« Timeless Causality | Main | Einstein's Superpowers »

May 29, 2008


Considering how vague and ambiguous it is, it's kind of hard for much of it to be wrong. How do I test e.g. the claim that "lots of small innovations" are "more important" than "a few big innovations"?

The dynamics of diffusion are also important. The genetic algorithm, while overrated as a search mechanism, is a good model for the creation and diffusion of innovation. A GA which uses global selection (choosing the best n organisms in each generation) typically reaches a lower optimum than one which uses local selection. This is because the best n organisms at any given time are largely related. Selecting the global best n thus severely constricts the gene pool.

For example, broadcast media can find and disseminate good performers and musicians rapidly; but the rate of musical innovation would probably be higher in the long run if we had more diverse radio playlists, and more cultural isolation between regions of the US.

What do you mean by "when innovation is used to signal"?


Are there any books that you would recommend to help me start thinking about innovation in this way? The only one I know of is Mokyr's Gifts of Athena, although I'm not sure if that fits with your thinking here.


growth rates commonly increase then decrease, usually the second factor eventually wins.

How can you tell that growth rates do not decrease then increase?

The idea that innovators should be paid "the full value gained by others" for their innovations seems likely to result in huge overpayments to me. I would rather see sponsorship of inventors based on their expected future productivity.

This list seems easily produced from a point of view that I might call economics (anyone have a better name?). Moreover, I doubt that it is of much use without that point of view (eg, Tim Tyler can't read it). I think it would be more useful as part of an explicit attempt to teach the point of view by example.

This seems like a pretty good list. Of course, not all innovations are "better part or structure designs", some are worse and some are merely different. I think systems may have control algorithms as well as parts and structure, and innovations may be in those algorithms as well.

A system is of course in the eye of the beholder. While an entrepreneur may think of his enterprise as a system, an economist might think of it as a subsystem in a larger system such as an industry. One of the imnteresting aspects of the information revolution is that innovations have made many systems smaller rather than larger. Thus firms have innovated by outsourcing functions and thus simplifying their systems and focusing on core competencies.

Innovations in organizations can also occur in ways people seldom seem to recognize, such as by hiring a new person, or replacing one employee by another.

I too have been thinking in this way for a long time, but can not judge how wide spread that thinking is.

As James Q. Wilson points out in Political Order in Changing Societies, innovation within an organization involves 2 acts; the proposal of innovative idea and the adoption and implementation of the innovative practice. The rate of proposal is directly proportional to the diversity of an organization. However, the rate of adoption of those innovative ideas proposed is inversely proportional to the diversity of the organization.
Diversity in this case is a measure of the complexity of task structure, control mechanism, and incentive system.
His views make complete sense to anyone who has tried to get bring about change to a business, political party, or military staff. The greater the number of entities in an organization, and thus the greater number of interactions and relationships, the greater chance that new combinations of ideas and methods will be thought of. However, the greater the number of entities, interactions & relationships within an organization, the greater the chance that there will be enough opposition to the change (perceived loss of status, resource, or power) to prevent adoption / implementation.

I can read. IMO, the problem is with the phrasing that "innovators are not paid the full value gained by others". Should the tax office regurgitate its sales taxes for the benefit of innovators? What about the wages of manufacturing/sales/marketing men involved in realising and propagating innovations? If this is an ideal, it seems like a silly one. Also, it ignores the whole controversy over intellectual property - and assumes something like the current legal situation.

Silas and ad, for many specific devices, such as the steam engine, researchers have done detailed inventories of the value added from each specific innovation to that device.

Phil, many people like to affiliate with innovative people, and so many people try to appear innovative.

John, yes of course systems are part of other systems, and a control system can be a part of a larger system.

Looks like you're articulating a lot modularity theory, which I know Henry Smith at Yale has been working on in the patent law field within legal academia.

But I would say that most of this stuff isn't accepted as obvious by practicing lawyers, economists, or business people. More specifically, I would say that almost all of the points you make are worthy of a separate blog post, if not a separate academic paper.

The larger problem that I see is that the dominant theories of innovation (and patent law) are supplied by economists, and economists since the 1940s haven't had a good set of tools for analyzing dynamics. Rather, they've been cobbling together models of very rich phenomena by using a bunch of coefficients that don't have apparent physical meaning. Paul Romer's work on spillover effects is an example of a success with this type of technique.

...but as every good physicist knows, sooner or later you have to go back and grapple with the underlying dynamics.

I would go so far as to say that even general observations about how and how much innovations decrease the period of time that it takes goods or services to be supplied would be new for a lot of people in the field.

You don't have to have a whole complex theory of modularity to make improvements in the existing theory of innovation.

Michael, can you be more specific about what claim I made "isn't accepted as obvious" and why?

It probably goes without saying, but read Schumpeter! He has had the most to say about the nature of innovation and entrepreneurship.

1-7 and 9. NOT 8.

Robin Hanson, some of this is standard cybernetics, starting from Norbert Wiener and Ross Ashby. There is a huge literature by now, and I have some questions and comments for you:

"Systems are parts in a structure" In this and what follows, your definitions appear to be contradictory: "systems are parts in a structure" (in #1); "innovations are better part or structure designs" (#1); there are "part innovations" and "system innovations" (#5); there are "system structures" (#6).

Exactly how are you using the word "system" in all of these?

Wouldn't it be better to say instead that systems are certain structures that have separable but interlocking parts, there may be separable subsystems, and that innovations are changes among the parts?

#1 "innovations are better part or structure designs" Why are innovations "better?" The answer is, usually because they REDUCE the space, time, or personal physical effort that is required to effect a transfer, transformation, transaction, or transportation between two parts of the structure. There is a reduction or condensation or concentration quality to it.

At the next hierarchical level up, this condensation may be seen as a specialization or differentiation, to be integrated with others, again via a larger type of transfer, transformation, transaction, transportation, in that higher level.

#2 The growth in systems is due to both (a) "collecting" innovations, and (b) the simple enlargement or increase of the system, including by accretion. A example of (b) is population growth, such as in societies or ecosystems, or the "opening" of new markets. It is hard to maintain that "most" growth is due to innovations, unless you restrict the discussion to certain systems along certain time scales: e.g. modern economic growth, or the eons of evolutionary time.

#3 In the case of institutions, which are homologous to innovations but instantiated at a different (usually higher) hierarchical level, the value often proceeds almost entirely from the single BIG change. Private property, for example, enables the easier transformation of resources and easier transfer of possession among individuals -- which Locke was on to. There was of course a host of "lesser" innovations to attend to private property, such as contracts and so on. But your point was about where most "innovation value" comes from, and institutions are a type system-wide innovation with an enormous importance and value; they are a sort of technology at the social level. (You may dissent from likening private property, or any institution, to an innovation, but then you cannot define innovation as "structure design" in #1 -- at least, not for all structures.)

#4 I wonder if you would please give an example, real or theoretical, from social systems or biology, where innovation rates "decrease when the most valuable easiest innovations tend to be pursued first." I am unfamiliar with this. Are you limiting this to a single technological lineage, as opposed to a whole system of them? Is there an example from biological evolution?

#5 Again, what characterizes "system innovation" in this list of basics? I.e., how and when do you say that the whole thing has changed?

Lee, most long term population growth and opening of new markets is due to innovation. Private property is a mix of lots of small innovations, not just one big one. It is trivial to describe a theory where agents choose the easy innovations first.

and to innovate too little, when innovators are not paid the full value gained by others.

But if innovators gain the full value their innovation creates, then what's the point of encouraging innovation? An innovator creates something that makes $100 million for the world economy as a whole, they get paid $100 million, so the rest of us are no better off than we were before. Furthermore, if any gains I get from an innovation simply go to the innovator, why bother adopting the innovation in the first place? Why buy a more efficient washing machine if the capital cost of the washing machine plus payment to the innovator is equal to the discounted value of the energy saved? Why buy a new medicine if I'm just going to be taxed enough that I'll be as miserable as I was without the medicine? So, under your scenario where the innovator is paid the full value of their innovation, there is in fact too little innovation as no one else has an incentive to make use of the innovation in the first place.

The only reason it makes sense to provide incentives for innovation is that the innovator does *not* capture the full value of their innovation.

Robin, thank you. It didn't occur to me until after posting my comment that you are restricting the use of "systems" to "economic systems," and "growth" to "per capita growth." Because surely, long-term growth in wildlife populations, or even bacteria, is due to reproductive rates and environmental conditions, which are not innovations -- unless you insist upon the fact that these things have evolved to do what they do; i.e., you make "evolutionary emergence" ("innovation," in biology) an obvious precondition of any statement; which would be trivial and redundant.

Even so, there is an exception to your basic rule in economic systems: comparative advantage. Here, each side can concentrate in the pre-existing differentiation where it has the best efficiency. The theory does not require the next step of an additional innovation to increase per capita income. After that, of course, growth is a steady alternation between innovation and trade.

Indeed I meant "opening of new markets" in the older usage of the term, i.e. of exploration or contact with other countries and trade liberalization -- opening a foreign country to trade, having more people to sell to, the prelude to comparative advantage.

On #4: with regard to a theory where agents choose the easy innovations first, can it be other than trivial? It is a tautology that, if innovation rates decrease, the "most valuable easiest" innovations were pursued first. But to write that the rates decrease "when" this is done, made me think this was more than mere definition.

I also did not realize that by "innovation value" (#3) you meant immediate monetary economic value, and not (as any old systems theorist would naturally be thinking!) something like what economists would call "positive externalities" or "network effects." But then how do you define "big" vs. "small" innovations? I think you are incorrect about the specific importance of private property in the web of institutions under its rubric -- but let's take another example. It seems to me that a big innovation was the telephone, and I would define it as big because of the extraordinary positive externalities, leading to an incalculable chain of growth and profits over a centruy. Yet you write that "most innovation value comes from many small innovations" (#3). So how do you draw the line between "big" and "little," in technological innovations?

I have found the work of Charles Leadbetter interesting on this subject (there is a Youtube video but here is his personal website)


The comments to this entry are closed.

Less Wrong (sister site)

May 2009

Sun Mon Tue Wed Thu Fri Sat
          1 2
3 4 5 6 7 8 9
10 11 12 13 14 15 16
17 18 19 20 21 22 23
24 25 26 27 28 29 30