Comment: “Towards a Comprehensive Model (…)”

workable foundation for further integration of decisional control modeling with outcome- oriented stress research. Stress can tenably be conceived of as driven by the obtaining of a desirable objective (a ‘positive’ goal) or the avoidance of a known peril (a ‘negative’ goal). Consequently, the understanding of how to shift the probabilities for successful goal pursuit in one’s favour through nested decision-making becomes highly relevant.

Having a functional structure (the mixture-model in the preceding manuscript) in which to place decisional-situation features provides the advantage of placing many types of situations on a comparable footing. Situational features such as number of alternatives (P,

p, q), information availability (C, N : yes; U: no) and executive power (C: yes; U, N: no) at different levels of decision-making are common in organized human social life. Companies, families, charities, schools, armed forces (police, military), and bureaucracies generally all apportion authority somewhat systematically and

hierarchically. This, not to mention the valuable footing provided to researchers who would seek to systematically vary these quantities in a cohesive, unified, formal manner.

The sequential linking of probabilities described in the preceding manuscript starts at the most basic phenomenological level: occurrence or non-occurrence of an event (either of which may be the desired outcome). Over this most common starting point for any type of data, simply counting ‘yes’ or ‘no’, are mounted threat values ti – the chance of a

‘yes’. Governing threat values, in turn, are the chance of obtaining the threat value in ordinal (aka “ranked”) position i, where i = 1 is best and i at a maximum value is worst. This level of ‘probability governance’ is denoted Pr(ti), the chance of getting ti.

Governing Pr(ti) in turn again, is the probability of a given decision structure, such as

CC, NC, CU, or in a three-level hierarchy, CNU or UUN, for example. To index these decision structures, the indexing variable j is recruited, similar to i for threat.1 That is,

1

Note that a ‘choice structure’ refers strictly to the node-by-node pattern of choice condition (C, U, or N) at each hierarchy level ( bin-sets (if applicable), bins, and elements). A ‘decision structure’ refers to the

what i is to threat, j is to decision structure. One difference is that j does not refer to an ordinal position of decision structure. There is no preferential ‘rank’, though this could be done. Rather, j is properly a nominal variable. Nonetheless, each decision structure can be identified, and an expectancy count of structure frequencies can be developed or estimated based on counting the number of occasions when the particular node and parameter configurations combine to produce a decision structure. Because Pr(ti) can

now be completely codified and located as a known distribution of probabilities within the set of decisional-control decision structures, Pr(j) is the likelihood of a particular

distributionPr(ti)j, characteristic of given decision structure j, being operative. As an

example, our initial work on distribution of decision structures assumed a ‘gentle prior’ (‘mild assumption’) of equal likelihood for all nine first-order scenarios (see p. 38). This meant a 1/9 chance of one of CC (j = 1), CN (j = 2), CU (…), NC, NN, NU, UC, UN, and UU (j = 9) determining relative access to the set of ti values for the decision-maker.

The assortment of j decision structures can itself be considered governed by the

availability of choice conditions (C, U, and N) and the set sizes at each choice node (P, as applicable, and p and q). These model parameters can be conceived of as being potentially in short or uneven supply, hence benefitting from prudent administration. Allotment of choice to a subordinate node may be costly to a super-ordinate decision- making unit, if overarching concerns are not being met or system-wide considerations become difficult to address. This may especially be true if error-free ‘maximax’ decision- making (selecting to obtain the best) is not occurring at subordinate nodes. Conversely, subordinate decision-makers may find their super-ordinate decision-makers make more

errors in their decision-making than subordinate agents. The rise of tyranny (removal of subordinate freedom) and groundswells of social upheaval (toppling of corrupt regimes) might well be influenced by comparative decision-making efficacy.

combination of choice conditions by node and a parameter set size values ( P(if applicable), p, and q). A ‘decision scenario’ is the expression used to refer to a particular arrangement of choice conditions and set sizes, such as CUN(4,2,3).

In sum, the capacity to expect and produce desired outcomes has been housed within five ‘levels of governance’ as modeled and distributed statistical quantities: event outcome (m

= 0,1), event probability (Pr(m = 1) = ti), access to event probability ( Pr(i) for Pr(ti) ),

likelihood of a given access to event probability ( Pr(j) for a given Pr(ti)j ), and the

availability of decisional control parameters C, U, N, and (P), p, and q for creating the distribution J of decision scenarios ZC,U,N;(P), p, q each with an indexing identity denoted

specifically with the label of lower-case j. This general distribution J then provides the context for relative frequency of a given decision structure j, with decision structure j

governing a probability distribution Pr(ti). Discrete probability distribution Pr(ti) in turn

allots the chance of obtaining a good event probability ti, and a favourable tihopefully

allows better-than-random chance of event non-occurrence (in the case of threat), or of event occurrence (in the case of a desired outcome).

3

A Dynamic Catalog of Decisional Control Values