A historical theory of technology

Technology is knowledge. Knowledge, as is well known, has always been a difficult concept for standard economics to handle. It is at the core of modern economic growth, but many characteristics make it slippery to handle.

The basic unit of analysis of technology is the “technique”. A technique is a set of instructions, much like a cookbook recipe, on how to produce goods and services. As such, it is better defined than the concept of a stock of “ideas” that some scholars prefer [e.g., Charles Jones (2001)]. The entire set of feasible techniques that each society has at its disposal is bound by the isoquant. Each point on or above the isoquant in principle represents a set of instructions on how to combine various ingredients in some way to produce a good or service that society wants. While technology often depends on artifacts, the artifacts are not the same as the technique and what defines the technique is the content of the instructions. Thus, a piano is an artifact, but what is done with it depends on the technique used by the pianist, the tuner, or the movers.

But who is “society”? The only sensible way of defining knowledge at a social level is as the union of all the sets of individual knowledge. This definition is consistent with our intuitive notion of the concept of an invention or a discovery - at first only one person has it, but once that happens, society as a whole feels it has acquired it. Knowledge can be stored in external storage devices such as books, drawings, and artifacts but such knowledge is meaningless unless it can be transferred to an actual person. Such a definition immediately requires a further elaboration: if one person possesses a certain knowledge, how costly is it for others to acquire it? This question indeed is at the heart of the idea of a “technological society”. Knowledge is shared and distributed, and its transmission through learning is essential for such a society to make effective use of it. Between the two extreme models of a society in which all knowledge acquired by one member is “episodic” and not communicated to any other member, and the other one in which all knowledge is shared instantaneously to all members through some monstrous network, there was a reality of partial and costly sharing and access. But these costs were not historically invariant, and the changes in them are one of the keys to technological change.

Progress in exploiting the existing stock of knowledge will depend first and foremost on the efficiency and cost of access to knowledge. Although knowledge is a public good in the sense that the consumption of one does not reduce that of others, the private costs of acquiring it are not negligible, in terms of time, effort, and often other real resources as well [Reiter (1992, p. 3)]. Access costs include the costs of finding out whether an answer to a question actually exists, if so, where it can be found, then paying the cost of acquiring it, and finally verifying the correctness of the knowledge.

The determinants of these access costs are both institutional and technological: “open knowledge” societies, in which new discoveries are published as soon as they are made and in which new inventions are placed in the public domain through the patenting sys­tem (even if their application may be legally restricted), are societies in which access costs will be lower than in societies in which the knowledge is kept secret or confined to a small and closed group of insiders whether they are priests, philosophers, or man­darins. Economies that enjoyed a high level of commerce and mobility were subject to knowledge through the migration of skilled workmen and the opportunities to imitate and reverse-engineer new techniques.

⁸ This cost function determines how costly it is for an individual to access information from a storage device or from another individual. The average access cost would be the average cost paid by all individuals who wish to acquire the knowledge. More relevant for most useful questions is the marginal access cost, that is, the minimum cost for an individual who does not yet have this information. A moment reflection will make clear why this is so: it is very expensive for the average member of a society to have access to the Schrodinger wave equations, yet it is “accessible” at low cost for advanced students of quantum mechanics. If someone “needs” to know something, he or she will go to an expert for whom this cost is as low as possible to find out. Much of the way knowledge has been used in recent times has relied on such experts. The cost of finding experts and retrieving knowledge thus determines marginal access costs. Equally important, as we shall see, is the technology that provides access to storage devices. the historical past.^[41] The nature of the books printed, such as topic, language, and acces­sibility, played an equally central role in the reduction of access costs. People normally acquired knowledge and skills vertically, but also from one another through imitation. Postdoctoral students in laboratory settings full-well realize the differences between the acquisition of codifiable knowledge and the acquisition of tacit knowledge through imi­tation and a certain je ne sais quoi we call experience.^[42] Improvements in transport and communication technology, that made people more mobile and sped up the movement of mail and newspapers also reduced access costs in the second half of the eighteenth century, a movement that continued through the nineteenth century and has not stopped since.

Techniques constitute what I have called prescriptive knowledge - like any recipe they essentially comprise instructions that allow people to “produce”, that is, to exploit natural phenomena and regularities in order to improve human material welfare.^[43] The fundamental unit of the set of prescriptive knowledge has the form of a list of do-loops (often of great complexity, with many if-then statements), describing the “hows” of what we call production.

There are two preliminary observations we need to point out in this context. One is that it is impossible to specify explicitly the entire content of a set of instructions. Even a simple cooking recipe contains a great deal of assumptions that the person executing the technique is supposed to know: how much a cup is, when water is boiling, and so on. For that reason, the person executing a technique is supposed to have certain knowledge that I shall call competence to distinguish it from the knowledge involved in writing the instructions for the first time (that is, actually making the invention). Competence consists of the knowledge of how to read, interpret, and execute the instructions in the technique and the supplemental tacit knowledge that cannot be fully written down in the technique’s codified instructions. There is a continuum between the implicit under­standings and clever tricks that make a technique work we call tacit knowledge, and the minor improvements and refinements introduced subsequent to invention that in­volve actual adjustments in the explicit instructions. The latter would be more properly thought off as microinventions, but a sharp distinction between them would be arbitrary. All the same, “competence” and “knowledge” are no less different than the differences in skills needed to play the Hammerklavier sonata and those needed to compose it. One of the most interesting variables to observe is the ratio between the knowledge that goes into the first formulation of the technique in question (invention) and the competence needed to actually carry out the technique.

The second observation is the notion that every technique, because it involves the manipulation and harnessing of natural regularities, requires an epistemic base, that is, a knowledge of nature on which it is based. I will call this type of knowledge proposi­tional knowledge, since it contains a set of propositions about the physical world. The distinction between propositional and prescriptive knowledge seems obvious: the planet Neptune and the structure of DNA were not “invented”; they were already there prior to discovery, whether we knew it or not. The same cannot be said about diesel engines or aspartame. Polanyi (1962) notes that the distinction is recognized by patent law, which permits the patenting of inventions (additions to prescriptive knowledge) but not of dis­coveries (additions to propositional knowledge). He points out that the difference boils down to observing that prescriptive knowledge can be “right or wrong” whereas “ac­tion can only be successful or unsuccessful” (p. 175). Purists will object that “right” and “wrong” are judgments based on socially constructed criteria, and that “success­ful” needs to be defined in a context, depending on the objective function that is being maximized.

The two sets of propositional and prescriptive knowledge together form the set of useful knowledge in society. These sets satisfy the conditions set out by Olsson (2000) for his “idea space”. Specifically, the sets are infinite, closed, and bounded. They also are subsets of much larger sets, the sets of knowable knowledge. At each point of time, the actual sets describe what a society knows and consequently what it can do. There also is a more complex set of characteristics that connect the knowledge at time t with that in the next period. Knowledge is mostly cumulative and evolutionary. The “mostly” is added because it is not wholly cumulative (knowledge can be lost, though this has become increasingly rare) and its evolutionary features are more complex than can be dealt with here [Mokyr (2005b)].

The actual relation between propositional and prescriptive knowledge can be sum­marized in the following 10 generalizations:

1. Every technique has a minimum epistemic base, which contains the least knowl­edge that society needs to possess for this technique to be invented. The epistemic base contains at the very least the trivial statement that technique i works.^[44] There are and have been some techniques, invented accidentally or through trial and error, about whose modus operandi next to nothing was known except that they worked. We can call these techniques singleton techniques (since their do­main is a singleton).

2. Some techniques require a minimum epistemic base larger than a singleton for a working technique to emerge. It is hard to imagine the emergence of such techniques as nuclear resonance imaging or computer assisted design software in any society from serendipitous finds or trial-and-error methods, without the designers having a clue of why and how they worked.

3. The actual epistemic base is equal to or larger than the minimum epistemic base. It is never bound from above in the sense that the amount that can be known about the natural phenomena that govern a technique is infinite. In a certain sense, we can view the epistemic base at any given time much like a fixed factor in a pro­duction function. As long as it does not change, it imposes concavity and possibly even an upper bound on innovation and improvement. On the other hand, beyond a certain point, the incremental effect of widening the actual epistemic base on the productivity growth of a given technique will run into diminishing returns and eventually be limited.

4. There is no requirement that the epistemic base be “true” or “correct” in any sense. In any event, the only significance of such a statement would be that it conforms to contemporary beliefs about nature (which may well be refuted by future generations). Thus the humoral theory of disease, now generally rejected, formed the epistemic base of medical techniques for many centuries. At the same time, some epistemic bases can be more effective than others in the sense that techniques based on them perform “better” by some agree-upon criterion. “Ef­fective knowledge” does not mean “true knowledge” - many effective techniques were based on knowledge we no longer accept and yet were deployed for long periods with considerable success.^[45]

5. The wider the actual epistemic base supporting a technique relative to the min­imum one, the more likely an invention is to occur, ceteris paribus. A wider epistemic base means that it is less likely for a researcher to enter a blind alley and to spend resources in trying to create something that cannot work.^[46] Thus, a wider epistemic base reduces the costs of research and development and in­creases the likelihood of success.

6. The wider the epistemic base, the more likely an existing technique is to be im­proved, adapted, and refined through subsequent microinventions. The more that is known about the principles of a technique, the lower will be the costs of de­velopment and improvement. This is above all because as more is known about why something works, the better the inventor can tweak its parameters to opti­mize and debug the technique. Furthermore, because invention so often consists of analogy with or the recombination of existing techniques, lower access cost to the catalog of existing techniques (which is part of propositional knowledge) stimulates and streamlines successful invention.

7. Historically, the epistemic bases in existence during the early stages of an inven­tion are usually quite narrow at first, but in the last two centuries have often been enlarged following the appearance of the invention, and sometimes directly on account of the invention.

8. Bothpropositional and prescriptive knowledge can be “tight” or “untight”. Tight­ness measures the degree of confidence and consensualness of a piece of knowl­edge: how sure are people that the knowledge is “true” or that the technique “works”. The tighter a piece of propositional knowledge, the lower are the costs of verification and the more likely a technique based on it is to be adopted. Of course, tightness is correlated with effectiveness: a laser printer works better than a dot matrix, and there can be little dispute about the characteristics here. If two techniques are based on incompatible epistemic bases, the one that works bet­ter will be chosen and the knowledge on which it is based will be judged to be more effective. But for much of history, effectiveness turned out to be difficult to measure and propositional knowledge was more often selected on the basis of authority and tradition that effectiveness. Even today, for many medical and farming techniques it is often difficult to observe which technique works better without careful statistical analysis or experimentation.

9. It is not essential that the person writing the instructions actually knows himself everything that is in the epistemic base. Even if very few individuals in a society know quantum mechanics, the practical fruits of the insights of this knowledge to technology may still be available just as if everyone had been taught advanced physics. It is a fortiori true that the people carrying out a set of instructions do not have to know how and why these instructions work, and what the support for them is in propositional knowledge. No doctor prescribing nor any patient taking an aspirin will need to study the biochemical properties of prostaglandins, though such knowledge may be essential for those scientists working on a design of an analgesic with, say, fewer side effects. What counts is collective knowledge and the cost of access as discussed above. It is even less necessary for the people actually carrying out the technique to possess the knowledge on which it is based, and normally this is not the case.

10. The existence of a minimum epistemic base is a necessary but insufficient con­dition for a technique to emerge. A society may well accumulate a great deal of propositional knowledge that is never translated into new and improved tech­niques. Knowledge opens doors, but it does not force society to walk through them.

4.