There can be real science behind Education. Ziegfried Engelmann’s followers (or: enthusiasts of his work) and collaborators would even say that education is a solved problem[1].
What follows is based mainly on the first chapter of Engelman’s “Theory of Instruction”, first published in 1982 – a dry theoretical tome I’m actually still reading.
The Theory of Instruction (ToI) has a few simple theoretical foundations.
The environment is assumed to be the primary cause of what is learned – but the learner is also considered to be a variable. In this context, how do we investigate the relationship between the environment and the learner? Since controlling the learner is actually “impossible”, we must perform the experimental control of the environment. One way to do this is by logically designing a “faultless” communication process so that only one interpretation is (logically) possible/permissible. A faultless communication process could be defined as one that is logically capable of transmitting the underlying concept or skill to any learner (provided they have certain minimum attributes that are going to be defined below). Note that the focus is on the logical analysis of the stimuli, instead of the behavioral analysis of the learner.
The performance of the learner in the learning task or how he responds to the faultless communication reveals important information about the learner. Specifically, information on how he deviates from the “perfect” response (derived from a behavioral analysis of the learner) shows the extent to which he does or does not possess the mechanisms necessary to respond to the faultless communication of the concept, as well as (in the later case) the way one must design instruction intended to modify his capacity to respond to that communication. The strategy is then used for designing instructional sequences and for deriving principles for communicating with the learner.
ToI also makes some assumptions about the learner. In order to provide for control of the maximum number of communication variables we must postulate a simple learning mechanism, as well as assume that the learner’s behavior is lawful (i.e. if he possesses the proposed mechanism then he will learn exactly what the communication is intended to teach). The proposed learning mechanism has two attributes:
A) The capacity to learn any quality from examples.
B) The capacity to generalize to new examples on the basis of sameness of quality.
Here “quality” is any irreducible feature of an example. The first attribute indicates what the mechanism is capable of learning so that the only limiting factor is the acuity of the underlying sensory processes. The second attribute suggests how learning occurs – in this case, the mechanism somehow makes up a rule that indicates which qualities are common to the set of examples used to teach a concept. This way, the generalizations achieved are completely explained in terms of the examples (and their possible common qualities).
ToI postulates that the primary difference between very different cognitive skills is the quality that is to be learned – and that comes from concrete examples, instead of the learner himself. The process of learning is considered to be exactly the same, regardless of what is to be learned. The postulate implies the type of structure we must provide to cause specific generalizations. The primary analysis of cognitive learning must be an analysis of qualities of examples and the communications that present such qualities to the learner (stimulus-locus analysis).
A communication aimed at inducing a specific generalization must meet the following structural conditions or requirements (derived directly from the assumptions about the learning mechanism)[2]:
A) The set of positive examples demonstrates only one identifiable quality.
B) The communication must also provide a signal that accompanies each example that has the quality to be generalized (e.g. saying “red” while presenting differently-shaped objects colored in red).
C) It must present a range of examples that show the physical variation of the examples that exhibit a common quality.
D) It must also present negative examples to show the limits of the variation in quality that is permissible for a given concept (boundaries for the range of permissible variations).
E) It must provide a test to assure the learner has received the information provided by the communication (positive and negative examples that had not been demonstrated earlier, but that are implied by the range of variation of the qualities to be generalized).
The techniques one may use to efficiently design a faultless communication (or to correct an inappropriate one) also follow from the above structural requirements: select the alternative that is most efficient and show uniqueness more emphatically. At the same time, the design of test examples can be reduced to some how-to-do-it formula once you know what test examples must do.
The observation of deviations from the predicted response (given a verifiably faultless communication) means that the learner does not have – or is not using – the assumed learning mechanism. It also means that we know (based on empirical findings on learning) how we must modify the learner’s capacity to produce the desired responses (response-locus analysis)[3].
Summary
"If the student has not learned, the teacher has not taught."
ToI assumes students are lawful things: given a certain environment with certain stimuli, they will react to them in entirely lawful/predictable ways. E.g., given an explanation, the student will arrive at an interpretation that is logically compatible with that evidence.
Fundamental problem: often multiple interpretations are compatible with the presented evidence, while only one interpretation is intended.
Solution: give carefully constructed evidence that logically rules out all but one interpretation (faultless communication). (The most important mistake of failed instruction is that it is logically ambiguous – the student can choose wrong interpretations, those are rarely found out early through tests and further understanding becomes impossible.)
So now I only need to apply the theoretical framework to the problem of self-education and figure out a way of deriving (and effectively testing) a number of different experimental course designs. How do I teach myself a complex discipline/skill without mastering the content itself (only knowing the logically correct way of organizing and presenting evidence – and of designing and allocating tests through time)?
——-
[1] I.e., the theoretical work has been mostly done already, so one just needs to master and modify it, in much the same sense that scientists figured out mechanics but people still need to become “engineers” to be able to use it (so to speak). And yes, Theory of Instruction enthusiasts are usually a evangelical in their language, but given the current (and historical) state of affairs in education (i.e. how horrible everyone else is – schools could conceivably be considered one of the worst institutions ever) one eventually “gets the hate” and accept the need to put some heads in spikes. Direct Instruction (DI) is the implementation of the ToI for (mostly K-8) schools, made possible by the same people (i.e. Engelmann and collaborators). ToI focuses primarily on the theoretical underpinning of any effective instruction, while DI deals with practical constraints. DI is considered to be the only theory in the entire history of education that has any serious empirical evidence whatsoever.
[2] ToI seems to be related to the principle of logical induction. As it happens, induction is pretty much what learning is. Also note that a logically unambiguous communication design is simply one that makes causes transparent (in the sense of Judea Pearl’s construction of causality). On the other hand, it’s one thing to abstractly know that a successful induction must obey certain laws, and then to actually remember to apply these laws when teaching concepts outside of logic itself.
[3] The modification involves specific guides about the amount of practice, massing and distribution of trials, schedules of reinforcement, and other variables that influence the response.