Scan barcode
erikars's review against another edition
5.0
This is one of those rare books that says little, but what it does say is well founded. Scale summarizes the work Geoffrey West and various colleagues have done over the year investigating how things scale. The first part of the book focuses on how organisms scale. The second part focuses on how cities scale (with a brief detour into companies).
When I say that this book says very little, what I mean is this: West takes a systems perspective to try to define models which explain some of the dominant factors of scaling. These models are, by design, meant to explain the coarse grained behavior of complex adaptive systems and do not provide predictions for specific instances. He also makes clear that he sees the models he and his colleagues have developed to be the starting point of modeling the systems under investigation. Although complex adaptive systems like organisms and cities will never be perfectly predictable, there is still a large gap between where we are and where we likely can be, especially for the scaling of cities.
One of the interesting observations about organisms that has been known for a long time is that many of their properties scale predictably, and they do so sublinearly. A whale lives longer than an elephant lives longer than a dog lives longer than a mouse. However, how much longer is not just a linear extrapolation from the size. Nor, for example, does an animal that is twice the size of another need twice the calories. It needs less than that. These and many other properties show regular scaling properties which are based on quarter powers -- x^(1/4) or x^(3/4) depending on the property.
This alone is a fun party fact but not necessarily useful. What West and his colleagues did is develop of model starting from a couple of key properties that it seemed reasonable to assign to a metabolic system: energy must be delivered to every terminal unit (e.g., cell), terminal units have approximately the same requirements, and the energy used on delivering resources should be minimized. From this, they derived a model of the sort of network structure which would best deliver nutrients and, by looking at the fractal dimension of these networks in relation to volume, provided a model which predicted the quarter power scaling laws seen in practice.
Even such a rough model can be useful. For example, it predicts that the maximum human life span is ~125 years, which seems to match what we see in practice. By understanding at least the most dominant factors for why that is the limit, we can see that we would need to fundamentally change our metabolic properties to significantly increase our life spans. Or, to put it another way, optimizing the system you have might help you live 120 years instead of 80, but it will not get you to 1000. For that, you need to change the system.
The second half of the book discusses cities and how they scale. One interesting aspect of cities is that they do show regular scaling properties. Given the historical contingencies that went into the development of each city and given that cities are young compared to evolutionary timescales (thousands of years compared to hundreds of millions), it is surprising that there is any regularity. Yet while the data is noisier than for various biological properties, there are trends which can be modeled usefully. Since less of this work has been done for cities, the explanatory models are less mature. What has been discovered so far is intriguing.
Physical infrastructure, such as roads, pipes, and power lines, tends to scale sublinearly like organisms. This makes sense since they are problems with a similar shape: within each physical system, the relevant terminal units have fairly constant size (even skyscrapers and tiny cottages have similar sized outlets and faucets), must all be served, and are trying to be optimized relative to physical properties size as distance or, in the case of water, very similar efficient flow models to circulatory systems.
More interesting is that social factors, both good and bad, tend to scale superlinearly. Although, to use a set of examples West mentions, New York, Los Angeles, and Dallas all feel very different, they are also roughly scaled versions of each other. Analyzing the data shows that many socioeconomic such as total wages, the number of professionals, patent production, the amount of crime, number of restaurants, and prevalence of diseases all scale with a factor of approximately 1.15. This means that a larger city doesn't just have more crime, wages, restaurants etc., it has more per capita than a smaller city, and the amount more varies in a predictable way. Note an important caveat here: for any particular property, the scaling relationship holds within a country, not across countries, but the scaling factor itself is fairly comparable. Thus, we cannot use the wages of New York City to estimate wages in Mumbai, but if we know how the wages in Tokyo, we can predict the wages in Osaka.
The model here is not as well developed as for organisms, but it these properties seem to derive from the properties of social networks. In particular, if you assume that humans have a limited capacity for social relationships (à la Dunbar), that people will generally have that capacity saturated, and make various assumptions about how information flow works in a social network (à la Granovetter), then you end up with a model that predicts superlinear scaling of various socioeconomic factors.
One upshot of this is that it is not really accurate to treat properties of cities as if they ought to scale linearly. It is not surprisingly, for example, that LA has more crime per capita than Seattle. It follows predictably from the dominant scaling characteristics. The flip side of this is that we should not expect cities to maintain the same properties as they grow. Some sort of essential character is not the primary determinant of various socioeconomic properties of Seattle. Another upshot of this is that if we want to figure out how to change things, it probably will not help much to look at another city which does that thing better unless it is an outlier after applying the dominant scaling factor. A third upshot is that such consistent scaling laws implies that major changes are unlikely to come from fiddling at the margins and are more likely to come with fundamental paradigm shifts.
Some of West's colleagues have worked on applying similar analyses to companies. The findings there are intriguing but sparse because detailed data on social structure within companies and on details about how companies fare on various factors are not as easy to come by as for organisms and cities. However, what the investigations have shown is that companies tend to be more like organisms in that they seem to scale sublinearly rather than superlinearly like cities. This may be because companies tend to be structured fairly hierarchically, although this is just speculation based on the findings.
The book ends on a note of caution. One obvious question which arises when facing superlinear growth is whether or not it is sustainable. As the failures of Malthus' predictions show, innovations can allow us to buck what would seem to be the natural and catastrophic end to a exponential growth curve. That is one source of hope. On the other hand, the growth of cities seems to be faster than exponential in a way that requires faster and faster innovation cycles to reset the timeline to where resources cannot keep up with growth. Innovation keeps pushing the margin, but every time we push the margin, we use it up even faster than before. We have shown that we can innovate, but it is a separate question to say whether or not we can innovate fast enough.
For a concrete example of faster than exponential growth, consider global population growth. Exponential growth is what it would look like if there was approximately the same amount of time between each population doubling. What we see in practice is that the amount of time it takes the population to double is shrinking. Each time the population doubles, we have less time to figure out how to deal with it. Now, there are indications that this trend may be starting to reverse and that the time between doublings will increase again. Thus, we can hope that even if faster than exponential growth is not sustainable, it does not necessarily need to end in collapse.
However, from a socioeconomic perspective, we have come to expect faster than exponential growth. Thus, it is likely that even without a catastrophic collapse, slowing to "merely" exponential growth will feel like a painful shrinking of resources. Can our increased creative output keep up? It is an interesting challenge.
When I say that this book says very little, what I mean is this: West takes a systems perspective to try to define models which explain some of the dominant factors of scaling. These models are, by design, meant to explain the coarse grained behavior of complex adaptive systems and do not provide predictions for specific instances. He also makes clear that he sees the models he and his colleagues have developed to be the starting point of modeling the systems under investigation. Although complex adaptive systems like organisms and cities will never be perfectly predictable, there is still a large gap between where we are and where we likely can be, especially for the scaling of cities.
One of the interesting observations about organisms that has been known for a long time is that many of their properties scale predictably, and they do so sublinearly. A whale lives longer than an elephant lives longer than a dog lives longer than a mouse. However, how much longer is not just a linear extrapolation from the size. Nor, for example, does an animal that is twice the size of another need twice the calories. It needs less than that. These and many other properties show regular scaling properties which are based on quarter powers -- x^(1/4) or x^(3/4) depending on the property.
This alone is a fun party fact but not necessarily useful. What West and his colleagues did is develop of model starting from a couple of key properties that it seemed reasonable to assign to a metabolic system: energy must be delivered to every terminal unit (e.g., cell), terminal units have approximately the same requirements, and the energy used on delivering resources should be minimized. From this, they derived a model of the sort of network structure which would best deliver nutrients and, by looking at the fractal dimension of these networks in relation to volume, provided a model which predicted the quarter power scaling laws seen in practice.
Even such a rough model can be useful. For example, it predicts that the maximum human life span is ~125 years, which seems to match what we see in practice. By understanding at least the most dominant factors for why that is the limit, we can see that we would need to fundamentally change our metabolic properties to significantly increase our life spans. Or, to put it another way, optimizing the system you have might help you live 120 years instead of 80, but it will not get you to 1000. For that, you need to change the system.
The second half of the book discusses cities and how they scale. One interesting aspect of cities is that they do show regular scaling properties. Given the historical contingencies that went into the development of each city and given that cities are young compared to evolutionary timescales (thousands of years compared to hundreds of millions), it is surprising that there is any regularity. Yet while the data is noisier than for various biological properties, there are trends which can be modeled usefully. Since less of this work has been done for cities, the explanatory models are less mature. What has been discovered so far is intriguing.
Physical infrastructure, such as roads, pipes, and power lines, tends to scale sublinearly like organisms. This makes sense since they are problems with a similar shape: within each physical system, the relevant terminal units have fairly constant size (even skyscrapers and tiny cottages have similar sized outlets and faucets), must all be served, and are trying to be optimized relative to physical properties size as distance or, in the case of water, very similar efficient flow models to circulatory systems.
More interesting is that social factors, both good and bad, tend to scale superlinearly. Although, to use a set of examples West mentions, New York, Los Angeles, and Dallas all feel very different, they are also roughly scaled versions of each other. Analyzing the data shows that many socioeconomic such as total wages, the number of professionals, patent production, the amount of crime, number of restaurants, and prevalence of diseases all scale with a factor of approximately 1.15. This means that a larger city doesn't just have more crime, wages, restaurants etc., it has more per capita than a smaller city, and the amount more varies in a predictable way. Note an important caveat here: for any particular property, the scaling relationship holds within a country, not across countries, but the scaling factor itself is fairly comparable. Thus, we cannot use the wages of New York City to estimate wages in Mumbai, but if we know how the wages in Tokyo, we can predict the wages in Osaka.
The model here is not as well developed as for organisms, but it these properties seem to derive from the properties of social networks. In particular, if you assume that humans have a limited capacity for social relationships (à la Dunbar), that people will generally have that capacity saturated, and make various assumptions about how information flow works in a social network (à la Granovetter), then you end up with a model that predicts superlinear scaling of various socioeconomic factors.
One upshot of this is that it is not really accurate to treat properties of cities as if they ought to scale linearly. It is not surprisingly, for example, that LA has more crime per capita than Seattle. It follows predictably from the dominant scaling characteristics. The flip side of this is that we should not expect cities to maintain the same properties as they grow. Some sort of essential character is not the primary determinant of various socioeconomic properties of Seattle. Another upshot of this is that if we want to figure out how to change things, it probably will not help much to look at another city which does that thing better unless it is an outlier after applying the dominant scaling factor. A third upshot is that such consistent scaling laws implies that major changes are unlikely to come from fiddling at the margins and are more likely to come with fundamental paradigm shifts.
Some of West's colleagues have worked on applying similar analyses to companies. The findings there are intriguing but sparse because detailed data on social structure within companies and on details about how companies fare on various factors are not as easy to come by as for organisms and cities. However, what the investigations have shown is that companies tend to be more like organisms in that they seem to scale sublinearly rather than superlinearly like cities. This may be because companies tend to be structured fairly hierarchically, although this is just speculation based on the findings.
The book ends on a note of caution. One obvious question which arises when facing superlinear growth is whether or not it is sustainable. As the failures of Malthus' predictions show, innovations can allow us to buck what would seem to be the natural and catastrophic end to a exponential growth curve. That is one source of hope. On the other hand, the growth of cities seems to be faster than exponential in a way that requires faster and faster innovation cycles to reset the timeline to where resources cannot keep up with growth. Innovation keeps pushing the margin, but every time we push the margin, we use it up even faster than before. We have shown that we can innovate, but it is a separate question to say whether or not we can innovate fast enough.
For a concrete example of faster than exponential growth, consider global population growth. Exponential growth is what it would look like if there was approximately the same amount of time between each population doubling. What we see in practice is that the amount of time it takes the population to double is shrinking. Each time the population doubles, we have less time to figure out how to deal with it. Now, there are indications that this trend may be starting to reverse and that the time between doublings will increase again. Thus, we can hope that even if faster than exponential growth is not sustainable, it does not necessarily need to end in collapse.
However, from a socioeconomic perspective, we have come to expect faster than exponential growth. Thus, it is likely that even without a catastrophic collapse, slowing to "merely" exponential growth will feel like a painful shrinking of resources. Can our increased creative output keep up? It is an interesting challenge.
szachary's review against another edition
5.0
Outstanding. Six stars.
Brings a new filter to apply to the world. Essential reading for those macro thinkers.
Brings a new filter to apply to the world. Essential reading for those macro thinkers.
sophie_pesek's review against another edition
4.0
At first, I thought this book was more or less this xkcd. But by the end, I was pretty engrossed. Still not sure I really understand complexity theory tbh but I'll keep listening to SFI's podcast and enjoying this messy, elegant field
nikakula98's review
challenging
informative
inspiring
reflective
medium-paced
4.0
Very informative and thought-provoking ideas are brought up in this book to bring to light the connections between things that are seemingly unrelated.
rick2's review against another edition
4.0
Fantastic and thought-provoking book that maybe could’ve used an editor to streamline and punch up some parts.
The theory itself is amazing. And it’s one of those things, having read this book a couple years ago, I noticed it seeping into so many aspects of my life. I keep looking at things as diverse as crime statistics and football scores and wondering what kind of the universal underpinnings are. Fortunately, for my league mates in fantasy football, I am an amateur with all of this. Whereas West actually teased out some signal from all the noise. More so on the crime statistics than the football scores, although my DMs are open, my team needs some help.
What results is a really interesting inverse of kind of a common accepted wisdom around organism scaling. To maybe butcher my freshman biology, but as in organism scales with size, it essentially is constrained by sort of a 3/4 power law. Basically saying that within the same sort of class, (mammals, invertebrates, etc.) the larger an organism gets the harder it is to spread around nutrients. Blood vessels get smaller and smaller the farther away they are from the heart in a mathematical way. This impacts things like metabolism, heart size, and so on, and so forth.
The really fascinating bit that I keep coming back to mentally is that the inverse is true for systems. I’ve been fortunate to work at a handful startups as they’ve grown from 10 people to 20 to 50 to 100. and it’s always interesting to reflect back and look at how the
problems you solve at 10 people and at 100 people differ so wildly. The last couple I’ve worked in an operations role so it’s everything from on boarding people to IT support. And with 10 people you can do it all by hand, FaceTime me personally and I’ll solve your computer issues by lunch. That doesn’t work when you have 100 or 200 people. You have to set up entirely different systems to support them, from wikis, to recorded videos, to paying for actual IT software. I have a nightmare where I would be forced to do this for an organization of like 10,000 people and from scratch have to figure out how to navigate that tar pit.
So the really cool part of this book is that there’s some actual mathematical underpinning for all of that. As an organization scales, the complexity scales exponentially. As a city scales, the resources to deliver essential services like crime prevention and fire fighters scale in an exponential way as well. I’m sure we all can instinctively tell that traffic does not get more congested on a linear path. As an aside, I would say it’s more of a jump diffusion model, but I think it also fits in here with Mr. West musings. It’s fascinating as a theory, and I do think it should be more talked about.
As a minor criticism, I do think the book was clearly written by an academic. Despite being one of the more fascinating books I read in 2019, it’s kind of like eating porridge without add-in’s, nutritional but oh so bland. And I wouldn’t necessarily comment on it because the core ideas do carry the day, except for the fact that, while reading, I found myself quickly rendered comatose on more than one occasion. Having to reread whole chapter because my mind had decided to run off elsewhere. In my version of the world, where everything runs perfectly and I always win my fantasy football matchups, this guy meets up with Randall Munroe to produce the next edition of this.
The theory itself is amazing. And it’s one of those things, having read this book a couple years ago, I noticed it seeping into so many aspects of my life. I keep looking at things as diverse as crime statistics and football scores and wondering what kind of the universal underpinnings are. Fortunately, for my league mates in fantasy football, I am an amateur with all of this. Whereas West actually teased out some signal from all the noise. More so on the crime statistics than the football scores, although my DMs are open, my team needs some help.
What results is a really interesting inverse of kind of a common accepted wisdom around organism scaling. To maybe butcher my freshman biology, but as in organism scales with size, it essentially is constrained by sort of a 3/4 power law. Basically saying that within the same sort of class, (mammals, invertebrates, etc.) the larger an organism gets the harder it is to spread around nutrients. Blood vessels get smaller and smaller the farther away they are from the heart in a mathematical way. This impacts things like metabolism, heart size, and so on, and so forth.
The really fascinating bit that I keep coming back to mentally is that the inverse is true for systems. I’ve been fortunate to work at a handful startups as they’ve grown from 10 people to 20 to 50 to 100. and it’s always interesting to reflect back and look at how the
problems you solve at 10 people and at 100 people differ so wildly. The last couple I’ve worked in an operations role so it’s everything from on boarding people to IT support. And with 10 people you can do it all by hand, FaceTime me personally and I’ll solve your computer issues by lunch. That doesn’t work when you have 100 or 200 people. You have to set up entirely different systems to support them, from wikis, to recorded videos, to paying for actual IT software. I have a nightmare where I would be forced to do this for an organization of like 10,000 people and from scratch have to figure out how to navigate that tar pit.
So the really cool part of this book is that there’s some actual mathematical underpinning for all of that. As an organization scales, the complexity scales exponentially. As a city scales, the resources to deliver essential services like crime prevention and fire fighters scale in an exponential way as well. I’m sure we all can instinctively tell that traffic does not get more congested on a linear path. As an aside, I would say it’s more of a jump diffusion model, but I think it also fits in here with Mr. West musings. It’s fascinating as a theory, and I do think it should be more talked about.
As a minor criticism, I do think the book was clearly written by an academic. Despite being one of the more fascinating books I read in 2019, it’s kind of like eating porridge without add-in’s, nutritional but oh so bland. And I wouldn’t necessarily comment on it because the core ideas do carry the day, except for the fact that, while reading, I found myself quickly rendered comatose on more than one occasion. Having to reread whole chapter because my mind had decided to run off elsewhere. In my version of the world, where everything runs perfectly and I always win my fantasy football matchups, this guy meets up with Randall Munroe to produce the next edition of this.
zb1113's review against another edition
4.0
Reads pretty academically. Lots of interesting data on scaling laws (power laws, exponential/super exponential/linear) and the underlying structures of our biology, our cities, our economies and our companies. Interestingly, these laws scale to both the physical structures and the socieo-economic/relationship based structures, albeit differing in important (potentially catastrophic) ways. The final chapter as a call to cross-disciplinary investigation and societal changes toward sustainability was a nice sum up.
mkesten's review against another edition
3.0
A very ambitious study. Could have done with a little more editing.
Are there universal laws of scaling chemistry, biology? Yes.
Do the same universal laws apply to the structures of urbanization and commerce? The jury’s still out on this one.
How important is it to figure out how to reign in the exponential growth our species?
I’d say pretty darned important.
Will technical innovation do it for us? Not bloody likely.
Will we negotiate the political and economic compromises needed with other countries to make this happen? Also, not so likely.
Are there universal laws of scaling chemistry, biology? Yes.
Do the same universal laws apply to the structures of urbanization and commerce? The jury’s still out on this one.
How important is it to figure out how to reign in the exponential growth our species?
I’d say pretty darned important.
Will technical innovation do it for us? Not bloody likely.
Will we negotiate the political and economic compromises needed with other countries to make this happen? Also, not so likely.
kahawa's review against another edition
5.0
This is one of those books that stays with you. I'll probably go back and read this again. I always enjoy learning about complexity, and West breaks things down in a very understandable way with no scary equations. I was particularly fascinated with the idea that mammals are scaled versions of each other, each with the same blood pressure and average heart beats per life time. I enjoyed the look at cities too. I'm not sure they're as mysterious as some make them out to be; they are, after all, simply expressions of brains - brains build them, to suit the needs of brains, so it shouldn't surprise us that they emulate neural networks. Loved his discussion about science and interdisciplinaryism. And group size dynamics. This book put into words things that I've felt intuitively since I was very young. It does take away some of the spark of reality though. Things are much simpler and mechanistic than we realise in daily life.