Some rough notes on how one could create a better model of technology than merely using it as the name for some kind of concave cost-of-production curve.
So you want to create a model of technology. That’s lucky, me too. Let’s jam ideas.
Now, presumably, in detail, technology is completely unpredictable. New technologies are definitionally often at the limits of our comprehension and require the work of many people and many computers to get functional. They often depend upon laws of physics we only just worked out, or possibly haven’t even worked out yet. (consider, e.g., the discovery of X-rays)
On the other hand, many things, in some loose sense, can be abstracted from the progress of technology over time, at least as seen in the current world market. Firstly, technology may or may not “grow” - consider steady-state societies- but if it does, it tends to increase in energy intensity, in length of supply chains, in specialisation of the labour of the people participating in it. Standards tend to form (rail gauges, screw threads, the metric system). Path dependence, descriptive complexity, brands and other tags to differentiate similar products, a profusion of different products, use of new materials. Success of a given technology that is not endogenous but a property of its interaction with a great many other technologies.
We suspect this process is the engine of progress. Perhaps it is. But on the other hand, perhaps its just more effective harnessing of available energy.
Can we specify a stochastic processes that has these characteristics? Can we measure its dependence on energy, on available money, on embodied information?
We could allocate resources by barter and pairwise collision, as in Sugarscape. It seems that this. might be a bit tricky when finding novelty is the goal. On the other hand, perhaps the cost of novelty could include this fiddling.
Another possibility is a market place. A fully connected market. then we’d need to bid, I suppose. Ties agents to claimed rates of return, and likelihood of them fulfilling the contract.
Shall we endow agents with primitive seed technologies? They can make loans to each other.
Where do unused materials go? What effect to they have on replication?
How is energy consumed/preserved?
Can we have a single resource, and that is energy in various forms?
How much are strings worth to agents? How can a consistent fitness landscape be carved out of the incredibly high dimensional space? Can its variance be made to increase with string length (Increasing both rewards and costs with length).
Who gives what to whom, and how do they find what the environment itself supplies? How can we get something like a good connectivity without having to go to the trouble of implementing a Walrasian auctioneer?
If we suppose that technologies happen by some kind of modeling of return on investment, then the prediction quality is itself something that technology enables. Can this be encapsulated?