Tuesday, August 27, 2013

Red Swans: Extreme Adversaries, Evolutionary Arms Races, and the Red Queen

The Red Swan of evolutionary arms races, where the
basis for competition is the innovation process itself.
As the Red Queen says: "...it takes all the running you can do,
to keep in the same place."
This is the third in the series "Many Shades of Black Swans", following on the introductory post "Think You Know Black Swans? Think Again." This one is named "Red" after the Red Queen Hypothesis in evolutionary biology, which itself draws from the Red Queen in Lewis Carroll's Through the Looking Glass (sequel to Alice in Wonderland).  But in this post I'll talk about competitive and adversarial innovation in general, including host-parasite systems that are most analogous to cyber security today.

In addition to the usual definition and explanations, I've added a postscript at the end: "Why Red Swans Are Different From Ordinary Competition and Adversarial Rivalry".

Definition

A "Red Swan" is a system of consisting of innovation processes and an estimation process where:
  1. The generating process is processes of emergent innovation process between competitors or adversaries, where the focus of competition or rivalry is on the pace and direction of innovation.
  2. The evidence is the history of evolution and the current and historical features and characteristics, either seen from one of the competitors/adversaries or from an outside observer.
  3. The estimation process is some method of reasoning that estimates the future trajectory of innovation based on the past trajectory, and perhaps some inferences about how that trajectory might change.
"Competitors" are players that benefit from consuming or controlling same resource (a prey community, a market, a set of information assets).  If there is plenty of resource to go around and, especially if it is growing in abundance, then the competitive rivalry will be mild.  However, if the resource is scarce and  especially if it is shrinking, then the competitive rivalry can be cut-throat.  Competitors' interests are exclusively rivalrous or antagonistic.  In business, the existence of other competitors can give you legitimacy in the eyes of customers.  Competitors often have common interests to grow the market and, where possible, to avoid costly cut-throat competition like price wars.  Competitors can also have areas of cooperation and symbiosis.

"Adversaries" are players that only benefit by "taking a bite" out of rival players, even if they don't receive equivalent benefits.  In biology, host-parasite systems are classic adversarial relationships.  In society and business, criminal communities are "parasites" on the "host" community or marketplace.

Red Swans can arise in either competitive or adversarial systems, though the coupling is usually tighter in adversary systems, and the incentives are usually more intensive because failure to "win" can mean death or extinction in rapid order.

Main Features

An evolutionary arms race has these characteristics:
  • A competitive or adversarial rivalry over a long time scale, and they are locked into the rivalry with few or no options to leave
  • The incentives for each side are dominated by the need to continually escalate or innovate to overcome the opponent’s capabilities. 
  •  In doing so, it’s possible that neither side gains a lasting advantage (i.e. the Red Queen effect).
There are two broad classes of evolutionary arms races:
  • Quantitative Arms Race -- Increasing rates of investment in existing capabilities. The strategic goal is to overwhelm the opponent.
  • Qualitative Arms Race -- Increased investment in new/different capabilities. The focus is on innovation, especially disruptive innovation. The strategic goal is to render the opponent’s capabilities irrelevant.
In the quantitative arms race, what is being generated is increases in some magnitude -- numbers of missiles, increase in volume and toxicity of a poison, etc.  In contrast, in qualitative arms races what is being generated are patterns and changes in patterns -- new architectures, new system configurations, new arrangements of performance characteristics.

Examples

The most famous example is the Red Queen Hypothesis evolutionary biology, which asserts that sexual reproduction was a strategic innovation to defend against the fast rate of parasite evolution (who mostly reproduce asexually, but at a very fast pace).  Host-parasite systems, in general, often generate Red Swans, and also some preditor-prey systems.

There are other examples outside of biology, including military arms races and technology innovation races.

Why Red Swans Can Be Extreme

Red Swans are capable of generating extreme outcomes because of coevolution which is anti-symbiotic.  In symbiosis, two species coevolve together to be compatible and complementary to each other.  In Red Swans, evolutionary innovation by one party makes things worse for the other.  This strong negative coupling of fitness landscapes dramatically accelerates pace of innovation, How much depends on the details of incentives, resources, history, and innovation capabilities of the players. By pushing each other into new and emergent areas in the landscape, it can lead to extremes and oddities, including overshoot and “evolutionary suicide”.

Why Red Swans Can Be Surprising

Red Swans, unlike some of the other colors of Swans, are capable of being perpetually surprising, and capable of being surprising in surprising ways ("meta-surprise").  The reason is that the coupled innovation processes result in emergent outcomes that restructure their fitness landscapes (a.k.a. spaces of possibilities), and this restructuring process can compound on itself, "bootstrapping" to more sophisticated innovation processes.  Essentially, these sorts of systems are capable of being deeply and continuously surprising, though not always "successful" in the ways they innovation.

In the face of this sort of "radical" emergence, our evidence is always inadequate to eliminate surprises, no matter how complete or detailed.  Likewise, our estimating system -- ways we mentally model the system -- will frequently be inadequate to encompass all the ways the system can innovate.  (This is the subject of intense research in Complexity Science in general, and in my field -- Computational Social Science.)

How Red Swans Can Be Rationalized In Retrospect

When looking back on a trajectory of innovation in an evolutionary arms race, it might appear that the innovations of each party are "obvious", "logical", or at least "well justified".  This rationalization can be done by outside observers, who might claim to have some wisdom about the innovation trajectory that the players in the system do not have.  Rationalization can also be done by players in the system, where they look back on the trajectory of innovation and see it as "inevitable" or somehow preordained.

A subtler form of rationalization is to believe that an evolutionary arms race was inevitable, discounting other strategic options open to players to "change the game".

How To Cope With Red Swans

If you find yourself locked in an evolutionary arms race, then you should consider the following strategies:
  • Boost your innovation capabilities, especially the "clock speed" of innovation and your ability to "reach broadly and fail fast".  All advantages are temporary so it's vital not to rest on your achievements or any momentary advantage.
  • Look for ways to neutralize your opponent's innovation capabilities, i.e. force them to "unlearn" skills and knowledge in which they have deeply invested.
  • Learn about the forces that are shaping the trajectory of innovation -- you may not be able to predict the exact trajectory itself but you might be able to see how and when it will make significant changes in direction.
  • Look for opportunities to break out of the arms race in such a way that you will have a sustainable competitive advantage, or at least parity.
In terms of analysis methods, I think simulations can be very fruitful, not to generate predictions but instead to help you understand the structure of the arms race and the forces that are shaping the innovation trajectory.

--------

Postscript: Why Red Swans Are Different From Ordinary Competition and Adversarial Rivalry

There are quite a few people, including experts, who claim that any competitive or adversarial setting with intelligent players are capable of Red Swans, and thus it is inherently impossible to estimate risk probabilistically.

This is flat wrong, both conceptually and empirically.

In the vast majority of settings, the nature of competition or adversary rivalries do not have the self-reinforcing structure to drive ever higher rates of innovation.  The players in most settings have better things to do and better options to pursue.  In business, companies can compete on price rather than innovation, or they can shift the ground of competition to geography, convenience, brand image, or a hundred other things.  In nation-state rivalries, a military arms race isn't inevitable since very often there are better, less costly policies -- alliances, cultural competition, trade competition, etc.

There's been a lot of research on the exploration/exploitation tradeoff.  "Exploration" is moving into new, explored territory, including anything that involves radical or fundamental innovation.  "Exploitation" is developing and maturing an existing strategy or capabilities, often by "learning-by-doing".   An evolutionary arms race, at least of the "qualitative" type (see above) is essentially a locked-in pattern of perpetual "exploration" by rivals.  But most of the time, rivals have plenty of reason to not commit 100% to "exploration" if they have limited resources and they care about getting the most benefit for costs (i.e. they are economizers with at least bounded rationality).  Instead, "exploitation" will be a significant focus of strategy.  To the extent that competitors/rivals are "exploiting" different strategies, or at least differentiated so that their utility does not solely come at the expense of the other, then the pace of innovation will be limited.

But don't "intelligent" competitors and rivals have the capability and incentive deceive each other and pursue surprising/unexpected strategies?  Yes, but in most settings their investment in deception and surprise is relatively small compared to their main priorities -- innovation itself ("exploration") and reaping the benefits of previous innovations ("exploitation").  In the next post in this series titled "Disappearing Swans", I'll explore the possibility adversaries that have nearly infinite capability to deceive. In the vast majority of competitive and rivalrous situations, each side has incentive to be "just surprising enough" to wrong-foot their rival.

Consider an example from American football.  Let's say that it's a passing situation -- 3rd down and long yardage, and the offensive team is trailing by 6 points in the game in the fourth quarter.  Focus on a single receiver (an End) and his defender (a Corner Back, assume man-to-man coverage).  When the play starts, what does this receiver do?  They run a pass route that is predefined (so that the quarterback knows where to throw or at least where to expect the receiver to be) and, along the way, the receiver tries to "shake off" the defender.  If they don't have a speed or height advantage, the "shake off" usually comes in the form of some "juke", "stutter step", "fake", or other slight deceptive movement that is just enough to wrong-foot the defender.  You never see receivers do anything truly surprising -- e.g. do cart-wheels down the field, or strip naked, or  get down on their knees and start singing "God Bless America".  Anything that surprising would be too costly relative to the main "mission", which is to run the pass route as planned.  Generalizing to offensive plays in general, this explains why very few plays in a professional football game are "razz-matazz" plays like "flea flicker", "double reverse", "Statue of Liberty", etc.  The vast majority of plays are straight forward (i.e. no elements of surprise or deception) or have just enough surprise or deception to wrong-foot the defense and yield a small local/temporal advantage for the offense.

In summary, ordinary competitive and adversary rivalries do not produce Red Swans most of the time.  Yes, there are episodes of disruption and surprise, but those mostly arise from different generating processes --  a Swan of a different color!

2 comments:

  1. First!

    (And with that out of the way, some serious considerations)

    Completely philosophical: I wonder what the long-term success v failure rate of a successful Red Swan is.

    A party that focuses its all on destroying its adversary will inevitably lose sight of developments that are tangential to that goal. In other words, I wonder what the ratio of innovation for self-preservation versus utter destruction of adversary has to be that will balance the scales one way or the other. ;-)

    ReplyDelete
    Replies
    1. Re: long-term success vs failure rate -- great point! It certainly isn't guaranteed that an innovation arms race leads to successful innovation. It could very well lead to an "evolutionary dead end" because it is prone to creating extremes and oddities that could lead to dramatic failure/collapse if the nature of competition and rivalry changes -- e.g. entrance of new competitor or adversary with very different capabilities.

      As you say, there's a myopia that is induced when most or all resources are devoted to destroying a given adversary. Worst-case setting for this might be blood feuds in criminal gangs.

      Delete