Mathematical economic thinking as one key need for survival
Last night I meditated some on the larger implications of some technical discussions we have had within IEEE, on the future of automotive technology, with direct implications for the larger issues of climate change and global security, which will substantially affect whether the human species survives or not. More concretely: on the Lifeboat list, I have summarized the most serious threats of human extinction as “H2S/NUC/AI/(bio?)”. The threat of death by H2S is not so simple as worrying about our carbon footprint, but reduction in greenhouse has emission is certainly one of the important variables affecting our probability of death by the same H2S mechanisms we have discussed before, which have several times already caused mass extinctions of species on earth. Likewise, conflicts related to oil directly affect the issue of death by series of nuclear wars and nuclear terrorism, more than most people understand when they live their lives in narrow stovepipes and try not to think too hard about scary unpleasant realities of life. Finally... I do not mean to discuss the scenario of death by Terminator AI, but there actually are some connections to what I will discuss here now.
To understand these concepts, starting from square one, really takes a lot of background. Therefore, this blog entry will be long, and I will send much briefer bits of it to a couple of places – referring to the blog for a more complete discussion.
One reason why this is important: Yesterday, at a Quaker discussion group, someone raised the question: “What good is it for retired people like us to speak truth to power anyway? We are not the ones doing what really matters.” I pointed out that folks who develop policy statements in the White House (like OSTP, which I visited a number of times in earlier years) could say the same thing. They speak... but are always in the shadow of the people who do real things. Real things like actually developing new types of space technology or solar technology, or coping directly with large scale realities of issues like education, health and poverty.
Auto industry technology is one of those places where real things are happening, or not happening and possible. It sounds small to many people to talk about AC versus DC in 480 volt recharge stations, yet this impinges directly on all three of the things which might kill us all or not.
At the end of a long listserv discussion, one person argued hard that pure electric vehicles like Tesla are much better than plug-in hybrids like the Chevrolet Volt or the BYD Qin, because they have a better carbon footprint. “Therefore we should make it a policy simply to shift directly to pure EV like Tesla.”
In my view, his argument is just one more example of the fact that humans in all walks of life can be brilliant about the details of what they are doing, and how they achieve their subgoals, even as they get totally mixed up about what their goals do to the larger reality of human life. I saw that over and over again at NSF, when many proposals were brilliant and world leaders in “how” to achieve some goal, but totally confused about the importance of the choice of goals, the “broader impact,” and the connection between their own work and that of others.
Yesterday, I responded to his argument using the same old valid but fuzzy principles which any well-informed energy economist would use. I wasn’t altogether stupid in my reply, but I was operating at my usual late afternoon level of consciousness:
I worry about climate change more than 99% of the population does, but even so I view national security, nuclear proliferation and emerging conflicts in the Middle East as adding up to a problem just as big and just as lethal as climate change.
Another way to think about EV versus PHEV is to think about how to use rational market design to get a kind of optimal mix. In reality, silver bullets and single products have always never been the best arrangement/mix for any major energy market; there are lots of segments. If we don't intend to pass laws requiring everyone to buy EV and not PHEV, the challenge may be more figuring out what kind of incentive or "externality payment" (ala tax breaks) would be appropriate for both. EVs and PHEVs both benefit if there are good and proper incentives for both. Also, the development of the two technologies is synergistic.
But again, a lot of the details (at least for the national security side) are in the transportation addendum already out there. (On the web site of www.ieeeusa.org, an addendum to the National Energy Policy Recommendations, NEPR, for which I was one of the many original authors.)
For climate change... I personally worry a lot about whether anything we can do now is enough. So far as I can tell, our best hope lies in finding ways to get the technology ready so that we can move quickly when and if people smell... not the roses... but the poison. That would include not only new energy technology, but also technology to make various types of geoengineering better and more available. Despite all the colossal lip service, most of what we need to be doing to make such breakthroughs possible is still not being done, due to all kinds of politics at many levels, and I find it ever more difficult to visualize a way forward with hope. I am glad that some of you still have hope and energy, at least for part of this.
But this morning, in meditation, I see that this reply, while valid, was shallow, and did not capture some of the real issues of life which will decide whether we live or die. One of those issues is our ability to think mathematically... something which STEM education should be trying to advance, at least for those of us able to learn to think more mathematically. People say “tell it like it is” – but when reality is ultimately mathematical in nature, that means making a place for the real truth, even when it means being more mathematical. Local priests and shamans, when trying to defend their power have often pushed the idea that ancient Hebrew or Arabic or Latin or Greek or Aramaic or Tagalog are the one true language of power of the spirit and the cosmos – but I would predict that none of those languages, even English and Spanish, would have much traction beyond the earth, but that mathematics (albeit with minor notational differences) is the one language we speak on earth which connects to the entire galaxy. Mathematics and images. So maybe I need to speak real truth, mathematics, at least to IEEE folks in such discussions.
Human species extinction is not the only big issue before us. There are also issues of spiritual growth, quality of life, etc. They also deserve attention, but to survive we need to be able to think about mundane survival, to focus with laser-like intensity on the issue of avoiding human extinction. In this post, I will limit myself to the issue of extinction.
The issue of extinction may reasonably be operationalized by trying to minimize the probability that the human species goes extinct within the next 10,000 years or so. (Why 10,000 years? To avoid distraction FOR NOW on issues like the Darwinian evolution of humans into some other species, and that kind of thing.) This is mathematically just one example of the classic problem of maximization over time in a nonlinear stochastic system. The most complete overview of workable methods to address such problems in given in the IEEE book, Handbook of Reinforcement Learning and Adaptive Dynamic Programming (RLADP), edited by Lewis and Liu, for which I wrote the first chapter giving an overview of the entire field.
Some people view RLADP as just a branch of control theory, but there are many areas of human endeavor, like economics, which also address optimization problems. Norbert Wiener, one of the founders of modern control engineering, explained a lot of the mathematics of what it takes to make a working thermostat which doesn’t go unstable, but he also had discussions with Von Neumann about how to push ahead to solve more difficult problems, leading, for example, to the creation of the neural network field. (Buried somewhere in my house I have a book containing dialogues of Von Neumann, Wiener, Warren McCullough and others which led to the McCullough/Pitts neuron model and the birth of that new field of research.)
Many people in control theory and in fundamentalist religion (two closely related streams of human thought!) got as far as the thermostat in their thinking, but find it uncomfortable to push further the way Wiener and Von Neumann did. “What is a probability anyway? What is the connection between what Von Neumann talked about and real life? Either we survive or we don’t. Where is there a probability?” But more realistic branches of engineering understand that there are stochastic factors even for humble systems like thermostats, and we need to cope with them. One of Von Neumann’s followers, Prof. Howard Raiffa of the Harvard Business School, did a magnificent job of teaching people the real meaning of probabilities and optimization in real life; his simple, readable book which created the field of “decision analysis” and decision trees gives lots of real world examples based on new work he did for real oil companies, where wildcat drilling is an uncertain game of probabilities – a game without which we wouldn’t have an oil industry.
One of the “thermostat people” complained to me a few years ago: “You optimization people are ‘way too optimistic. You say you want to make outcomes better, but in the real world it is hard enough to keep things from falling apart altogether. In the real world, we need a system of order, a stable state which may not be the best, but won’t fall apart. We need solid, ironclad linear degrees of stability – something we can prove, because only when we assume things are linear can we really prove anything with confidence. Stability is all.”
I even remember as workshop on electric power where one of the power engineers (NOT representing all power engineers!) said “you guys can forget optimization. If someone comes to me talking about anything but stability and provably stable linear systems, I won’t even talk to him.” At that same meeting was an executive of a power company who had funded his work, and he said: “I’m glad you told me that, because you need to know that no one will get any more money from US unless they show they can include maximizing value added and consumer benefits in the equation.”
More seriously, when the thermostat guy spoke to me, I replied: “No, it is YOU who are too optimistic. Life is not really certain or linear. There is NO WAY that you can actually give an absolute guarantee of survival or stability under realistic assumptions. The best we can in the real world is to maximize the PROBABILITY of survival, and that’s an optimization problem. Even a mouse in the field cannot guarantee it will survive the next day; the best it can do is to maximize the probability that it survives, and its entire brain evolved to cope with that kind of real-world problem as best it can.” (I tend to view sharia and certain kinds of Vishnu thinking, and JudeoChristian fundamentalisms, all as a search for a stable fixed point solution – to challenges which do not have that kind of solution.)
So – what happens when we look at the problem of maximizing the probability of human survival through the lens of ADP mathematics, the mathematics which is appropriate to that kind of problem? To get real, I need to get a little mathematical.
It turns out that there is a very tight connection between economics and ADP. In fact, there is a Greek letter “lambda” which embodies that connection in a very powerful way. In first year market economics, people sometimes use the letter “p” for price and “q” for quantity, but more advanced work (like Ken Arrow’s seminal books) the vector “lambda” represents a vector of prices. In optimal control theory, there is something called the “Pontryagin equation”, which uses essentially the same lambda vector – but obeying a condition which ensures optimal policies and allocations across time. (Microeconomics usually pays more attention to optimal allocations in equilibrium, as in general equilibrium theory or the newer dynamic stochastic general equilibrium.) In RLADP, I developed a stochastic generalization of the Pontryagin equation, first described in chapter 10 of the Handbook of Intelligent Control (White and Sofge eds, Van Nostrand, 1992). That generalization is the basis of Dual Heuristic Programming (DHP), the algorithm which led to the breakthrough in hit-to-kill missile interception by Balakrishnan, and a number of other successes.
A well-trained energy economist would explain to my IEEE colleague that real markets entail a diversity of prices and suppliers, reflecting the diversity of costs and consumer preferences across a complex world. That’s embedded in what I actually said. But in actuality, to maximize the probability of human survival across a dynamic changing landscape... the real story is that we face a future movement in prices/lambdas if we follow a policy course which really maximizes our probability of survival. Now only are there mixes of prices across the world, as the economist envisions, but there is a trajectory of future prices. The best strategy is not simply to just make believe we are living in utopia now, but ... it’s like planning the orbit of a spacecraft (one of the real applications of DHP!)... and yes, I agree with my old colleague that we do need to be especially careful to develop what we need to be flexible, to be able to survive and prosper under a wide range of unexpected conditions, good and bad. That’s reality... and we need to be able to be clear about it, and not fall back into unrealistic, unviable fundamentalist ways of thinking when we make the move from general principles to actual implementation and action and market design. Market design itself is absolutely fundamental, as is honorable and open competition in general. If we only have honor, or we only have competition, we die.