Jul 24, 2002
- whether the practitioners are employing what may be called the scientific method - whether the discipline has produced results that qualify as what may be called scientific principles or scientific knowledge - whether the discipline has a domain of enquiry that could be studied using the scientific method and could produce results that may be called scientific knowledge
With regard to the first point, I have already stated that I do not believe that most current work in education qualifies as a science. I have no interest nor desire to defend the contrary. Adducing numerous examples of the non-scientific nature of contemporary educational research does not therefore impact on my current discussion.
With regard to the second point, I argued that there are some aspects of education that approach knowledge with certainty. While agreeing with me that such knowledge had been obtained, various writers argued that it had not been obtained in the correct manner, and therefore did not count as scientific knowledge. However science is more than the context of discovery: our knowledge of gavity counts as scientific knowledge even though it was discovered non-scientifically long before Newton.
But I am not especially interested in debating the second point either, for reasons that will become apparent below. I am rather more interested in the third question: whether education is a domain of enquiry that could be studied using a scientific method and whether such an investigation could produce results that could be called scientific knowledge. I argue that it is, and that it can. But I must heavily qualify this argument in order to state clearly the sort of scientific enquiry I have in mind and the sort of scientific knowledge I have in mind.
A science consists of what I will refer to below as a knowledge web. A knowledge web is a set of interconnected statements that form a coherent body of belief and that are related according to a set of rules or principles appropriate to a domain. In the physical world the set of statements consists (mostly) of a taxonomy of entity and a set of causal relations between those entities expressed according to the language and rules of mathematics. Some people (eg., Fodor in Psychosemantics) cannot envision any other constitution for a science, yet it is clear that a variety of structures is possible.
For example, the following forms of inference are all acceptable within a scientific domain: - deducation, as in formal (propositional and predicate) logic and mathematics - induction (including projection, inference by analogy and similarity, probability) - abduction, or 'inference to the best explanation' including confirmation and falsification - definition, or a formal stipulation of entities and rules
A science is typically thought to consist of a set of general (preferably causal) principles from which predictions of all events in a domain or discipline follow. The paradigm case of this is the set of principles we call the laws of nature, and it is held (as an urban myth) that we can deduce all states of nature from such principles. Insofar, however, as there is significant evidence suggesting that nature does not obey laws of nature, it is unreasonable to define a science as requiring a law of nature. Rather, a science consists of a set of statements about the (relevant) properties of entities related by one or more of the forms of inference listed above.
Even more to the point, however, is the suggestion that a science consists of one body of knowledge, a set of statements that is (univocally) true, and a contrasting set of statements that is (univocally) false. This is a misapprehension of science. Outside of an appropriate domain of enquiry, a statement is neither true nor false, but meaningless. My assertion, for example, that 'the Sun caused the grass to grow' is neither true nor false unless you understand the domain of my enquiry. If I am talking about the best way to raise a child, and you say to me 'The Sun caused the grass to grow,' I will look at you with a blank stare (or try to come up with an analogy, any analogy, that will help me understand your point). More on this below.
Far more relevant to the current discussion is whether our research into education will bear the sorts of fruits one normally expects and desires from a science. That is, we must jusge the discipline according to what it could produce, as opposed to its current methodology or content. In a science, we normally expect at least some of the following: - predictive capacity, at least, to a reasonable degree of probability, subject to some standard of verifiability (Ayer) or falsifiability (Popper) - explanatory power, within a domain of relevant alternatives, preferably with some degree of simplicity (Ryle) and possibly some causal efficacy - a taxonomy or vocabulary for describing a reasonably coherent ontology, including typical taxonomical relations (is a type of, consists of, is a member of)
My hope for the future of education as a science is based on the supposition that it is now possible to represent the sum of human knowledge in a language which catches this complexity. The language, a simple grammar, is generally known under the title of 'the semantic web' (Berners-Lee). The idea of the semantic web is that what we know in a discipline can be described. Insofar as it can be described, we can form the necessary set of interrelated statements. A visual representation of the semantic web may be found in McGuinness. It is the possibility of capturing knowledge in the semantic web that offers an answer to those who claim that education is by necessity not a science.
When questioned about the semantic web as a whole, most people probably agree that there will be culturally relative components. But they seem to think that their little corner of it will not admit of any sort of relativity, if my experience talking with educational standards people is any guide. I argue frequently, for example, that SCORM is only the first of many application profiles to emerge and suggest that vendors ought to prepare themselves to adapt to different standards on the fly, and yet almost everyone I speak do cannot imagine any sort of alternative educational object metadata. Their loss.
I've done a lot of work in the philosophical underpinnings of knowledge representation. There is no common understanding for even common words, such as 'Paris', much less more controversial words denoting, say, educational units (if you doubt this assertion, have both you and a friend list the concepts relevantly associated with the word 'Paris' then compare lists). There is a whole other class of words that do not allow for a fixed definition even if you hold the denotation fixed to current usage. Such words - the irreducably vague - are simply not used consistently. A good example is the word 'river'. There are no necessary or sufficient conditions for the meaning of the word 'river'. Such words are best defined by what Wittgenstein would call 'family resemblances' - and family resemblances require a set of alternative metadata.
Once we get beyond the definitions of basic words and move into more advanced concepts (Berners Lee envisions knowledge generation layers above RDF and the semantic web, recall) we run into serious difficulties. The meanings of a whole range of statements are context-dependent (my first published paper was an illustration of the contextual factors impacting the meaning of a standard conditional sentence). When we look at things like the meanings of words (Quine), categories (Lakoff), causal relations (Hanson), counterfactuals (Stalnaker) and explanations (van Fraassen) it becomes clear that a description of the cultural surround is irreducably part of the meaning of the sentence (van Fraassen: How do you explain X? You say P caused X. But the sentence 'P caused X' is only true with referance to a relative alternative: 'P caused X to happen instead of Y' But relevant alternatives are defined within the cultural context.
What's important to keep in mind is that this is not an artifact of some sort of theoretical relativity. It is not the case that, if only we had all of our facts and theories straight, that we could settle into a single vocabulary (that is an error committed in the Veltman paper). Rather, it is an artifact of knowledge itself: an artifact of the fact that knowledge itself is relative to state of the knower. Knowledge, as my colleague Rod Savoie like to put it, is information applied to a purpose. Change the purpose and you change the nature and content of an item of knowledge (this is what Steve Eskow is trying to get at, but he concludes incorrectly that since knowledge in a domain could be variable, it is not a science).
To know something isn't merely to know 'S is a P'. If one knows that 'S is a P' then one also knows that 'S is not not a P' and 'If x is an S, x is a P', and a myriad of other statements that follow. But more: it is also to know that 'P is not Q' and that 'S is not Q'. The knowledge that 'S is a P' is embedded in a myriad of related statements, what Quine would call a 'web of belief', and this web is not, cannot be, anchored. When a person says to you that 'S is a P' you can at best appeal to a set of cultural foundations, what Wittgenstein calls 'riverbed propositions', to try to get at what that person actually means. Certainty is not attainable; unanimity is impossible. At best, what we get is a sort of 'meaning by agreement' (Davidson). And where the impetus for agreement ends - where once common goals are no longer shared, for example, meaning by agreement dissolves.
None of this is to suggest that knowledge is not possible, nor is it to suggest that the representation of human knowledge is not possible, nor is it to suggest that education is not possible. After all, this is the system in which we already work: indeed, it is a system that children learn at a very early age (when Daddy says "no" it means one thing, when Mommy says "no" it means something slightly different, and when Grandma says "no" it means something totally different). No, rather, it suggests that human knowledge is much more nuanced that people suppose, and that education is therefore similarly nuanced. It means that the achievement of an education is not the mastery of the facts and theories within a given domain, but rather, a facility to work with a variety of knowledge webs that roughly overlap.
This is something we try to teach students in university when we expose them to competing theories (and therefore, different vocabularies) but even then we suggest that one of them of the other is correct (if only we had enough experimental data). But probably no theory is universally correct, not even in the 'hard sciences' (physicists have long gotten over their angst after discovering that Newton's theories had exceptions, and are now comfortable disucssing singularities when none of their theories apply). All theories (based on the appropriate sort of knowledge web and empirical observation) are true in their own domain (which is why so many of them appear to be unfalsifiable). De Bono's method actually works, in a context: some lateral thinking is better than no reasoned assessment at all, which is the norm where De Bono is applied.
Eskow and some of the others are completely correct to say that education is goal directed, and that there is therefore no single theory that will describe all educational situations. Any educational situation is a mixture of goals: the student's goal, the teacher's goal, the school division's goal, the parents' goal, the President's goal. These define a particular mixture appropriate to an individual student (indeed, I would seek a characterization of learning characteristics defined in terms of what a student wants in an education, as opposed to a student's learning apptitudes). It is not surprising that education resists formalization. But it does not follow that we cannot learn, understand and know how to educate a person: it's just that such an understanding is incredibly nuanced.
What appeals to me about the semantic web - broadly conceived as I see it here - is that it captures these nuances. If we resist the urge to create one true theory of learning (and one true taxonomy, and one true pedagogy) then we are able to stand back and apply a mixture of what we know to each person. The semantic web captures this. When you seek the property of an object, then, because the semantic web is based in RDF, and because RDF supports multiple context-based vocabularies, you may be prompted for a context (or merely presented with a variety of contexts, from which you may select the one most similar to your own). RDF also supports cross-contextual discourse, so that (for example) a Catholic may criticize Lutheran theology, or (for example) a sociologist may incorporate a biological taxonomy.
So how does this get us to a science of education? I adduce two relevant points:
1. We are able to educate ourselves and our young. This is indisputable and is easily verified by observation. There is a phenomenon, called education, that we are trying to study in order to understand how it works. It is an observable phenomenon. Little children and even pets manage at least some of it. We acknowledge that there are superior educators and that in certain conditions they achieve superior results. We see spectacular successes in isolated cases (Mill). And failures in others.
2. We are poised on the verge of being able to describe this phenomenon in a (scientific) language vastly more flexible than ever before. It seems clear that the effort to describe education as a set of (pseudo-)scientific principles has failed, just as it has failed in a variety of other disciplines. This, however, does not preclude us from being able to express what we know as a knowledge web using the language of the semantic web.
Our description of education within the structure of the semantic web will lead us away from the desire to create one or a dozen overarching 'laws of education' and to identify a more precise statement of what constitutes successful (and unsuccessful) education in different contexts. Just as I can say from observation that that works (and what counts as success) differs on a First Nations reserve from a university classroom, I can say that a description of education within the semantic web will draw out those differences. As I once commented in a previous article, while I can see Guy Bensusan's method working in Arizona, I can't see it working for me: but this is far from a refutation of Bensusan, it is simply a statement that what works there doesn't work here, just as an astrophysicist would say that what works on the Moon does not work in a black hole or at the subatomic level.
References (you'll have to find the publishers, dates & page numbers yourself)
Ayer, A.J. Language, Truth and Logic. Berners-Lee, Tim. The Semantic Web. Davidson, Donald. Inquiries into Truth and Interpretation. De Bono, Edward. Lateral Thinking and Parallel Thinking. Hanson, Norwood Russell. Patterns of Discovery. Lakoff, George. Women, Fire and Dangerous Things. Popper, Karl. The Logic of Scientific Discovery. Quine, Willard van Orman. Word and Object. Mill, John Stuart. Autobiography. Ryle, Gilbert. The Concept of Mind. Stalnaker, Richard. Counterfactuals. van Fraassen, Bas C. The Scientific Image. Wittgenstein, Ludwig. On Certainty.
And look up...
McGuinness, Deborah L. Ontology Comes of Age. http://www.ksl.stanford.edu/people/dlm/papers/ontologies-come-of-age-mit-pre ss-(with-citation).htm
Veltman, Kim. Challenges for a Semantic Web. http://www.cultivate-int.org/issue7/semanticweb/
