Wednesday, January 20, 2016

Terrorist Threats: Limits of Rationality

There's a rather famous quote by D. S. Wilson, the prominent evolutionary biologist.

If there is a trade-off between the two forms of realism [factual & practical], such that our beliefs can become more adaptive only by becoming factually less true, then factual realism will be the loser every time. … Factual realists detached from practical reality were not among our ancestors. 

For context, here's how he frames the two forms of realism:

What do I mean by factual and practical realism? A belief is factually realistic when it accurately describes what’s really out there (e.g., there are no people up there sitting on clouds). A belief is practically realistic when it causes the believer to behave adaptively in the real world.

This has some rather important implications in properly analyzing relative threat levels from domestic right wing vs. Islamic terrorism.

First off, we have to ascertain what type of threat is imposed: existential or conventional (eg. money, lives, freedom, etc.).  While I don't know of studies that directly answer this, there are a fair number that circle around the bush.  I'll amalgamate these into a simple function:

threat = damage per event * frequency * potential growth rate * probable damping curve

Upper bound represents in-group orientation metrics.
Lower bound represents out-group orientation metrics.
 
However, this leaves out a key aspect from yesterday's discussion: risk vs. cost metrics.  As mentioned, early intervention toward an out-group has minimal direct negatives. In disproportional scenarios, at worst you lose some unknown future potential benefits (retaliation risks vary directly with out-group strength & penetration potential).  Late intervention toward an out-group has potentially severe direct negatives, including uncertain existential threat levels.  Thus a better threat
function is:

threat = (damage per event x event frequency x potential growth rate x probable damping curve) x power law cost-benefit curve

The power law cost-benefit vs. intervention point time curve shows why good people often talk past each other in terms of right-wing vs. Islamic terrorism.  Path future is more a function of one's own out-group in-group bias than it is a function of rational judgment.  Error rates swamp precision/validity; complexity rules and reductionist rationality is naive fiction.

While future prediction is bounded by complexity recognizing historical patterns definitely facilitates bayesian heuristics. In other words if history shows that immigration (which includes some violent elements) hasn't caused existential problems before, it probably won't again.  This is an out-group position.  The in-group position is sensitive to black swan events;  Sometimes, out-group positions cause complete existential destruction. Tension between these two positions accentuates the role of minor clues. 

threat = (damage per event x frequency x potential growth rate x probable damping curve x power law cost-benefit curve) x black swan frequency x black swan damage

In regard to the role of minor clues, researchers like Haidt would add that the emotional tail wags the rational dog. Kauffman leverages this ex post facto rationality limit by suggesting our emotions are optimally primed to pick up fuzzy signals.  When operating at the edge of chaos, this leads to better-than-rational predictions. This means we shouldn't be so quick to discount "naive" prejudicial out-group heuristics. It simply means good people will disagree on the weighting they apply to short-term history vs. genetic-term history.  Out-groupers may favour a combination of the short-term history & risk-reward.  In-groupers may favour genetic-term history & risk-aversion-cost-aversion.  (Obviously long-term group survivability in competitive landscapes requires both positions and their dynamic tension.)

Turchin's cliodynamic work on secular cycles throws in some interesting data points.  His work suggests ideological conflict is correlated as a pre-cursor to societal collapse.  While this seems obvious, it throws a useful wrench into the right-wing vs. Islamic terror equivalency debate. Which is more ideological foreign & incompatible? 

Turchin's work also highlights the evolutionary fitness of practical rather than factual weightings.  Thus, even if right-wing terrorism were to be perfectly equivalent to Islamic terrorism the main deciding factor is whether, over a long-average, an out-group will treat you better than an in-group? Societal collapse happens when elites split into one group and commoners into another. Each serve their own interests at the expense of the other. Unfortunately for the plebes, the elites have the power (they still excel while societal collapse progresses)!  Eventually even a strong rule of law can't handle the polarization nor its Malthusian trap.

This illustrates some of the unique dynamics occurring in the debate. People within society who appear to favour out-of society actors over in-society actors raise people's freeloader heuristics (specifically, traitorous dynamic heuristics).  This is especially true of people in elite and quasi-elite roles.  Treachery in these groups can have especially nasty consequences for the middle class and non-aligned elites & quasi-elites. This exaggerates the importance of out-group in-group framing.  

For elites, a little out-group entry has: 
  • minimal immediate costs to moderate immediate benefits (more slave-like labour & taxes),
  • potential good pay-offs if out-group swamping occurs (keep status & grow wealth)
On the other hand, for the middle class, out-group entry has:
  • minimal long-term pay offs with out-group swamping (only some will excel during destabilization)
  • moderate negative immediate costs to minimal positive immediate benefits (now the slave or more work)
One doesn't have to be a game theory expert to see the preferred solution for both parties. Add in Turchin's findings that in societal collapse most of the middle gets seriously screwed, and bigoted xenophobes can appear like risk averse bankers.  

Additionally, the middle class should be especially sensitized to actions which appear to devalue historic in-groups.  This is a pending signal that elites, quasi-elites, and your own peers will screw you. Once social contract binding loses effectiveness those on the margins are disproportionately impacted (say via simple rent increases which push you into poverty and then into serfdom due to evaporation of income buffering).

The net result is the addition of a non-linear "non-rational" factor into the threat function.  Wilson might suggest calling this a "practical reality over factual reality weighting".  As potential discrepancy between the two increases, practical reality is weighting more heavily.  This is due to the costs of missing a false negative versus the cost of missing a false positive. So,

threat = (damage per event * event frequency * potential growth rate * probable damping curve * power law cost-benefit curve * black swan frequency * black swan damage) * treasonous elite freeloading probability * disproportional socio-economic impact * nonlinear non-rational sensitivity

Now if you're like me, this function is getting pretty messy.  So messy in fact, that rational choice theory application is non-sensical.  Rational choice theorists (of which I'm not one) suggest people group terms.  Here's one way grouping:

threat =  event frequency * disproportional socio-economic impact * probable damping curve * damage per event * power law cost-benefit curve * black swan damage * black swan frequency* potential growth rate * treasonous elite freeloading probability * nonlinear non-rational sensitivity

One way to interpret this grouping is: 
threat = minimization of out-group threat * rational judgment * maximization of out-group threat * non-linear sensitivity


How conservatives and liberals differ in this weighing seems straight forward.  Confounding simplistic us=good them=bad polemics is the fact that people on different ends of the political spectrum simply switch in-group out-group demographics.  For many on the left, right-wingers are the out-group.  For many on the right, most of Islam is the out-group.  Thus, the real distinction is the non-linear sensitivity term. As Wilson illustrates, history selects against factual realists detached from practical reality. Thus people clue into the potential each group has to cause existential extinction.

Progressive leftists are legitimately concerned about the extinction of progressive society.  A status quo is simply not acceptable.  It represents real disintegration of what the West is to them.  Conservatives are legitimately concerned about the extinction of Western values and culture. A status quo is acceptable. Social progression is valued less than status quo freedoms.

Thus the only seemingly valid metric (which is not paradigm based) is the propensity of right-wing and Islamic terrorists group to infiltrate and attack other states/socieites.  This reveals the worst case level of existential threat (at some far future time).  Islam is clearly different here. San Bernadino type events suggest a non-zero infiltrate and leverage tendency. Insurgencies in many Islamic population centres re-enforce this idea. Fears of localized submission walls within a growing demography are cited.  Right-wing terrorists just have not shown a tendency to go to other countries (unless you count national military as an example of right-wing terrorism, which is certainly a logical position to take - it just legitimizes criticisms about freeloader treachery risks).


No comments:

Post a Comment