The book by Thompson of the constructs of models is an interesting read. It is clearly targeted for the general audience but does provide some enlightenment to the world of models, especially ones that have played a part in our day to day living. Two of the most recent are those related to COVID and those related to climate change.
Overall the books is well written and the author gets their point across. However, there are many points of views on models. I have worked with models for well over sixty years. Thus I have the distinct disadvantage of experience. Early on I had to model the dynamics and control of spacecraft, some of which is controlled by the laws of physics and some by random or stochastic processes. Namely there are event we model to account for things that happen beyond our control or that we have no idea about other than trying to anticipate them. These models got people to the moon and back.
The next level of models are those in business plans for startups. I spent several decades in that space both as an entrepreneur and as an investor. In over hundreds of business plan models I do not think a one ever was even close to what happened. Success came from the entrepreneur being able to recognize issue and adapt accordingly.
The author interjects models of COVID, especially those from Ferguson and the London cabal. In later March of 2020 they predicted in their model deaths in excess of 2.5 million in the US in less than a year. So far three years out we have less than 1.1 million. That is three years and more than two dozen variants. The flaws in their models are extensive. They did not anticipate the mutation rates we see but anyone with minimal skill in virology would have done so since it was a single stranded RNA virus and was easily mutated. Secondly, even today, we really do not know the details of transmission. Large crowds yes but aerosols and virions and then what? Third they made recommendations from those models which resulted in enormous societal impacts and economic disasters. Their models were specious at best and lacking in any details to have reached such conclusions. They did get a lot of Press however.
The author discusses LTCM, and the Black Scholes disaster in 1998. I recall that quite well, I was in Moscow watching banks collapse as well. In 1972 I was at a conference at the University of Chicago with a team from MIT. We were to work with economists on systems and models, especially in a stochastic domain. Some of the folks later at LTCM were there as we had a discussion of an Ito process, and the formula. My parting remark having had the experience of spacecraft and especially Apollo XIII was that they should beware of the long tails on the probability distribution, that are never Gaussian. Thus LTCM was not a surprise. The model was flawed and they knew it.
The author then moves to climate and the related models. Of course one must be careful here since there are many committed souls who have faith in these models. Yet one thing we know about models is that are reflections of their makers not necessarily of the reality they are attempting to describe. In the lates 60s I had the experience in some of the early models, and worked on getting data on aerosols and heating. I learned that each time I gathered mor data I had more questions. I suspect global climate models evoke similar responses.
Let me now make some specific comments.
p.15 Ockham dies in 1348 thus he lived and was active in the 14th century NOT the 13th. In fact his Nominalism was a major factor in examining each entity as unique and the existence of the abstract was denied.
p. 19 Regarding mouse models we have the issue that they are used to examine, for example, immune system responses on specially genetically engineered mice. That in itself is an interesting tale of models. Mice ae mice, and genetically engineered mice are just that. Humans have unique characteristics and each is unique unto themselves. Thus mouse models explore just what the model is capable of stating.
P 36 The single story statement is interesting. Perhaps that is why we have 4 Gospels?
Pp 50-51 Are models sufficient to tell us something? Yes, like an exercise. One must recall that the Navy had a model for war with Japan, War Plan Orange. This was a model of how to fight Japan. So how well did that work out? Think Pearl Harbor. Likewise the Big Data and AI issues are themselves models, but models devoid of any paradigms. Namely AI is nothing more than lots of input with little output. For example thousands of mammograms and ultrasounds with the training for breast cancer. Place a new set in and it will tell you if it cancer or not. Hopefully.
Pp 52-53. Feedback in reality is a key factor found in nature. The Cybernetics of Wiener is a critical element lacking in many models of physical systems. However it is a critical reality in almost all if not every organic system.
P 83 is part of the Black Scholes model I discussed above. The author finds that a model wanting but perhaps for just some of the right reasons. Nature and the markets have that feedback capacity that often changes the underlying model elements one starts out with. Then again there are those nasty Black Swans.
P 108 The author calls AI and autonomous system. I would say it goes well beyond that. AI is not based on underlying first principles proven buy clear and repeatable physical experiments. In contrast it is oftentimes just an amalgam of data weighted in a manner suitable to the designer and “trained” to produce a “proper” answer.
p 114 et seq Economic models are complex abstractions of reality. Demand, supply, etc are abstractions and measured at some gross level. Each economic crisis we go through evokes some alternative economic reality that is an attempt to justify actions. But alas one wonders where this specific economic paradigm was before the crisis and what should have been done to prevent it.
The latter part of the book is an attempt to justify climate models and discusses what governments should be doing. The reality unfortunately is that climate models are just that, models limited by data and the stochastic dynamics of reality.
Overall the book is a good take on models in general. Some
models are good, some necessary, some useless, and some cause undue harm. COVID
models were example where anyone could be their own epidemiologist, albeit
lacking any knowledge of the virus itself. The virus has taken on a life of its
own and may very well outlast all it models.
Posts
