For those who work in this field, “design for learning” may seem an awkward phrase. We prefer it to “learning design” for what may seem a trivial reason, but is actually crucially important. “Learning design”, like “experience design”, can be read as meaning that it is possible to design other people's learning. This is not the case (Beetham 2008). One cannot design someone else's experience. One cannot design their happiness. One cannot design their learning. Only the person who is learning can learn. Someone involved in the design for learning can design things that help other people learn. We say that this is crucially important for two connected reasons. First, recognising that there is a gap between:

that which has been designed; and

means that one can then try to analyse the relations between (a) and (b). Second, it is very rare for (a) to determine (b). One can argue that classical work in instructional design assumes (quietly) that (a) determines (b). In which case, there is no reason to try to see the relations between (a) and (b) as involving anything subtle or complex.

Understanding the relations between (a) and (b) mirrors a central problem in the human sciences – how to relate structure and agency. How should we understand that people make history, but not in circumstances of their own choosing? When thinking about design for learning, how should we consider the ways in which learners are influenced and supported by artefacts (etc.) in their learning environment, and at the same time, how their activity changes that learning environment (for themselves and others)? Design for learning often involves the creation of scaffolds, scripts and other supports for parts of learners’ activity – how should we think about the functioning of these scaffolds (the way they shape activity) and about the ways in which activity marks, erodes and rebuilds the landscape in which it is set?

If design for learning is to be part of a sustainable educational practice, then it should not routinely be thought of in time-limited, project-based terms. So large a proportion of innovation in education, especially where new technology is involved, is funded through short-term projects, that we do not have a shared professional sense of how design would look if it was a core part of the mainstream work of education. There are three consequences of this limited vision. First, it means that we tend to see design as something that begins with a fresh project. We do not so often think of design as a natural part of on-going cycles of educational provision and enhancement. Second, the imperatives of projects are such that people involved in them work hard for success. Among other things, they create a niche for the project, to protect it from the complexities of everyday educational life. This is one of the reasons that things, which work well, in the project's niche, tend to fail when the project funding ends and the innovation has to prosper (or not) among the unpredictable exigencies of the real-world. Third, and connected to this, it means that projects typically fail to plan adequately for life after the project's funded period ends. Crucially, the design work that dominates the life of a project tends to under-prepare for: (1) the complexities of embedding or enacting the design in new contexts, (2) the customisation or co-configuration that is done by learners and teachers to make the innovation work in new contexts, (3) supporting the work of teachers who have to make the innovation function, after the end of the funded project, and (4) periodic review, modification and/or redesign of the innovation.

We have picked this phrase for two reasons. Classically, it refers to a literary device, through which a narrative begins “in the middle of things,” rather than at what one might take to be the logical start of a story. (In Horace's Ars Poetica, in medias res is contrasted with ab ovo – literally “from the egg”; from the very start.) Most of the classic texts of instructional design start ab ovo. Agreed, they will often structure the process as a sequence of Analysis, Design, Development, Implementation, Evaluation (ADDIE), within which there are meant to be multiple feedback loops, but they nevertheless sit best within a project-based framework. We use in medias res to try to normalise design – as something that should have value, and be understood, within the regular on-going flow of educational activities. This is one of our key messages. But we also use the phrase for some secondary resonances – that design creates things and these things influence the activity that results in learning. In short, design needs to be understood from a socio-materialist perspective (Fenwick, Edwards, and Sawchuk 2011).

Understanding the relations between (a) that which has been designed and (b) the activity that results in learning is complicated (one might say confounded) by the uncertain role of teachers. The best way we can make this point is to ask you to imagine design for learning that takes place in situations where there is no teacher – just the learner and whatever the design process has helped create, and whatever else the learner brings to hand. Without the realtime (learntime) interventions of a teacher, the relations between (a) and (b) become clearer.

One of our concerns is that many approaches to “learning design” assume that a teacher will be around at learntime and that they (the teacher) will be able to improvise remedies to shortcomings in the design – filling in gaps, explaining details, etc. In some cases, what purports to be a design for learning is actually just a design for a teacher to enact: a lesson plan.

Now we need to make a subtle point. In our view, a comprehensive theory of design for learning needs to include the ability to describe a role for a teacher at learntime. A teacher can, of course, play a crucial role in making a success of a lesson. The distinction we want to make is between: (1) situations in which designs are (unconsciously) incomplete or underspecified, assuming that a teacher will be on hand to fill in details and make repairs on the fly, (2) situations in which designs explicitly treat the teacher as a knowledgeable, helpful actor, but with bounded capabilities. We think that many (weaker) examples of designs for learning are of type (1). We think that designs of type (2), which some call “design for orchestration”, represent an important subclass of designs for learning (Dillenbourg et al. 2012).

A comprehensive, coherent, defensible view of design for learning therefore needs to pay careful attention to a number of major issues that are not uniformly well-handled in the existing literature. A related issue, which we examine more closely in our companion paper (Dimitriadis and Goodyear, 2013), is that design thinking and design practice would benefit from being even more forward-looking. By this, we mean that design work should not just focus on what students are expected to do (in their learning activities), it should also include design for configuration (what students and other agents do, to prepare, customise or otherwise modify that which has been designed, to suit specific needs), design for orchestration (providing support for the teacher's work at learntime), design for reflection (ensuring that actionable data is gathered at learntime, to inform system evaluation) and design for redesign (consciously making design choices at time t, to make modifications at time t+1 easier).

The nature of the students’ activity is what affects whether or not they succeed in learning, and what they actually learn. Outcomes are dependent on activity. Different kinds of knowledge are acquired in different ways (Biggs and Tang 2007; Ohlsson 1995). It therefore makes sense to put student activity at the heart of things: to embrace “activity-centred design.” As Lehtinen (2011) has pointed out, many of the things we ask students to do involve (and result in) complex learning – in which multiple forms of knowledge and ways of knowing, at various grain sizes, have to harmonise when needed (Pellegrino and Hilton 2012). This means that activities and their outcomes are rarely simple or singular. Most designs for learning need to deal with two or more learning goals. If there is only a single learning goal, then optimisation of a design around that goal can be quite algorithmic/formulaic. Design acquires some of its complexity (wicked problems are the norm) because designs for learning are often trying to balance multiple desired outcomes. For example, a learning task might have the goals of: (1) learning to apply a specific Physics principle, (2) developing teamwork and communication skills, and (3) enhancing the learners’ ability to regulate their own learning. Optimising for (1) may make the achievement of (3) very difficult, and vice versa. (Some of the heated, but rather ill-conceived, debate about direct instruction vs. “minimally guided” learning can be brought into focus here. Direct instruction may turn out best for objective (1) but not necessarily for objective (3); self-managed learning may usually be better for objective (3) than (1).) As different sub-fields of the learning sciences have focussed on different aspects of learning, designers need to be able to integrate multiple, ostensibly competing, accounts of learning in order to map outcomes to activities to design.

A downside of our use of terms like “teaching-as-design” (Goodyear and Retalis 2010) or “teaching as a design science” (e.g. Laurillard 2012) is that it can erase some important distinctions between design and interactive teaching. For example, if the same person is designing and teaching, they can fill holes or repair flaws in the design at “learntime”. This ability to make repairs “on the fly” can be of great practical (educational) value, but it also makes it harder to see where the design work stops and the interactive teaching work begins. It can make it harder to see exactly what has been designed, and how the realised form of what has been designed actually functions in practice. This, in turn, makes it harder to re-use designs in other contexts.

A clear view of “what has been designed” is easier to obtain if one places definitional limits on what can be designed. As mentioned earlier, we prefer to avoid implying that behaviour, activity, experience or learning can be designed. Rather, design activity results in the creation of things which can influence behaviour, activity, experience and learning.

The most important kind of designed things (with respect to learning) are tasks: suggestions to the learner of good things to do. Learners’ activity can be strongly influenced, though rarely determined, by the tasks they are set. Design can also result in the creation of, or, more commonly, the bringing into place of already existing, tools, artefacts and other material/digital objects that constitute the “learnplace” – also known as the “learning environment” or the “setting” for learners’ activity (Carvalho and Goodyear 2013). When we say that “learning is physically situated,” we mean that what can and cannot be done, what is easier and harder to do, is often influenced by the stuff that comes readily to hand in the learnplace. As learning is also socially situated, design can also have an effect when it creates things that influence whether and how learners work together – for example that suggest they should work in pairs, or in jigsaw groups, or consider themselves active, legitimate, albeit peripheral members of a community of practice.

Some designed things are influential because they work as/through texts. Tasks are often communicated through written specifications, for example “please write a 5000 word essay on the history of instructional design.” The material form in which the task specification is inscribed is not (of course) the same as the task itself. Usually, it is the characteristics of the task, rather than the medium in which it is inscribed, that most influence the learning activity and its outcomes. But the material form is not unimportant. If a teacher communicates a homework task to students verbally, those who were not present that day may fail to do the work; those who were there may misremember what was asked for; those who were given the task on a piece of paper may lose it. A task inscribed in a course website has a more robust existence.

Interpretation of texts is not the only way that students’ activity is influenced by designed stuff – far from it. The tools, artefacts and other resources that come to hand are influential. The qualities of the learnplace can be influential.

The fields of educational technology and human–computer interaction make much use of the concept of “affordance” in explaining how the material world can be said to have an effect on human activity (Ingold 2000; Laurillard 1987; Norman 1999). More recently, both Collins (2010) and Oliver (2011) have questioned the coherence and explanatory power of “affordance” – preferring to stress the interpretative powers of human agents. As Goodyear and Carvalho (2013) argue – theorising the relations between the material world and human activity needs both concepts (interpretation and affordance).

Some activity proceeds smoothly without much conscious thought. Some kinds of learning task work best if they demand hard thinking – if they invoke “desirable levels of difficulty” (e.g. McDaniel and Butler 2011). Design for learning is likely to have best results if it helps learners deploy their mental efforts where they are most needed. For example, if successful completion of a task depends upon both hard thinking and the use of an unfamiliar computer-based tool, it would be best if use of the tool did not also require hard thinking. The same applies to the design of learning environments, whether virtual, material or hybrid. The affordances of a learning environment should help students focus mental effort where it is most needed.

Two related theoretical perspectives help with this kind of design challenge. Kahneman (2011) distinguishes between two systems that are involved in human cognition. “System 1” is intuitive, fast and emotion-laden. “System 2” involves deliberation, is slower and more rational. In a similar vein, Sweller (2008) draws on David Geary's evolutionary psychology to distinguish between “biologically primary” and “biologically secondary” forms of knowledge. “Biologically primary” knowledge is knowledge that we have evolved, as a species, to acquire (e.g. our mother tongue). “Biologically secondary” knowledge is culturally important and taught in educational institutions through formal instruction. Sweller argues that proponents of discovery learning are making the mistake of assuming that biologically secondary knowledge can be learned through the “natural” methods our species has evolved for acquiring biologically primary knowledge.

These distinctions – fast and slow thought; biologically primary and biologically secondary knowledge – can help understand how to approach the design of scaffolds for learning (including scripts, which we take to be a special kind of scaffolding). More generally, they help distinguish between areas of design where the goal is to invoke “System 2” and areas of design where success is measured by the extent to which “System 1” supports smooth navigation of a learning environment. Mapping this onto Sweller's terms, design cannot always rely on the automatic invocation of biologically primary learning processes. “Natural” learning methods will not always be adequate to solve the challenges at hand.

Returning to our account (in the previous subsection) of affordance and interpretation, we might now say that affordances are what connect designed materials to System 1 and interpretation is what connects designed materials to System 2.

An additional key aspect or dimension to be considered is the locus of control – with the teacher/designer, with the student/learner, with devices/materials/virtual agents in the learning environment. This is not a simple matter, because learning activity can be thought of as nested and the locus of control can vary between levels/layers at any one time (Jones, Asensio, and Goodyear 2000; Laurillard 1988). For example, one can imagine a course within which there is a tight specification of the learning goals and of the ways in which students will be assessed, but where the students themselves are given a high level of freedom over how they learn. The degree of structuring, or the locus of control, can be seen as applying to each of the design components in design for learning. For example, a teacher might specify the learning goals, but give the students freedom to choose the tools they want to work with; or students may be told who they have to work with (e.g. in teams of four, or in a jigsaw group) but allowed to pick their own learning project. The existence of such complicated freedoms means that design for learning usually needs to be considered as an indirect practice (Goodyear 2000). Only where the locus of control is firmly with the teacher/designer does this indirection become of marginal importance. In most other circumstances, there is necessarily an uncertain relationship between what gets designed and set in place for the learner and what they actually end up doing – and how, and with whom. For example, learners will often construct their own interpretations of the requirements of a designed task and work accordingly. Or a certain kind of virtual space for learning may be designed, involving a number of different tools, and then one finds that the actual spaces of interaction are “contingent and locally inhabited” (Jones and Dirckick-Holmfeld 2009). Similarly, designs proposing certain kinds of divisions of labour will sometimes be taken by students as suggestions rather than commands – the students’ subsequent activity, with its various acts of communication, collaboration and coordination may result in very different social arrangements. (It is worth noting that many areas of design, outside education, have learned to live with such uncertainties.)

There are various ways of thinking about, or discriminating between, the key agents – designers, teachers, students and technological agents. It is important to pay careful attention to the goals and capabilities of these agents. For example, a designer may or may not be involved as a teacher at learntime. Whether they will be able to provide extra guidance at learntime can have a significant effect on how and what they design. As another example, avoiding cognitive overload of the teacher at learntime is one of the starting points for talking about design for orchestration; the ability of the teacher to monitor and intervene is part of the rationale for orchestration (Prieto et al. 2011). Yet again, students often need to experience an appropriate combination of self-management (autonomy) and just-in-time support. If a teacher or technological agent is able to help them balance autonomy and guidance at learntime, then designs may not need to pay such careful attention to reconciling these forces.