Today: Several riffs on options and “multivocality”.

Tomorrow: A little bit on “totalitarianism” and my recent readings.


I. Options

I’ve painted a worldview that’s very messy, but not so messy that it demands inaction and trembling. I do think that engineering drastic moves is dangerous when you can’t see very far. (I also believe that we can’t see very far.) Still, there are things worth doing and strategies worth pursuing. Some local actions are apparently better or worse. Some strategies involve discovering and holding many options. Other strategies are explicitly about culling bad options.

Playing: Exploring, opening options. Creative.

Gaming: Culling options. Actualizing a small set of options. Destructive.

Humans are not particularly excellent information processors, and so if presented with many options we automatically initiate behaviors to cull many of them and then apply some more deliberative thinking to deal with the limited set we have left. Again, the world is messy: any given situation has too many options to deal with pre-processing. Naturally, many of the options we cull will be bad options, and many will actually be good options that we have arbitrarily failed to consider for lack of good pattern recognition. On the whole, these culling strategies (heuristics, culture, folk knowledge, institutions) were at least wily enough to have survived. Some strategy sets have survived by sheer dumb luck or by their incidental relationship to actually-fit strategies (call them “Cargo Cult” strategies). Some successful strategies, to improve their reproductive fitness even further, have attached themselves to dumb (but memorable) stories. Nuance doesn’t preserve well. Some strategies were once fit but are no longer adaptive due to changes in the environment. We should not always presume to know which strategies are Cargo Cult strategies. Collisions between behaviors and reality are the most important test, much more important than abstract appeals to our fancy.

Option culling is usually more important to humans than finding new options- humans are very fragile, and many options are real’ bad. But, discounting restrictions of time and energy and information, a perfect and rational agent likes options and ought to want to have as many as possible.


II. Entropica

Have you heard of Entropica? It made a bit of a splash last year. It’s based on the ideas expounded on here, and one reasonable (although I think poorly presented) critique is available here (nearly arguing that “causal entropic force” theory is a very presumptive totalizing scheme).

The paper suggests an odd theory of intelligence. Condensed: “How many future histories can you seize control of?” [“Intelligent behavior emerges as an effort to keep future options open (src)”]  “When the best computer programs play Go, they rely on a principle in which the best move is the one which preserves the greatest fraction of possible wins.”

It’s an interesting idea, anyway, even if wrong.


III. Narrative Optionality: Multivocality and Robust Action

Quotes from this paper, excerpted by this excellent post:

Yet the puzzle about Cosimo’s control is this: totally contrary to Machiavelli’s portrait in The Prince of effective leaders as decisive and goal oriented, eyewitness accounts describe Cosimo de’ Medici as an indecipherable sphinx …
… Lest one conclude that this implies only savvy back-room dealing, extant accounts of private meetings with Cosimo emphasize the same odd passivity.’ After passionate pleas by supplicants for action of some sort, Cosimo typically would terminate a meeting graciously but icily, with little more commitment than “Yes my son, I shall look into that” (pp. 1262-1263)

Regarding Francisco Franco’s inscrutability, along the same lines, from the same post:

“It is said that if you meet a gallego on a staircase, it is impossible to deduce if he is going up or down. Franco perhaps embodied that characteristic more than most gallegos.”

It makes survival sense for Franco, as the coalition he assembled was wildly contradictory. His enemies (and his allies) would over-commit to their own actions and crash-and-burn around him.

Cosimo “never said a clear word in his life” (p. 1308). But not only was Cosimo inscrutable; his actions, especially after 1434,
… appeared extraordinarily reactive in character. Everything was done in response to a flow of requests that, somehow or other, “just so happened” to serve Cosimo’s extremely multiple interests. (p. 1263)
We use the term “robust action” to refer to Cosimo’s style of control. The key to understanding Cosimo’s sphinxlike character … is multivocality-the fact that single actions can be interpreted coherently from multiple perspectives simultaneously, the fact that single actions can be moves in many games at once, and the fact that public and private motivations cannot be parsed. Multivocal action leads to Rorschach blot identities, with all alters constructing their own distinctive attribution of the identity of ego. The “only” point of this, from the perspective of ego, is flexible opportunism-maintaining discretionary options across unforeseeable futures in the face of hostile attempts by others to narrow those options.
Crucial for maintaining discretion is not to pursue any specific goals [my emphasis]. For in nasty strategic games, like Florence or like chess, positional play is the maneuvering of opponents into the forced clarification of their (but not your) tactical lines of action.  Locked-in commitment to lines of action, and thence to goals, is the product not of individual choice but at least as much of others’ successful “ecological control” over you … Victory, in Florence, in chess, or in go means locking in others, but not yourself, to goal-oriented sequences of strategic play that become predictable thereby. (pp. 1263-1264)

Of course, robust action will not work for just anyone. For the flow of requests to be channeled, only some network structures will do. And for the resolution of judge and boss to be credible, coherent interests must remain opaque as far down as it is conceivable to peer. Contra Machiavelli, even Cosimo himself did not set out with a grand design to take over the state: this assumption reads history backward. … Cosimo’s political party first emerged around him. Only later, during the Milan war, did Cosimo suddenly apprehend the political capacity of the social network machine that lay at his fingertips.


IV. Quiet Decision Patterns

I only once really engaged with Venkat’s four decision patterns, but I’ve found them to be really useful tools for thinking about behaviors.

I think that we tend to tell stories with only two decision patterns: deliberative and reactive. They are very mechanical, and both involve a single, clear model of directed action. Stories are linear and simple, this-then-that-then-that. Each action, in this Newtonian model, causes a simple result in the next step, which might be the cause of a new event.

The other two decision patterns are more complicated to deal with narratively:

Opportunism involves holding more than one model at a time in your head, and switching tracks. There may not be a discrete moment of realization. Retroactively, opportunism can be written as a grand plan (deliberative) or as an incidental/immediate event (reactive), but this writes out the serendipity and recognition that the decision pattern actually entails. Multivocality is an Opportunist strategy.

Proceduralism is also difficult to talk about, because unlike action-reaction there is often a difference of scale involved that makes it easy to omit: (a ritual inherited from my culture un-examined; a procedure instilled by an organization automatically; a habit humming along beneath my awareness). There isn’t usually much attention paid to the procedural layer in our storytelling, although it contains the most information/behavior. As Venkat notes in the above link, breaking into the opponent’s OODA loop usually involves hacking their procedural-level decision lair, where their attention is usually missing.

“Robust Action”, it seems to me, makes the most sense framed by the Opportunist and Procedural decision patterns. To someone with a simple direct-action-oriented view of only deliberative/reactive decision patterns, the Opportunist/Procedure-hacker will seem illegible, perhaps like a divinely lucky or otherwise off-script person.


V. Robust Action: Sociopaths

In the Gervais Principle series on Ribbonfarm (another must-read), Venkat defines the characteristics and dynamics between three archetypes in the Organization: the rationally-disengaged Loser, the earnest middle-management Clueless, and the amoral [read: outside-the-system] Sociopath, who drive patterns that birth, scale, and eventually kill Organizations.

From GP-1:

The Sociopath (capitalized) layer comprises the Darwinian/Protestant Ethic will-to-power types who drive an organization to function despite itself. The Clueless layer is what Whyte called the “Organization Man,” but the archetype inhabiting the middle has evolved a good deal since Whyte wrote his book (in the fifties).  The Losers  are not social losers (as in the opposite of “cool”), but people who have struck bad bargains economically – giving up capitalist striving for steady paychecks.

The Sociopath character (not a diagnosis, just a title) is the most interesting one to me, because they set terms and build social systems. Importantly, their status allows them to extract resources from a system and then exit successfully when it stagnates and dies. They achieve this through robust opportunism and careful procedure-legislating.

The Clueless class has no understanding of the source of the arcane power of the Sociopath (“they must have worked extra hard and earned their success!”), and it is this innocence and earnest over-performance that leads them up the ladder to middle-management. The Loser may have had a slight enough glimpse of the horrid, unmediated, messy, godless universe of the Sociopath to realize that the cards are stacked against them- which is why they mentally “check-out” and never over-perform within the Organization.

Towards the end of the series, the behavior of the Sociopath is described.

From GP-5:

For [Sociopaths], HIWTYL [“Heads I Win Tails You Lose”] is not about hacking reward/penalty structures after the fact. It is about proactively engineering systems and processes that reliably, predictably and stealthily generate HIWTYL outcomes.  In other words, they look for ways to systematically claim paternity for successes, and orphan failures.

Good sociopaths foster plausible deniability in their speech. They reap benefits personally and distribute losses throughout their Organizations, sometimes through the brilliantly-conceived “Hanlon Dodge”:

The basic mechanism by which Sociopaths transfer blame to the Clueless, while reducing the overall severity of the penalty, is an application of Hanlon’s Razornever attribute to malice what can be adequately explained by stupidity.

Because Hanlon’s Razor is often true, it is a believable dodge even when it is not.  Coupled with another uniquely human trait, the tendency to link penalties to intentions rather than consequences (eg. first-degree murder vs. vehicular manslaughter), Hanlon’s razor can be used to manufacture predictable HIWTYL outcomes out of fundamentally unpredictable situations.

How? By shifting blame from a locus where it would be attributed to malice, to one where it can plausibly be attributed to incompetence, the severity of penalties incurred is lowered.

Hanlon’s razor is double-edged, and Sociopaths use it to feign incompetence themselves or to charge others with incompetence, as necessary.