out the specific points in the learner’s pattern of reasoning or behaviour that make the greatest contribution to the quality of the work. If a learner cannot differentiate continuous functions, for instance, and taking derivatives is an integral part of the chain of reasoning that leads to the right answer on an equilibrium calculation problem, then feedback that promotes learning should single out the learner’s skill gap in differential calculus.
• To the learner herself — to patterns of reasoning, calculation or behaviour that are specific to the learner’s own way of thinking or being. Good feedback is not generic — it is highly tuned into the learner’s patterns of thinking and behaving.
• To the consequences of behaviour or output and their interpretations. Good feedback on interpersonal, social or relational tasks points out the consequences of the learner’s behaviour on others’ feelings, behaviour and likely thoughts, allowing the learner to make textured inferences about the causal chain that links her behaviour to their social consequences.
The highest-value tasks performed by humans have become predominantly social, relational and interactive.
• ACTIONABLE: Good feedback provides prompts for behavioural or conceptual changes that are intelligible, clear and executable by the learner. It does not merely provide an appraisal of how successful an answer or behaviour was, but also a set of suggestions or injunctions for changing thought or behaviour patterns which are likely to lead to a better result;.
• CREDIBLE: Good feedback is persuasive to the learner in virtue of being:
• Legitimate. It is connected to the learning objectives of the course or module or learning experience and to the learning objectives of the learner;
• Justified. It is buttressed by valid reasons, drawn from disciplinary research and/or research on optimal learning;
• Objective or impartial. Good feedback can be validated by others of comparable expertise to the feedback giver, and is not thus prone to personal biases that render it partial or unfairly slanted.
• CREDIBLE: Its intent is to help the learner improve her performance on a task, or enhance her skill or competence in a domain — rather than merely to provide an ordinal or cardinal ranking of learners’ effort and talent levels for the purpose of providing discriminant value to recruiters or other programs of training.
• ITERATIVE: Good feedback is not a one-shot deal. It proceeds in iterative fashion. Just as neural networks and automata learn from multiple rounds of feedback that build on each other, learners require sequences of feedback sessions that help them refine their skill or capability.
• RESPONSIVE: Good feedback is responsive to the learner’s objections or interpretations of the feedback. It is neither opaque nor definitive, even if and when it is legitimate and impartial.
Two Routes to One Big Opportunity
As indicated herein, the current system of professional and higher education is very far from embodying the insights of feedback science. Given the foundational importance of feedback to learning and the gap between current and optimal feedback practices, we are faced with a significant opportunity to make a $2 trillion-industry massively more effective by changing its feedback practices.
What if the learning outcomes that the current lecturehomework-quiz-exam course achieves in 25 hours of lectures and 50 hours of homework and testing can be replicated in a feedback-intensive environment with just four-to-six hours of learner-teacher time? The opportunity is significant both educationally and financially. Several organizations with Facebooksized revenue streams could live well from even a 10 to 20 per cent reduction in the costs of education driven by changes in feedback practices.
There are two routes to the realization of this opportunity, and both are likely to emerge and develop within the next five years. Each has the potential to radically change the way teaching and learning are done. The first makes use of the semantic, dialogical and conversational capabilities of AI agents and enhanced formal and natural language-processing technologies, while the second relies on a new generation of teachers and educators making feedback the centerpiece of their curricular designs and teaching plans. Let’s take a closer look at each.
1. FEEDBACK BECOMES ALGORITHMIC. Walking in the footsteps of IBM’S Watson and Bluemix, and making use of deep learning ecologies of algorithms and platforms like Google’s Tensorflow and Microsoft’s Cognitive Services, adaptive feedback agents (AFA’S) will take the learner’s ‘stream of thought’ attempt to solve
a problem and give targeted, immediate, specific, objective, accurate feedback on each step of that learner’s process of reasoning or calculation, along with suggestions for remedial exercises or drills that develop each sub-skill or competency required for the successful execution of a task.
Powered by a database of questions, problems, answers and solutions from some 58 million learners taking some 13,000 massively open (MOOC) and small private online courses (SPOC) offered by 700 universities around the clock, AFA’S will be trained to address patterns of errors, idiosyncrasies and reasoning styles that learners exhibit. New results from feedback science can be embedded into feedback practice via updates to algorithmic platforms without the need to train up armies of teaching assistants and graders. Feedback can thus be liberated from the fluctuations of quality, mood, resources and acumen of human graders, for those skills that are sufficiently explicit and cognitive in nature to be tracked by algorithmic agents.
2. THE FEEDBACK-CENTRIC LEARNING FACILITATOR EMERGES The Fourth Industrial Revolution is not only one in which many tasks previously performed by humans can be performed by algorithmic agents hooked up to server farms, but also one in which the nature of the highest-value tasks performed by humans have changed, becoming predominantly social, relational and interactive.
Eighty per cent of the work managers now do in organizations is performed in groups and teams, and hence, the skills most prized by organizations are communicative and relational in nature. They comprise as many and even more affective skills (empathic accuracy, expressiveness) and executive skills (like problem structuring and quick task switching) as they do cognitive skills.
With affective computing still in a turbulent — though promising — infancy, there is a need to rapidly develop the language and base of expertise for giving feedback on interpersonal, relational and communicative ‘genres’ of work — such as board presentations, sales pitches, negotiations, deliberations, processes of collaborative inquiry and debate — that will enable and foster real learning of skills that are (still) quintessentially human and very ‘hot’ in the labour market.
‘Communication skill’ is now used as a catch-all label, which makes the development of all of the skills that go into ‘communicating’ very far from the elaborate evaluation rubrics that have been developed over a century of practice in teaching and grading Calculus, Microeconomics, structured language programming or thermal system design quizzes. But progress on creating the practices that will promote the rapid acquisition and transfer of these in-demand skills requires that we think carefully about the semantic and syntactic (e.g. coherence and completeness) and dialogical and interactive (e.g. responsiveness, informativeness) aspects of the learner’s behaviour in a social context — and that our feedback practices reflect a much higher level of precision.
Mihnea Moldoveanu is the Desautels Professor of Integrative Thinking, Professor of Business Economics, Vice-dean of Learning, Innovation and Executive Programs, Director of the Mind Brain Behaviour Hive and Academic Director of the Self Development Laboratorytm and the Leadership Development Laboratory at the Rotman School of Management, as well as Visiting Professor at Harvard Business School. Maja Djikic is Associate Professor and the Executive Director of the Self Development Laboratory at the Rotman School of Management. Over the past six years, they have designed, developed and implemented feedback science-based learning in the Self-development Laboratory — the Rotman School’s personal development engine for its professional students.
Rotman faculty research is ranked #3 globally by the Financial Times.