figure a

This chapter focuses on what the MELT mean for education theory and practice. A number of contemporary learning theories (in addition to those discussed in Chap. 4) sit comfortably on the learning autonomy continuum of MELT. Some of these sit at the prescribed end, some at the unbounded end and others in the middle. What this means is that the MELT can function as a kind of conceptual glue that holds these often-competing theories in tension: by placing them on the same page, the intention is to open up the conversations between people who hold to each one. Then, based on this complementary view of educational theories, the second half of the chapter considers what MELT means for enhancing learning and improving curricula and pedagogy through teacher action research. This addresses one of the major issues identified in Chap. 1: how to use MELT to engage with educational theory in ways that make practical sense to educators.

5.1 Situating Contemporary Learning Theories/Ideas

This chapter focuses on four educational conceptualisations that have influenced teaching practice in recent years. These conceptualisations entail competing perspectives on teaching and learning and were chosen from a range of possible contenders because they sit fairly well on one end of the learning autonomy spectrum or the other and so provide a kind of ‘argument from extremes’ perspective. However, there are still substantial differences between those conceptualisations at the same end of that spectrum. These four conceptualisations are Meyer and Land’s Threshold Concepts [2], Sweller’s Cognitive Load Theory [3], Siemens’ Connectivism [4] and Schön’s reflective practitioner [5]. These three theories and Schön’s conceptualisation have influenced education at various levels, from the national curriculum level, to the level of teacher decisions on a minute-by-minute basis. Because of their history and potential future influence on practice, seeing how these theories and ideas may connect together from the perspective of MELT is vital.

5.1.1 Threshold Concepts [2]

Threshold Concepts (TCs) are the pivotal learning concepts in each subject or discipline. Within Meyer & Land’s theory, these concepts are difficult for students to grasp, making it hard for learners to ‘enter through’ into discipline-specific ways of seeing the world. For example, in physics, the fundamental concept of inertia (that objects keep moving at their current speed unless a force slows them down or speeds them up) seems counter-intuitive to many. Without understanding the concept of inertia, it is impossible to comprehend Newtonian physics, even if students master the equations associated with motion. TCs are usually difficult or troublesome for learners, and pedagogical structuring is frequently required to enable students to cross the threshold of understanding. The learning involved in crossing this threshold may be frustrating and frightening, and the concept to be learned can seem alien or wrong. In the case of Tara in Shrink, the correct explanation of phenomena was opposed to her own interpretation, and so she deemed it ‘stupid’. Sometimes, threshold concepts are very difficult to understand in ways that can be operationalised, as in the case of ‘controlled variables’ in Parachute. Sometimes they involve arbitrary conventions represented in absolute ways, such as the left-hand numeral in a two-digit number being worth ten times its face value (as in Place Value). Once a TC is crossed over, the concept often seems ‘obvious’ to the learner, after which it is easy to forget the struggles that such learning requires.

As ‘thresholds’, such concepts are considered non-optional entry points into learning. Without a fundamental understanding of these concepts, subsequent learning will be superficial and may eventually collapse. In Shrink, the idea that metals shrink when heated would ultimately dismay Tara if, for example, she built a brick cooking place with a metal plate measured to fit snugly between the bricks. She ‘knows’ that the metal plate will ‘shrink when heated’, leading her to believe that there is plenty of room for the plate. However, in reality, the plate will expand, buckling or breaking the brickwork. If Tara were to become a designer working with metals—a jeweller or architect, for example—her misconceptions could have dangerous and expensive consequences.

Threshold concepts are by definition conceptually demanding, and this demandingness needs to be interpreted according to the age and experience of the learners. For example, an MBA student who has twenty years’ industry experience may find it easy to grasp the concept of ‘distributed leadership’, while MBA students with no professional experience are likely to find the concept much more difficult. Conversely, people with twenty years’ experience of top-down leadership may find the concept of distributed leadership an even more difficult threshold to cross. In this latter case, where a TC is counter to a person’s experience, educators may need to plan for the student to ‘unlearn’ an old idea before learning a new one. If there is one thing that is clear from the literature, it is that prior understandings are resilient and difficult to displace [6]. Awareness of TCs helps educators to consider students’ existing preconceptions; as such, educators will be less likely to allow students to stumble around with incorrect ideas.

TCs help to explain why students need more than an internet connection if they are to become adept at a particular discipline. Accessing information can lead to a bewildering array of relevant or irrelevant knowledge. Without grasping fundamental threshold concepts, students risk building incorrect models. Threshold concepts require knowledge that is frequently best attained through repeated exposure, experience and educative guidance from an experienced hand.

In terms of MELT, the educative guidance required to cross over thresholds (from a TC perspective) falls within the prescribed and bounded scope provided by educators. Here, modelling of and guidance into content and concepts are a part of the educational experience that students may need before they are ready for self-initiated discovery. Merely ‘telling’ a student about a threshold concept does not guarantee that they will internalise it. True understanding of a TC requires more than hard thinking; it may require every facet of MELT, including the application of current ideas and observation of where these fall short. Most commonly, this happens in an environment of low student autonomy, where teachers provide students with a highly prescribed learning environment and the students emulate. This can be a good way to help many students move across a threshold, because the teacher will be aware of conceptual sticking points. However, if a lesson pertains to a threshold concept that some have already grasped and others have not, there can be frustration. If the teacher assumes the TC as background knowledge, those who have failed to cross the threshold will feel left behind. Conversely, if the teacher approaches the TC as something that all need to learn, those who have already ‘crossed the threshold’ will feel bored. In such cases, it can be necessary to provide students who understand the TC with more scope for their learning autonomy.

Once attained, TCs allow students to operate in a whole new way and, in terms of the MELT, the attained concepts enable students to operate with higher autonomy, more ownership and empowerment, until the next TC is encountered. MELT’s perspective on autonomy, as unpacked in Chap. 2, suggests a shuttling back and forth between low and high student autonomy. Shuttling back to lower learning autonomy (where teachers prescribe and/or students emulate) shows a recognition of the need for students to be guided through new TCs.

5.1.2 Cognitive Load Theory [3]

Cognitive Load Theory (CLT) considers the complexities of students retaining, processing and applying information in sophisticated ways. CLT builds on the fairly stable idea that human brains dedicate a lot of storage to long-term memory (LTM), but that there is also ‘working memory’ (WM) which has very short-term retention of new information and limited storage. WM is a key place where new information is processed, as well as connected and applied to information stored in LTM. Currently, it is thought that humans only deal with around three chunks of new information simultaneously, but that they can also draw simultaneously on their almost unlimited LTM. People learn new information using their senses, and in educational contexts, visual and aural inputs predominate. CLT theorists talk about the ‘visual scratchpad’, which has a memory of half a second, and an ‘auditory loop’ which can retain information for up to thirty seconds. If this information is held temporarily in the WM, it can be juggled conceptually.

There are several instances in Place Value where the cognitive load is associated with conflicting concepts of value accorded to left–right positions of numbers. This was in evidence when the teacher asked, ‘is there any other number in the wrong place?’ At first, the whole class called out ‘no’. There were multiple loads on their working memory, and together, these may have been partly responsible for the group’s unanimous incorrect answer. Some of the cognitive load at that point relates to the question asked early in the lesson: ‘Is place value on your right or on your left?’ Within a two-digit number, the digit of higher value is on the left (e.g. in ‘44’, the left-hand four is worth forty and the right-hand four is just worth four.). However, in another mathematics convention, numbers of a higher value are placed on the right in a sequence (e.g. 9, 14, 27). This right–left difference for these two conventions is hard enough to grasp when students are learning about double-digit place value for the first time, but much more so when some of these six-year-old students may not be clear about which side is left. In Place Value, this set of inherent complications is magnified: students on the italitali mat viewed the number sequence back-to-front—it decreased from left to right because it was sequenced from the perspective of the eight students that were facing them. To students on the mat, it appeared as ‘57, 41, 29, 20, 17, 11, 9, 5’. With this extra extraneous load on each student’s WM, the processing time required for the analysis became longer, increasing the risk of inaccuracy, confusion and a feeling of dissonance, all of which may provoke early maths anxiety. In effect, the ‘grammar’ of mathematics, like any language, is complex and often has internal inconsistencies which can overload WM. Not until students internalise this grammar can the load on WM be lessened when students are dealing with two-digit numbers and be more able to, say, add together two-digit numbers.

We can only process small amounts of new information at a time. These small chunks of information build-up and eventually become part of our personal knowledge base. For short-term memory to be consolidated into LTM requires time, rehearsal and the application of information.

From a CLT perspective, the problem with providing students with higher levels of autonomy is that their STM can become overloaded with new bits of information. Like threshold concepts, CLT provides a rationale for laying down the foundations of a knowledge base, with minimal distracting elements, and slowly building up opportunities to apply that knowledge. From the perspective of these theories, teachers should not consider allowing students to initiate until students have demonstrated a mastery of the appropriate content. Shelly struggled in Parachute because she did not have the necessary basic concepts in place, such as the formula to calculate the area of a square. Kevin’s success in Silver Fluoride is based on the fundamental knowledge and skill base that he developed during his degree, and which therefore prepared him to initiate sophisticated learning.

CLT and TCs are contemporary learning theories/ideas that provide an impetus for learning to be engineered, prioritising the lower end of the learning autonomy continuum, where teachers prescribe and students emulate. But another current and influential theory sits at the opposite end of the continuum of learner autonomy: connectivism.

5.1.3 Connectivism [4]

Connectivism focuses on knowledge being distributed in networks and doesn’t differentiate information from knowledge and learning. Siemens positions ‘knowledge’ as a construct that can exist externally to a learner. Information and knowledge can be stored, for example, in papyrus scrolls and hand-copied books. Information and communications technology (ICT) merely speeds up the sharing of information, as did the move from hand-copied books to the printing press. However, modern ICT is qualitatively different from those technologies in that it also enables the consumer to be a producer, and to dynamically engage with and change the knowledge representation. This can involve commenting on, modifying or creating from scratch, uploading a video, or compiling mash-ups.

For Siemens, the goal is to put knowledge into action at the point of application. Where the student lacks that knowledge, he or she draws on ‘the ability to plug into sources to meet the requirements…’ (p. 2). When knowledge is not the Bloom’s Taxonomy-like foundation for learning, but rather the learner’s purpose is, control shifts from a knowledge-giving teacher towards the learner. For connectivism, then, knowledge is not a fundamental basis for learning. Rather, knowledge is an enabler, a part of the learning process, not the beginning of it. Importantly, knowledge is, in effect, injected when needed in the learning process, and this resonates with non-sequential, multifaceted learning. For example, if a student is exploring how to code a robot so that it is able to walk up a ramp, the student may go online, or visit other knowledgeable students in the class to get that information at the time the student perceives she needs it. This could be contrasted with the teacher providing all class members with the information that he thinks necessary about uphill coding, at the time he thinks it needed. For MELT facets, connectivism elucidates that ‘clarify… the knowledge required’ does not necessarily come first. It may be well into an investigative process before further knowledge needed is factored in by the learner themselves.

According to Siemens, the acceleration of knowledge production means that current knowledge becomes quickly outdated. This short life expectancy of knowledge currency means that learning ‘can now be off-loaded to, or supported by, technology’, contrary to theories like CLT and TC, which focus on students’ cognitive processes. ‘Know-how and know-what are being supplemented with  know-where (the understanding of where to find knowledge needed)’ (p. 2). From the MELT perspective, finding others’ information requires all six facets, especially clarification of what is sought, methodologies to find and the interrogation of sources and their content. Siemens further probes the times when knowledge is not the starting point for learning: ‘How do learning theories address moments where performance is needed in the absence of complete understanding?’ (p. 5). In contemporary, interconnected learning, there is certainly a case for high levels of autonomy and prioritisation of just-in-time searches to get the job done, and this is true for professionals as well as students.

Connectivism emphasises interconnectedness among students, as well as between students and educators: ‘We can no longer personally experience and acquire learning that we need to act. We derive our competence from forming connections of each other’ (p. 2). Students connect to each other in a way that is analogous to the connection between partners in a business. The bigger the business, the less each partner knows about the whole enterprise, and the more specialised individual knowledge becomes. The company’s knowledge resides in the interconnectedness, in the networks between different employees, as well as between employees and knowledge-containing devices. At face value, for MELT, this interconnectedness enables foregrounding the self-determination of what to know, how to know it, and therefore of higher student autonomy. However, given the nuances of autonomy, such enabling really depends on the nature of the relationships, and if there are any educative relationships.

5.1.4 Schön’s Reflective Practitioner [5]

Schön’s reflective practitioner (RP) is an individual who has powerful internal resources at hand. Such an individual has the capacity to reflect on the action and teach themselves as they go. In some ways, the RP concept is the opposite of connectivism, in that the RP’s knowledge is not distributed and networked, but localised. In other ways, RP and connectivism are complementary, as they both defer towards higher learner autonomy (improvise and innovate). Within both perspectives, student ownership of the learning is paramount.

Schön’s two big self-teaching tools are reflection-in-action and reflection-on-action. Reflection-in-action occurs when a professional faces indeterminate decisions, that is, ones in which the correct choice is not obvious. According to Schön, practitioners make off-the-cuff decisions using a detailed and nuanced ‘know-how’. Whereas some may write off this idea as a ‘gut-feeling’, Schön considers such decisions to reflect an underlying expertise and experience. Like practitioners, students can learn to take what they know and act. However, Schön’s second reflective component is a vital component of this process.

The second component of Schön’s model loosely corresponds to the standard view of reflection: reflection-on-action. This type of reflection is carried out after the effect, sometimes spontaneously (in musing or conversation) and other times systematically (in diarising or recording). Reflection-in-action can be done with specific frames of reference in mind, such as conceptual frameworks or theoretical positions (like CLT or connectivism). The ‘reflective surface’ one uses for reflection determines the character of those reflections. For Schön however, the emphasis of reflection-on-action is the process of looking back at the indeterminate decisions made through reflection-in-action, and teaching oneself. Therefore, knowledge external to the learner and others’ theories are secondary to internal decision-making, which provides a powerful impetus to learn. Together, reflection-in-action and reflection-on-action make Schön’s Reflective Practitioner a strong learner, able to improve and exercise high levels of autonomy.

At the same time, not only do each of the theories sit on the continuum of learning autonomy, but also in their various foci, each requires all six MELT facets. For example, given MELT’s ancestry of Information Literacy standards, with their focus on ‘information-seeking behaviour’, the know-where of connectivism requires all six facets of MELT. As stated earlier, know-where is ‘the understanding of where to find knowledge needed’, and in many ways it is a form of ‘know-how’. One reason for this is that ‘where’ is not merely the location of a source, but the level of relevance and trustworthiness possessed by that source. Likewise, ‘know-where’ cannot connote something effective for learning if it were to merely ‘relocate’ knowledge from one position to another. In addition, there must be analysis and synthesis that are dependent on effective organisation and that are enabled by, and lead to, effective communication. Across all these MELT facets, the students increasingly need to clarify their purpose, especially if the beginning is vague due to a lack of clarity from teacher instructions or due to the student’s own lack of specificity. As noted in Chap. 2, in the flood of irrelevant information, clarity is power [7]. Einstein makes the same point on the Chap. 3 title page: ‘Out of clutter bring simplicity’ [8]. To paraphrase and synthesise both quotes ‘powerful thinking results from learning how to clarify the clutter’ and this requires all six facets of the MELT. Thus, the MELT perspective about sophisticated thinking is not so much ‘higher order thinking’ in keeping with Bloom’s taxonomy, but multifaceted thinking.

5.1.5 Corporately Destructive or Mutually Informative?

The four contemporary learning theories/concepts described above are current and appealing. However, when we compare them to each other, we find apparent tensions between them in terms of what teachers ought to emphasise. These tensions may be relieved by placing each of the theories/concepts along MELT’s continuum of student autonomy. Doing this provides an understanding which propels shuttling back and forth along the continuum. There are various challenges to each of the four perspectives above, especially from each other. While two are especially information and knowledge focused, one of these––Cognitive Load Theory––posits that learning occurs in an individual’s brain, while the other—Connectivism––defines learning as something distributed outside any one’s brain. The other two perspectives emphasise deep understanding, and where Threshold Concepts prioritise the inculcation of foundational concepts by teachers, the Reflective Practitioner employs and values learners’ existing resources and intuitions. Conceptualisations that have overlap, then, occupy ‘competitive spaces’, and so, in effect entice adherents to warn about the perceived dangers or deficits of neighbouring competitors. For example, one article focused on the new learning taking place due to learning technologies is titled ‘A Pedagogical Paradigm Shift from Vygotskyian Social Constructivism to … Siemens’ Digital Connectivism [6]. This ‘shift’, demanding a conceptual migration, shows that the notion of ‘academic tribes’ [9] is alive and promotes competition fostered in defending or expanding the conceptual territory. This competition is useful if one theory or concept may be or become the best, most parsimonious, most informative for teaching and learning. Given the histories of education and most disciplines, this is not likely, and so treating theory as literal, as really describing the learning condition, rather than as more metaphoric in nature, diminishes the potential for mutually reinforced connections that the MELT continuum of learning autonomy makes possible. This competitive stance of territorial tribes has until now merely diminished and made disparate the energies in education for teaching, learning and research and reduced their effectiveness but could, in the near future, have more sinister effects.

5.1.5.1 Epistemologies of Machine Learning

Disparate and competing views make the educational enterprise, broadly speaking, more vulnerable to vested interests and to the emerging new world order, especially that mediated by machines. Just like the theorists above, humans currently working on AI and Machine learning are working on issues of epistemology. Machine Learning has to be on some learning platform or other and the issue of epistemology––or how learning happens––dynamically influences and is influenced by that platform. At a mechanical level, some AI learning platforms mimic the neuronal architecture of the human brain, whereas some are altogether different. Quantum computers will probably be a key component of AI operating systems, with possibilities including ‘a neural network encoded in the quantum properties of light’ [10]. We do not know what will comprise the intelligence of AI because the possibilities are numerous and broad, and include hybrid versions of old-style analogue computing and quantum computing [10]. With the advances in quantum computing, it is difficult to know which platforms or combinations will learn best in which situations. At a processing level, it has been known for a long time that machine epistemology is or can be very different from human epistemology [11] but because the structure of AI will not be clear for a long time, the epistemologies of AI will likewise remain unknowable until they fully emerge or diverge. Well before we can understand Machine Learning, however, we will have to understand Machine Pedagogy, for this will dynamically influence student learning in classrooms.

5.1.5.2 Epistemologies of Machine Teaching

AI used in schooling and university education for teaching will face unresolved epistemological questions just like all AI, but it is the pedagogy that machine teachers use on children that will impact first. We do know that the machines that teach children will need to choose or have chosen for them learning theories that are suited to humans not machines. An AI teacher may derive human learning theories itself, discarding all those that have preceded, and use a grounded theory approach [12] because, after all, machines will use data from children learning, in order to formulate their theories of teaching. Alternatively, specific learning theories may be prioritised by some programmers of teaching machines. For example, one study found that ‘The results suggest that integration of certain behavioral theories as features in machine learning systems provides the best predictions’ [13] (italics added). Currently, students learning to read, as mediated by reading robots, seem to develop a strong bond to the robot, and these robots could  be programmed with a Social Constructivist epistemology where correction may be secondary to connection. A robot programmed with a behaviourist orientation, however, would favour correction over connection in order to provide correct stimuli leading to correct response. This raises an issue that has been endemic in education. What works ‘best’ depends on what you value in your measurements, with objectivists valuing, and so measuring, quite different things from constructivists.

In the short to medium term, it is likely that parameters for learning theories that machine teaching will operate by will be set by human programmers. This may result in some incredibly consistent teaching, and maybe even highly creative, varied machine teaching, engaging students within the boundaries set. But it also risks escalating paradigm wars to new levels, with AI programed to follow parameters that prioritise one end of the learning continuum or the other. It could result in unequivocal understandings about the superiority of a theory in a specific set of circumstances, and/or in vested interests competing for the commercialisation of their AI/paradigm package.

AI could, of course, be programmed with parameters that allow for and encourage the full spectrum of theories, interpreted along the MELT continuum of learning autonomy. The window of opportunity for educators to play a role in determining such parameters is closing and with late 2023 being when the earth is predicted to hit 8 billion people alive at the same time, that provides 2020–2023 to make decisions that will impact on the education of the next billion human brains. The main readership of this book, teachers of young children to supervisors of PhD students, have a little time to inject a deeper sense of humanity into the debate about Machine Teaching. Maybe parents and other citizens would like to think too about the ramifications of narrow or competing sets of learning paradigms for machine teachers. In addition, all of us have to wonder if we want a future where human teachers are increasingly irrelevant, and where maybe there is little point of human learners when machine learners get all the jobs.

Each perspective in 5.1.1–5.1.4 provides a currently useful consideration for engaged learning and teaching, each can be challenged by the other theories, and each can be placed on the extent of autonomy continuum. Placing them in such a way allows each theory to be held tentatively, without considering any of the four to represent ‘the truth’. Weighing together different educational theories helps us to understand how they can inform and improve engaged learning and teaching. When we bring in Machine Learning and Teaching, the stakes about differences in learning theories become much higher, mission critical. While we have human teachers, or maybe to keep teachers human, what are the insights that learning theories can provide to improving student learning, when viewed in a complementary way through the MELT?

5.2 MELT for Curriculum Design and Improvement

Using MELT to connect different learning theories has practical implications for student learning. Firstly, curricula need to have strong conceptual underpinnings. Such underpinnings help teachers to facilitate student acquisition of a contemporary knowledge base and investigate areas of interest in ways that develop sophisticated thinking and rigour. Secondly, curricula themselves need to improve and adapt over time in order to enhance learning and teaching. By using MELT, educators can hold a variety of perspectives, such as Direct Instruction and discovery learning, so that instead of conflicting, the perspectives can work in unison to inform curricula and their improvement. MELT reduces educators’ obligation to choose one perspective or theory only, and increases their capacity or willingness to hold several in productive tension. Such multiply-informed curricula, taken together as a set over time, may better scaffold the development of sophisticated thinking when compared to a set of curricula informed by a narrow range of strategies. This is only true if a way of connecting, like MELT, enables teachers to conceptually unite a variety of perspectives so there is a legitimate coherence between them.

Which curricula are best able to provide learning that spans learner autonomy, that enables the acquiring and construction of students’ own knowledge, and the simultaneous development of the skills associated with sophisticated thinking? The answer is not in the curricula, which vary enormously, but in how teachers bring a particular curriculum to life, improving it over time through planning, implementation and revision. Ongoing improvement of this kind is here called ‘action research’, and maybe an imperative to keep teaching human.

5.2.1 Teacher Action Research

Action research entails teachers intentionally enhancing the conditions for learning through spirals of action and improvement, in contexts for which they have direct responsibility. Much published educational research comes under this definition of AR, although such research is frequently labelled as something else. The component of the definition from which some people distance themselves is ‘have direct responsibility for’. One reason for this distancing is that a phenomenon for which the researchers have teaching responsibility entails a ‘subjective’ engagement rather than ‘detached’ observation. In terms of standards of objective research, subjectivity is a no-no, but in terms of curriculum improvement, there are many advantages to teacher-as-researcher, with a strong interest in attaining positive outcomes for student learning.

The rich contextualisations and nuanced descriptions possible in AR mean that it is more likely than many other research approaches to overcome the theory–practice gap [14]. Through AR, practitioners can put theory into action in ways that make sense to them. This means that the typical problem of ‘translation’ of research findings is minimised, because during AR, theory is translated as the research is carried out. AR should be communicated not as an account that is generalisable to any context, but with rich descriptions to enable another practitioner to read, understand and transfer thoughtfully to their own context [15].

Action research is a useful methodology for consolidating existing good practice, because teachers often try one intervention at a time, leaving most of their practice untouched. Some people with a strong policy orientation push for wholesale change [16] rather than gradual change, but this can be highly problematic for at least five reasons. One is that the useful, evolved and adapted knowledge of a community can be thrown out in big change, whereas AR is good for consolidating practices which already work. Second is that teachers can struggle to understand initiatives imposed by others, whereas AR typically is the teacher’s own initiative. Third, policies are easy to implement badly, but difficult to implement as intended. Fourth, different teachers have differing personalities and theories of education, both of which influence implementation, and so large-scale changes can lead to highly variable outcomes. Even providing a uniform curriculum to multiple classes through forced training and tight invigilation of teachers is very difficult, and can be counterproductive. Fifth, those who pilot initiatives and show good outcomes are often those who chose to be involved. There is a risk that the second generation of implementers demonstrates less effective outcomes than the self-selecting first generation [17]. Action research inevitably leads to change, as suggested by ‘action’. However, such change can simply mean the maintenance or tweaking of existing approaches. This may be one of the features that make AR effective.

AR accounts for who and what teachers actually are: people with emotions and a teaching sense. They may slide between professional pride and care to apathy, disengagement and anxiety. A motivational starting point in AR is for teachers to resolve an issue that is important to them. This builds on their existing thinking and can lead some to find a way out of anxiety or apathy. A disengaged teacher is an ineffective teacher, and telling teachers what to do more forcefully or with greater incentives does not always work. Giving a disengaged teacher some real say in design or implementation can be a way to improve their teaching. Degrees of rigour and sophistication need to be added to spirals of AR, which should increasingly reference quality criteria [18] and enact these over time.

The ownership of a teacher AR is empowering and has the potential to produce something akin to the Hawthorne Effect [19]. The Hawthorne Effect was named after a phenomenon found during research on productivity in the Hawthorne factory. Researchers found that while an increase in luminosity increased productivity, decreasing luminosity had the same effect. They surmised that the workers responded to the actions or presence of researchers, and this response was independent of lighting levels—rather, it was cued by the change in levels.

The researchers’ findings from the Hawthorne factory may have been unhelpful for employers seeking to fine-tune lighting levels. However, the findings are very useful for those who want to improve student learning in contexts for which they have direct responsibility. Students can tune into the fact that the teacher is doing research and trying to find ways of improving their learning. Teachers will be interested in the outcomes of their designed intervention. But the Hawthorne Effect suggests that any intervention, if perceived as an innovation or change in practice, alerts students to something, and this alert enhances their learning. As a result of the Hawthorn Effect, one of AR’s apparent weaknesses—subjectivity due to teachers’ ownership of the research outcomes—becomes a strength, actually enhancing student learning while AR is happening.

Typical AR spirals proceed as follows: identify problem or issue, plan an intervention, implement, evaluate and identify further cycles. But there are three problems with this standard sequential model. First, such a sequential representation does not capture the messy, recursive nature of actual research (including AR). Second, this simplified sequence does not reflect forms of research other than AR, and so it reduces the status of AR to something less than research, something you do if you can’t do anything else. Third, and most importantly, this neat and linear conceptualisation minimises the effectiveness of AR because it does not incorporate the conceptual prompts or challenges needed to increase rigour and sophistication of research and action over time. This compromises possibilities for improvement.

MELT’s six facets can help teachers inform their action research into curriculum improvement so that they add rigour and sophistication over time. From the perspective of MELT’s six facets, AR involves messy, recursive processes where teachers: identify the issue, problem or aspect that could be improved; imagine improvements, find information that could be relevant, find relevant data or generate new data; evaluate information for pertinence and for trustworthiness; organise information so that trends can be seen for detailed analysis and manage the time, resources and strategies used; look at information from other classes, the web or the literature and synthesise new, creative answers to address the issue; and weave communication and application of knowledge throughout the whole process.

MELT, if used as a scaffolding for the structure of teachers’ AR conceptualisations, has a four-fold imperative. Firstly, teachers can use MELT to inform curriculum design. Over time, this could include making connections to others’ curricula, for the purposes of planning and implementation. Secondly, when used as a thinking routine, MELT help teachers to facilitate student awareness of their own thinking. Third, teachers may use MELT skills to inform their own action research in terms of planning, implementing and evaluating improvements in teaching, learning and the curriculum. This enables an intersection of the learning, teaching and research through MELT-informed conversations between educators and researchers. Finally, by using MELT to inform their AR, teachers can come to understand how their individual research may better connect to others’ curriculum research.

5.2.2 Conjoined Action Research: From the Transferability of Individual Studies to Generalisability When Using an a Priori Framework

Educational researchers tend to value generalisable results over idiosyncratic research such as AR. This may be a big problem for education, because applying generalisable results to the classroom is difficult [20]. While the notion of generalisable results is highly appealing, methodologies designed to achieve such results have serious limitations. And while individual AR is not generalisable, each instance of AR has the capacity for transferability. Moreover, if there are ways that different instances of AR can ‘speak together’, then they may have elements in common that can, over time, reveal trends that are generalisable.

In education, generalisability is sought through Randomised Controlled Trials (RCTs). Whereas RCTs are often thought of as a gold-standard of research with humans, in education they have many limitations because they possess low ecological validity, are highly resourced and typically conducted in educationally insignificant timeframes [16]. Here, ‘low ecological validity’ means that the research is far away from ‘business as usual’ due to: the presence of researchers not normally on-site; extra resources or funding that may not be available once the research is complete; contact with parents and students to achieve informed consent; randomisation into treatment and control groups, creating artificial groups of students or temporarily isolating individuals; and a quick injection of treatment or control group learning protocols [18].

RCT designs in education typically take place over educationally insignificant timeframes, such as a lesson, a day or a week. The real-time measures immediately after an intervention are a time-bound way of capturing student learning. RCTs provide an immediate extrinsic motivation to engage in short-term learning and testing in ways that cannot easily be mirrored in the standard classroom. Informed consent normally means that some students and parents opt-out, and so the cohort is less representative than if all students were involved, and those who opt-out maybe those for whom the intervention is less suited. Moreover, even with those who consent, true randomisation is very difficult to achieve, due to timetable constraints.

A Nature journal survey of authors who published scientific, generalisable research found study results were not replicated in follow-up studies conducted by other researchers (70% non-replication) [21] or even by the same team (50% non-replication) [21]. This means that generalisability in experimental work such as RCTs ‘depends’ on the research team. RCTs within the complexities of education are unlikely to be more rigorous than medical RCTs which involve, say, an intervention requiring regular doses of a medicine.

Action research, on the other hand, is conducted in naturalistic ways and timeframes. Here, ‘naturalistic’ means ‘implemented in classrooms with levels of resourcing that are not added to by researchers’ budgets and collaborations’. The associated timeframes tend to be educationally significant: a term, a year or multiple years. Extended timeframes provide far more credence for classroom practice, because teachers engaging in AR have to deal with aspects that RCT approaches see as ‘confounding’, such as student motivation to engage long-term in the learning.

Action research is powerful because it is fitted to the specific context in which it is enacted, by the teachers who have responsibility and ownership of the learning environment. This leads to ownership of the research, a strong interest in positive outcomes and steering towards the best student learning possible. While the researcher has a vested interest to confirm that their own approach works in AR (compromising the study’s objectivity and generalisability), this is also AR’s strength, for positive learning outcomes provide an endorsement of teacher professional judgement. The teacher’s ownership of the research speaks to the heart of their educational prowess, a self-validation but not a generalisation to other contexts. Table 5.1 shows the features of AR when compared to RCTs.

Table 5.1 Comparison between RCTs and AR
Full size table

Analysis of multiple MELT-informed AR studies could inform not only individual implementers, but assist a move from transferable to generalisable research. A systematic review of studies about research-based learning found that the amount of guidance provided by teachers to students was reported in an ad hoc manner. The review’s authors, therefore, recommended that an a priori framework be used for reporting [22]. MELT’s continuum of learning autonomy provides a cross-studies language to describe ‘amount of guidance’, and along with the six facets, it provides an a priori framework that is both adaptable and readily useable for reporting. Meta-analyses of AR, if MELT facets and autonomy are used to connect multiple studies together, may show that these studies have some generalisable findings. MELT as an a priori framework for analysis and reporting can connect the reportable outcomes of otherwise non-generalisable AR, providing a trend over many studies.

Action research can begin with low rigour and sophistication, but as AR continues, it should demonstrate increases in rigour as part of the learning inherent in research as a process. A practical way of adding rigour is ensuring that AR researchers use and report on all of the facets of MELT. Increases in rigour can be assisted by reflection-in-action, and simultaneously by adding to personal, idiosyncratic approaches, those informed by the literature. As an example of increase in rigour, Table 5.2 considers pertinent questions, facet by facet, early and later in AR cycles of actions. With MELT to inform ways to lift rigour of AR processes, and to enable connections between different AR studies, this gives AR the means to improve student learning in ways that RCTs cannot. Multiple AR studies of contexts with educationally significant timeframes may over time show the benefits of interventions more effectively than RCTs. If MELT is used to frame the reporting, multiple AR studies conducted in naturalistic settings with little extra resourcing may coalesce on the same findings over time. This would then provide rich descriptions for reported AR outcomes so that they can be transferred to the classroom, and synthesis that shows trends across many AR studies and contexts.

Table 5.2 Questions and issues for each facet and level of learning autonomy in relation to commencing AR and maturing AR
Full size table

5.3 Conclusion: Multifaceted Use with the Same Overarching Purpose

MELT helps teachers to interpret educational theories by plotting them along the learning autonomy continuum. This allows them to connect to and use theories as part of their teaching repertoire. For example, if a teacher needed to choose between CLT or Connectivism to inform a contemporary curriculum design, each theory implies a very different design with correspondingly different learning opportunities for students. However, connecting these theories through MELT using the continuum of learning autonomy to inform when and how to use each theory can help teachers design a curriculum which is more holistic and helpful for engaged learning.

Action research, informed by one or several theoretical frameworks, can be conceptualised and reported in the terms of MELT. This use of MELT will not only allow different educational theories and perspectives to speak to each other, but will provide the scope to draw together separate instances of teacher AR and report the analysis of the trends and synthesis as more generalisable results.

The huge untapped resource in education is the sum of separate efforts. These efforts sometimes exist in conflict, but can reinforce each other through a consideration of the continuum of learning autonomy and the six facets of MELT. The conceptual power of MELT, as a consolidation of 100 years of educational research and 100,000 years of human learning, is the connections it forges. This power can enable students to connect all their learning together across the years and teachers to connect to each others’ curricula and to educational theory as they engage in systematic AR informed by MELT, in order to develop student sophisticated thinking.