7.2 Emergent Phenomena
More actions
Many phenomena in science are emergent, i.e., visible only at higher levels of organization. This tends to occur when large numbers of elements interact, e.g. as in individuals on social media.
The Lesson in Context
After discussing more concrete forms of causation, we now introduce a type of causation whose outcome is not measured on an individual level, nor on a statistical level (e.g. averaging over individuals). In emergent phenomena, the outcome cannot be found by measuring any number of individuals, but is purely a property of the entire system. We tie this concept of causation to society by recognizing that many sociological phenomena may be of this type.
- The idea of emergent phenomena calls back to the "raft vs. pyramid" descriptions of science. From a reductionist viewpoint, all our understanding breaks down when we end up being wrong on a smaller scale. However, scientists may work on problems at different scales of explanation and still have meaningful results. Harkening back to the raft viewpoint, the areas where these explanations overlap should be consistent. If the explanation at any scale turns out to be wrong, it does not automatically invalidate the other scales.
- Singular causation is individual intervention leading to an individual result. General causation is the same intervention on many individuals leading to results on these individuals that can be seen only statistically. Neither cover the case where the result is purely a property of the whole system, rather than any number of individuals.
- When something goes wrong in society, we may blame a particular leader or policy. This is can be helpful in determining how to intervene for future prevention. However, one shouldn't overlook the possibility that the cause is emergent, and therefore the blame may be misplaced.
- When deciding what simple objects and rules to include in a model of an emergent phenomenon, it is helpful to separate the first-order causes from the higher-order ones. For example, in modelling the persistent traffic jams on highways as an emergent phenomenon, we do not include honking.
Takeaways
After this lesson, students should
- Understand that simple things governed by simple rules can, in aggregate, result in surprisingly complex behaviors, which can be studied in and of themselves.
- Be aware of when complex behaviors in some physical and sociological systems may be the consequence of relatively simple rules on the constituents, and therefore not fully explicable by either reductionism or deliberate agents.
- Be aware of humans' tendency to over-perceive agency in external phenomena in general (e.g. anthropomorphizing), making us prone to mistaking emergent phenomena as intentional.
- Understand that there is value at larger and intermediate scales of explanation despite the fact that larger scales may be reducible to smaller ones.
Scale of Explanation
Emergent Phenomenon
Phase Transition
Scientific Reductionism
Social media
- When people are highly connected to each other in social networks with simple rules of interaction (likes, friending, retweeting, etc.), unintended emergent phenomena, such as widespread misinformation, are likely to arise.
Water
- Water is wet, but water molecules are not. The property of "wetness" must emerge from the relatively simple rules that govern the interactions between water molecules.
Conway's Game of Life
- A classic demonstration of cellular automata, in which simple local rules result in complex patterns arising on a larger scale.
Shelling Model of Segregation
- A model wherein large scale racial segregation can emerge even when individual people have no explicit desire to segregate. When the agents are fine being in the minority as long as at least some fraction of people that live around them are of the same group, even this mild preference can lead to segregation.
Consciousness
- Consciousness as a result of local interactions between neurons.
YouTube is promoting late night talk show hosts over smaller individual creators. It must mean that the YouTube executives are deliberately stifling small content creators.
An arrogant physicist would say that all your thoughts and emotions are nothing more than complex interactions between subatomic particles.
"Let's say chess is the rules of the universe. After two thousand years, we finally figured out how the pawns move. And then I suppose one day we'll have the God equation and that'll tell us how the whole chess board moves and then we'll become grand masters." -Michio Kaku
Useful Resources
Recommended Outline
During Class
| 15 Minutes | Have the students walk through the introductory discussion on Conway's game of life. |
| 30 Minutes | Guide the students through the Shelling neighborhood segregation example. |
| 35 Minutes | Guide the students through the conspiracy theories discussion and modelling activity. |
Conway Discussion
Instructions
Show your students the following video. Pause the video whenever it stops and some text that says "Paused" appears in the top left corner.
Every time the video pauses, have your students spend one minute in small groups answering the following questions about the current state of the video.
- What are the pieces?
- What are they doing?
"Pieces" is asking what the objects of interest at the current scale are.
Once the video is done, ask your students the following discussion questions.
Conway Discussion Questions
Arrogant Friend
Suppose you're looking at the largest scale when your arrogant friend comes to you and says "I know what's going on here. There's tiny square pixels that turn on and off based on their neighbors. I can run an exact simulation of this and get everything you see on the larger scales. All your descriptions of intermediate pieces and interactions are superfluous."
As a scientist studying this system, do you feel that your friend's explanation fully captures all there is to know about the system? Why or why not?
There's a sense in which your friend is right, but they're being overly reductionist. There's a tendency to think only about the most basic rules and think that all the work is done. While that may in principle give you everything you need to know to simulate other scales, it doesn't tell you anything useful about how phenomena at those scales actually happen. If you're trying to intervene on a scale other than the smallest one, then just knowing the simplest rules may not be helpful.
Solids and Transistors
Let's say you're a physicist that knows all about electrons in solids and how transistors work. Your grandma comes to you and says that she's having trouble getting a game to open on her phone. She assumes that since you understand the technology that makes her phone function, you should be able to fix her issue.
Is this correct? What are the assumptions she's making about scientific understanding?
She's falling for the same trap that your friend was (though in a more innocent way). She's assuming since you understand electrons you understand everything that results from their interactions as well.
What are some intermediate levels of understanding between your understanding of electrons/transistors and her understanding of user interface?
There's lots of levels that can broken up different ways. Here's one possible breakdown.
- Quantum Mechanics
- Electrons and Individual Transistors
- Logic Circuits
- Hardware Architecture
- Machine Languages
- Assembly Languages
- High Level Programming Languages
- Software and User Interfaces
- Your Grandma's Game Not Opening
Shelling Neighborhood Segregation Model
Students will play around with a simulated model of neighborhood segregation. Have them mess around with it in small groups and answer the following discussion questions as they do so. Then call the class back together for a larger discussion.
Shelling Discussion Questions
Question 1
What are the components of this model/simulation?
- What is the overall system, and what does it represent?
Some grid of pixels representing a city.
- What are the simple objects, and what do they represent?
Pixels that can be either red or blue corresponding to some binary feature of a person, e.g. ethnicity.
- What are the simple rules?
A pixel moves to the nearest unoccupied space if an insufficient fraction of its immediate neighbors are the same color as it. This corresponds to some preference that agents have to be near those that are similar to them.
Question 2
What is the phenomenon that emerges? (What complex pattern emerges as a result of this simple rule?)
Pixels segregate into neighborhood clusters of the same color.
Question 3
Set the model to the following values:
- Similar: 75%
- Red/Blue: 50/50%
- Empty: 10%
- Size: Any Value
And then also try setting it to these values (changing only the "Similar" item):
- Similar: 76%
- Red/Blue: 50/50%
- Empty: 10%
- Size: Any Value
Is the phenomena that we see different? If so what's the difference? How did we get such a drastic change just be tweaking one value by one percent? Do we think this would apply to the real world? Why or why not?
In the first set of values, you eventually reach a steady state of maximized segregation. In the second set, agents keep moving forever and are never satisfied. They would like to have similar neighbors, but it's too hard to find enough similar agents to make this happen. So the city stays diverse (but in motion) forever. We have observed a "phase transition." This is a point where a tiny change in values can cause a huge change in the behavior of the system. The change from the first set of values to the second is analogous to the change from frozen ice to liquid water.
Question 4
Continue playing with the model and see how the emergent behavior changes. What parameters have the biggest impact? Under what conditions does the emergent phenomenon collapse/vanish?
One thing students may observe is that the phase transition is shockingly robust with regards to changes in the red/blue ratio. There's other conditions where the model has interesting behavior as well. For example, if you have a low similar percent threshold and really imbalanced red/blue percents then the the city might not reach a steady state.
Question 5
What features are missing from this model? Do those features matter with regards to the phenomenon that emerges?
There's lots of missing features. For example, there's no terrain to create natural neighborhood boundaries. There's also no schools, work, stores, or other things that would cause agents to leave their neighborhoods. Other oversimplifications are that there's only two types of agents, everyone has the same threshold, the agents opinion of each other doesn't change, etc. The list goes on.
Question 6
How well do you think the agents in this simulation model real agents picking their neighborhoods? Does this model tell us anything about how agents behave and what sorts of interventions it might take to reduce neighborhood segregation?
The model suggests several possible interventions. One option is to enforce rules around whether or not agents live in diverse environments. Given the simplifications of the model, this could deal with neighborhood segregation, but would leave many of the agents unsatisfied. An alternative option would be to intervene on the agents and see if it's possible to change their similarity thresholds. In this model, that would reduce neighborhood segregation while also leaving agents satisfied.
Note that we don't claim that agents's local preferences is the only (or even dominant) cause of neighborhood segregation. You could also enforce segregation by enforcing neighborhoods on a "global" level (such as with red lining).
Conspiracy Theories
From the example of the Schelling Model above, we have learned that a model of causation by emergence involves many actors following simple rules of interaction. When we observe a phenomenon we suspect to be emergent, we can build an emergent model by first hypothesizing simple rules on individuals, and then simulating the outcome of the interactions of many such individuals.
Consider a widespread conspiracy theory. (For example, 36% of Americans polled in 2006 believed that it was somewhat or very likely that federal officials assisted in the 9/11 attacks or knowingly let them happen. source) We would like to model the popularization of this belief among Americans as an emergent phenomenon.
In small groups, try to build such a model. (Think about the way the Schelling model works.)
- What is the overall system that you are trying to model? What features do you want to include in your model?
- What are the simple objects (actors) that make up this system?
- What are the simple rules that each of these objects follows (regarding the forming and sharing of beliefs)?
- What are some parameters you could change in your model that might affect how beliefs spread through the system (i.e. society)? What do these represent in the real world sharing of beliefs?
- How do you think the patterns of forming and sharing beliefs are likely to change as you move from smaller to larger scales? (i.e. consider how belief spread is different in a hunter-gatherer society of small nomadic bands vs. the contemporary globalized, internet-connected world.)
Bring the whole class together to share their models with each other.