%% date:: [[2022-12-26]], [[2023-03-26]], [[2023-04-05]] Author:: [[Steven Johnson]] %% # [[sources/Book/Emergence|Emergence]] ## Summary > [!abstract] Summary > Emergence is the development of new and complex behaviors or characteristics within a system that were not present in its individual components nor specifically intended by a higher controlling force. It is the evolution of the whole beyond its parts in unexpected ways. Natural selection is an example of emergence. > > There are a few conditions that facilitate emergence: a system of *organized complexity* (rather than a simple system or a disorganized but complex one), diversity, bidirectional feedback, networked communication, random encounters, and reproduction or ability to adjust to feedback. > > Emergence involves a type of "swarm logic" that is bottom-up rather than top-down, and thus is fundamentally subversive. ## Thesis > [!question] What are the main points of the book? > What was the author trying to say? Identify the overarching ideas and how the author connected them. ### Criteria for emergence ![[emergence-how-to-be-a-good-ant.png]] #### A system of organized complexity There are three different types of problems in the world. ![[emergence-types-of-systems.png]] - *Simple* problems are those that can be solved by simple mathematical calculations. These problems are one- or two-variable problems and it doesn't take much to solve them. - Example: Given a planet's orbital distance and duration, where will it be in position to the sun in 300 days? - *Disorganized complexity* problems are those that involve too many unpredictable variables to be able to math everything out. While they can't be solved with 100% certainty, we can still use statistics to come up with models to arrive at an answer with a high degree of certainty. - Example: In a game of billiards, how many balls will end up still on the table after the first break? - *Organized complexity* problems are ones that, at first glance, might *look* like sheer chaos, but actually behave according to an overaching logic, whether intended or not. These problems often involve other, interrelated problems, that must all be solved to unravel it all. Organized complexity problems are not feasibly solvable in the traditional way (computing all the variables), and so we have to be a bit more creative. - Example: Evolution is an organized complexity problem. #### Diversity [[Sample sizes for load testing|Sample sizes]] must be sufficiently large to provide an accurate assessment of the body as a whole. Large sample sizes facilitate [[Diversity]] in the way that strategies are implemented. ![[emergence-strategies.png]] Even when you have sufficiently large numbers, diversity in *approach* must be encouraged. Ignorance is useful; individuals must be allowed to specialize or have biases and preferences that affect how they do things. #### Bidirectional feedback loops Emergence requires a way for individuals parts within the whole to receive feedback about their performance so that they can learn from it. Feedback can come in the form of: - a numerical disparity between the intended and actual results - a number or rating to be improved upon by future generations - a statement of success or failure ![[emergence-natural-selection.png]] It's interesting to note that the response to feedback does not have to be perfectly logical. TRo the contrary, [[Natural Selection]] works best when the responses are *imperfect* and allow for some variation. ##### Negative feedback may be more useful than positive feedback Negative feedback may be more effective in facilitating behavioural change than positive feedbackl This is partly because [[Natural Selection]] also works best when there's an element of risk or real danger for a species. ![[emergence-negative-feedback.png]] #### Networked communication Individual components need to be able to communicate with each other to see what has already been tried, and allow that knowledge to reflect their strategy. They must have access to their neighbour's information and make a decision about a future action accordingly, without having an order passed onto them from above. #### Random encounters The previous requirement, networked communication, must be random in nature: no overarching manager must be deciding which neighbours an individual component comes across. Emergence requires a mechanism that facilitates [[Serendipity]]. #### Reproduction Components must be able to act on the information from networked communication by adopting some parts of how neighbours acted into their own behavioural patterns, such that there is a distinction between one generation and the next. ![[emergence-reproduction.png]] ### [[Bottom-up approach]]: "Swarm logic" ![[emergence-demon-hierarchy.png]] The author speaks of [[Oliver Selfridge's]] [[Demon Hierarchy]], in which a hierarchy of demons trying to guess a letter organise themselves such that they have "lower-level demons shrieking to higher-level demons shrieking to higher ones". Emergence involves learning that begins from individual components and spreading to the ecosystem as a whole, not the other way around. Swarm logic includes the ability to assess feedback, find patterns, and *change behaviour accordingly*. Swarm logic allows each "ant" to develop its own [[Weltanschauung]]. ### Emergent software Software can be *evolved* instead of written. The book talks about an experiment involving "virtual ants" that were trained based on an pheromonal obstacle course, then allowed to "breed". Researchers then counted how many generations it took for *all* virtual ants to complete the obstacle course perfectly. ### Emergent systems aren't always good The book talks about some examples where emergent systems aren't a good thing, including the proliferation of fake or sensational news and tornadoes-- both of which are feedback-heavy and emergent, but not necessarily a positive effect. Some emergent systems are destructive. ### Similar ideas > [!question] Similar ideas > Have you heard these ideas before? List similar concepts from other authors, applications of the same ideas, or arguments that support the author's ideas. #### [[Learning in public]] Just like the networked communication that virtual ants lean on to sniff out the right way, learning in public fosters [[Scenius]] by contributing knowledge to a community that can iterate on that information. ![[emergence-scenius.png]] #### [[Zettelkasten]] The Zettelkasten methodology uses a [[Bottom-up approach]] to [[Personal Knowledge Management|PKM]]: instead of starting with a topic and fleshing it out, the Zettelkasten approach involves cultivating several avenues of interest, and then waiting to see the structure or clusters that emerge. The related concept of [[Principle of Atomicity|atomic]] notes is the PKM counterpart to the individual "ants" that emergence takes root in. #### [[Obsidian]] backlinks Backlinks in Obsidian are [[sources/Book/Emergence#Bidirectional feedback loops|bidirectional feedback loops]] that allow notes to "interact" with each other and be influenced by each other. #### [[Napkin]] The PKM app [Napkin.one](https://napkin.one) uses [[Artificial Intelligence|AI]] to automatically generate tags and categories based on content. This could be seen as a type of "emergent intelligence" that is reliant on the amount of content and input (in the form of pre-vetted or pre-entered tags) that it has received. #### [[Artificial Intelligence]] Artificial intelligence is a type of emergence in software. #### Play encourages emergent behaviour Good [[Play]] leaves enough space for individual decisions and creativity, and such conditions often foster emergence. ## Antithesis ### Emergence is just chaos that's been given enough time The author talks about emergence being distinct from [[Chaos Theory]] in that emergent behaviour involves the development of *new* behaviours that did not occur at random. However, firstly, chaos is not random. It merely appears it at first glance. Very few things are truly random. In this sense, chaos can be thought of as fertile ground for emergence. Secondly, one could say that the author draws a distinction between chaos and emergence based on their output. While it's true that emergence requires a coherent new characteristic to develop, it's also true that that development likely took several iterations to emerge. What we call "chaos" can evolve the same way. ### Emergence takes too much time to be useful While the idea of emergence is enticing, the reality is that we can't always put it into practice because of the sheer amount of time it involves. Cultivating an environment ripe for emergence requires a lot of initial investment, significant maintenance, and then a lot of patience in waiting for new behaviours to develop. ### The results of emergence are unpredictable The nature of emergence makes it impossible to forecast beforehand what the end result would be, precisely because there is no top-down proliferation of instructions. This unpredictability, added to the fact that emergence takes a lot of time, calls into question its usefulness and practicality as more than just a fun experiment. ### Emergence treats individual behaviour as a black box Emergence treats individual behaviour as a [[Black Box]]: ants are given a task, something happens, and *poof!* new behaviour emerges. It almost treats the process of evolution or adaptation with reverential awe. While it's true that emergence is generally unpredictable, that doesn't make individual behaviours unpredictable or unexplainable. ## Synthesis > [!question] Middle ground > How would you reconcile conflicting ideas? How is this relevant to you? ### The opposite of emergence is reductionist hierachy Emergence is the type of radical evolution that occurs from the bottom-up. A reductionist hierarchy is the opposite in two respects: - Hierarchies are top-down, and instructions are passed down accordingly. - [[Reductionism]] is the belief that the whole can be explained by an analysis of its parts. It opposes emergence because emergence is the phenomenon of entirely new behavior arising from parts that do not exhibit it. ### In what situations is emergence better than a reductionist hierarchy? - When a system is exceedingly complex. There is a tipping point beyond which it's too difficult to try to maintain or manage all the intricacies of a system. - When a system displays some element of intelligence or organization. A completely arbitrary system cannot develop truly emergent behaviors. - When there is sufficient time for experimentation, reproduction, and learning. Emergence takes time. - When individual liberty is more valuable than productivity. - When the goal is exploration rather than to achieve a predetermined and specific goal. ## Related - [[readwise/Books/Emergence|My highlights on Emergence]] - [[emergence-scribe-notes.pdf|My notes on Emergence on the Kindle Scribe]] - [[Emergence]] - [[system/cards/Emergent load testing - Rules for organized chaos]]