Drilldown: //Grand Framework 0.1.0/COBRA
The COBRA framework presents an extended form of cost-benefit analysis which can be used to model decision making processes that are involved in the other frameworks.
Framework Navigation
Top: Grand Framework
Up:
Down:
Siblings:
Child:
Attributes
| Attribute | Value | Notes |
| Class | CobraFramework | |
| Name | COBRA | |
| Type | Framework | |
| Symbol | ||
| Namespace | GrandFramework_0_1_0.COBRA | |
| Version | 0.1.0 | |
| Description | ||
| Example | ||
| Navigation | Drilldown: //Grand Framework 0.1.0/COBRA |
Properties
Context
The presenting situation environment or predicament. Often this means a change in the state of the environment to something less favourable or the emergence of an opportunity.
Operant
Some behavioural response or action. Usually in response to some change in the environment in an attempt to resolve a problem or capitalise on an opportunity. There may be many potential operants to a context, but this model assumes that only one is applied.
Benefit
A benefit is mostly about what positive aspects can be derived from responding to a situation. However, in this model, the benefit can also be negative. Different operants, when applied to a context, produce different levels of benefit.
Risk
The COBRA framework does not automatically assume that the operant in a context guarantees an outcome or benefit, but rather that there may be some probability involved. The risk factor considers the chances that a particular operant will work out; it can be considered as a measure of confidence.
Allocation
Performing an operation requires the investment of resources, which are usually of limited supply. Resource allocation is about those resources that can be given to attempting a particular operation. It might not be sensible to allocate all resources to a particular operation, especially if there are further or conflicting needs, or that operation is one carrying a high degree of risk.
Specific Classes
In brief
Symbol system
Details
[more detailed use of COBRA analysis on seperate page – not the framework documentation page]
keep this bit:
COBRA is an elaboration on the more traditional cost-benefit analysis (CBA), but has been adapted and expanded for contextual behavioural science in decision making and also to be consistent with the other cliogical frameworks. The cost-benefit analysis method was developed for making decisions about investments involved in public policy. It is a process of economics and is usually reduced to financial considerations. Some variations also consider other factors such as risk and uncertainty.
COBRA can be reduced to the quantifiable metrics of a common currency, and can be automated using software applications; but one of the points of COBRA is that strict quantification is not essential and vaguely defined or poorly understood, non-quantified, and intuitive parameters can also be applied to produce, while not as accurate, at least some guidance. This is of heuristic benefit where precision is not available. The essence of COBRA is that such decisions are being made by people without economic training all the time (even though they do not know it). Indeed the principle is intended to generalised to unconscious system 1 decision making, and further to most, if not all, behavioural response selection, whether neural or genetically determined, of all life-forms. Hence, behaviour can be assessed in terms of this “equation” whether the decision-maker is conscious of it or not – it is reduced to “mechanistic” terms and considers intelligent decision making to be on a continuum.
Moreover though, for individual humans and complex social organisations, the equations can be applied systematically with the aspiration of making better decisions.
There are similarities to CBA.
- Benefit is roughly the same concept, although intangibles are incorporated. The ultimate bottom line here is Darwinian – that of fitness and survival. Financial gains and losses subsume this survival need.
- Allocation of resources roughly maps to Cost. The C, in the extended version, no longer refers to cost, but rather to context.
- The ideas of Operant and Risk are also introduced.
- Context and Operant come from contextual behavioural science: the context being the situation, and the operant being the behavioural response to that situation. The COBRA analysis though considers a range of contexts, a range of responses, and the suitability of those responses to the contexts.
In essence, COBRA analysis surveys a landscape formed across a range of parameters. It can be viewed as a weighted matrix whereby each cell is calculated from Benefit, Risk and Allocation. The ideal being that, given a particular context from among the set of possible contexts, then the optimum response can be selected. This corresponds to the noam framework of selection and enactment of a kinetic response (rho) from potential responses (k).
The model considers that from the many possible contexts, one actual context, consisting of a cluster of contextual traits, is presented to a system (individual, organisation or organism) at one time. The system has a repertoire of potential behaviours it can draw upon, and can select one “kinetic” behavioural response, consisting of a cluster of behavioural traits. This is performed in an attempt to maximise the systems survival chances. It tries to do what is best. COBRA then is how the system selects the kinetic from the potential in an attempt to optimise benefit.
A decision is based on the presenting information meaning that the response is a reaction to a situation such as an emergency. COBRA can also be used in what might be seen as a “proactive” mode. This is where a sequence of situations tend to occur in a predictable order: where given event A (such as lightning), then event B becomes likely (ie thunder). COBRA would be a reaction to the foreshadowing event but be proactive, preparing a state of readiness, for event B. However, as the foreshadowed event is only a probability given the foreshadowing event, then proactive behaviour can be termed in a range between “insurance” or “investment” which would be based on the users model (experience) and confidence.
Further work needed when main section done.
Move all this:
Elaboration of properties
There are 5 properties in this model, 4 of which (CORA) are factors which go into the remaining factor (B) is the result of a “calculation”. The aim is to find the optimum benefit from among the array of choices which would give an indication of what operation to chose. The four contributing factors can be thought of as internal or external to the system; or as qualatative categories or quanatative values. Ideally, the calculated result of benefit would be numerical.
- External factors are the context that the system faces, and the risk (or certainty) of a particular option, and are such that the system has little control over.
- Internal factors, the system has more control over, and are the operant and the allocation of resources.
- The qualatative categories are the context: what type of situaion is being presented, and the operant: what type of behaviour should be enacted. These is a one of many type parameters; a Boolean type.
- The quantative values are the assessment of risk and the amount of resources to be allocated. These can be thought of as numerical values.
The relationship between the factors can be seen as a grid.
| Qualatative categories | Context | Operant |
| Quantitative values | Risk | Allocation |
| External | Internal |
Benefit is not on that grid as it is a resulting evaluation.
A way of thinking what is happening is that the external categories factor, that of context, is the input coming into the system from the environment: the sensory information that a person or organisation may experience. The internal categories, the operants, are the behavioural output to the environment chosen from the potential repertoir of responses. This is from the system’s point of view as the actual dynamics would be cyclic. The external and internal categories of contexts and operants can be rearranged in a different grid to provide the major horizontal and vertical axes respectivly. Between the range of possible contexts and the range of possible operants would be a table whereby each combination of context and operant would have an intersecting “cell”. These cells would be a measure of how appropriate a given response would be to a certain situation. hence, presented with a situation and the table values, then it would be possible to infer the best course of action.
| Internal | ||
| operants | [ intersection table] | |
| contexts | External |
From a cliological perspective, the internal and external categories are clusterable – that is, they can be depicted as a tree. The contextual tree would depict the similarities between situations, while the operants tree would show how closely behaviours were related. These structures are used in determining “eka-stratagems”, of novel situations and novel ways of dealing with them. The intersecting table between the two axes could be viewed as a fitness landscape – of how befitting an action is to an environment or emergency and would allow for hill-climbing optimisation algorithms.
The external and internal axes are primarily categories, but they have their values counterparts paired up with them. Context can be paired up with risk; operant can be paired up with resource allocation. It is from these values that we can attempt numerical manipulation and derive the benefit measures, which is the value of the intersecting cells and basis for the decisions. Here, we start to move into game-theory payoff matrices.
| operant C | |||
| operant B | |||
| operant A | |||
| context A | context B | context C |
COBRA range from intuitive rule of thumb with many poorly defined, understood, or measured parameters but is best with accurate numerical values.
The intersecting cells are calculated from the numerical values of benefit, risk and resource allocation. The simplest cases are where the allocation and benefit are well known and there is absolute certainty in the results of an action; this is similar to traditional cost-benefit analysis. Here we are interested in the amount of benefit per unit allocation. The “break-even” point is where the ratio equals one: a “profit” where more; a “loss” where less (other factors may be involved in commercial investment).
Risk, or rather the reciprocal of certainty, comes into play where the outcome is probabilistic – for example, the success of a particular operation may be one in ten (ie 1/r=0.1); where success is assured, as in the simple case then the risk factor is 1. The formula for each cell then is:
(Benefit/Allocation) * 1 / Risk or Benefit/(Risk * Allocation) ie b/ra
Consequently, a low benefit per unit resource allocation with low risk may be equivalent to a high benefit per unit resource allocation but with a high risk attached.
(1/1)*1 = (10/1)*0.1
Each intersecting cell would then have its b, r and a values which would yield the b/ra calculation as a payoff. The highest payoff would be a good candidate of choice.
Reaction and proaction
Spread betting
Resource allocation (A) concerns what resources are put into an operation given a context. In general, there is a finite amount of resources available which must be optimised against a series of contexts. These resources are budgeted according to likely contexts, benefits, and risks. For example, a herbivore such as a rabbit, surviving in the wild, at any time, will only have a certain amount of energy available, and a limited degree of attention it can pay to what is going on. This rabbit must forage for food in order to replenish its energy, while at the same time be alert to predators. It also must make baby bunnies (play, dig burrows, sleep and all that other rabbit stuff). Its energy and attention would be divided between such activities – over-concentration on any one activity could result in being fox-food, or starvation, or a decline in the population. The rabbit’s involvement is ongoing and not just a one-off activity. This means that it must allocate some energy into what it is presently doing while reserving some energy for further action. So, even if feeding yields the highest return at the present, the rabbit cannot put all of its energy into feeding. It must hold back some energy for in the event that a hungry fox turns up. There is a trade-off here between finding food, avoiding becoming food, and reproducing. While emergencies and responses are likely (but not necessarily) sequential, allocation is done in parallel but requires constant re-evaluation as events transpire.
Two factors are intertwined with the others and influence how resources are budgeted. These are risk (the probability of an eventuality and success of an action) and time, which involves the immediacy of the consequences. A delay in gratification does increase risk as there is the heightened probability that some subsequent event may inhibit the pay-off. An immediate pay-off, on the other hand, has greater certainty. Therefore, there is a priority in resource allocation. The rabbit must prioritise preditor alertness and response otherwise it would be denied the further opportunity to digest or gestate. There are serious disadvantages to being a fox’s dinner (the benefit to the rabbit might be thought of as minus infinity) but when any immediate threat has passed, it can concentrate on avoiding starvation. Once safe and nutritionally satiated then there may be time for love.
There is a range of stress levels as presented by the current context from mortal danger to leisure. The present state is known the subsequent states are not.
risk consists of unknown factors including the likely success of an action
Traditional cost-benefit analyses do not explicitly consider the comparative situations and responses and does not formulate a spread of allocation. COBRA does not consider the relationship of cost to benefit in isolation but rather takes a wider, more contextual picture so as to evaluate what might be best from a range of options.
| fornicate | No | No | Yes |
| feed | No | Yes | no further benefit |
| flee | Yes | waste energy | waste energy |
| fox present | hungry (no fox) | satisfied (no fox) |
{victor/victim}{insure/invest}
To abstract, the table for game-theoretical purposes, the two axes of context and operant can be arranged from “negative” to “positive” with the “negatives” to the bottom leftmost corner but ranging up to the least stressed, or most leisurely. The operants can be abstracted to “insure” to “invest” with taking no action in-between. The outcomes would be to lose “victim” or to win “victor” again with a neutral value between them of neither winning nor losing.
Suppose in a thought game (a mundane and non-paradoxical one) that a rabbit can focus its attention on either looking for carrots (“invest”) or watching out for foxes (“insure”), or it can just sit around. The more advanced warning it has of a preditors presence, the less it has to scramble for its life should a fox come to victimise it. On the other hand, if it concentrates (“invests”) on finding carrots, it may be victorious. However, it cannot concentrate on two things at once. Neither the presence of a carrot nor the absence of a fox (“victor” or “victim” contexts) are assured – the situation may be neutral.
This game plays out to nine configurations. Some fictitious costs and benefit figures have been allocated to illustrate how payoffs might work – these are chosen to give a somewhat neutral example in terms of decision making in that there is no particular advantage to any choice; they are of equal merit. The outcome probabilities (r) are balanced; both insure and invest cost one; the benefit (or lack thereof) of not finding a carrot or not being chased is zero; other configuration payoffs are as given in the table.
| victor | a=1; b=0; r=1/3: net=-1/3 |
a=0; b=0; r=1/3: net=0 |
a=1; b=8; r=1/r: net=7 |
| neutral | a=1; b=0; r=1/3: net=-1/3 |
a=0; b=0; r=1/3: net=0 |
a=1; b=0; r=1/3: net=-1/3 |
| victim | a= 1; b=0; r=1/3: net=-1/3 |
a=0; b=-3; r=1/3: net=-1 |
a=1; b=-6; r=1/3: net=-7 |
| insure | neutral | invest |
- insure: the rabbit has been looking out for danger (allocation: cost=1)
- victim: a fox turns up but the rabbit has plenty of time to scurry off to its burrow.
- neutral: no foxes turn up but there were no carrots to find any way
- victor: there is a carrot but the rabbit was not looking for it
- neutral: the rabbit is idling (allocation: cost=0)
- victim: a fox turns up but the rabbit still has enough time to hop away (benefit: =-3).
- neutral: no foxes turn up but there were no carrots to find any way
- victor: there is a carrot but the rabbit was not looking for it
- invest: the rabbit concentrates on finding carrots (allocation: cost=1)
- victim: a fox turns up starling the rabbit which has to run for its life (benefit: payoff =-6).
- neutral: the rabbit searches for a carrot but there is no one to find.
- victor: the rabbit finds the carrot (benefit: payoff =+8)
This idealised scenario is deliberately unrealistic for a number of reasons, but as each operant accumulates to -1 then the population would go extinct (a more realistic dynamics would feature the Lotka-Volterra curves of fox and rabbit populations). However, we might alternatively think of the rabbit as “keeping an eye out for predators” while looking for carrots which would increase the risk compared with a pure insurance strategem, but yield fewer carrots than a pure investment strategem.
| victor | a=0.5; b=0; net=-0.5 |
a=0; b=0; net=0 |
a=0.5; b=4; net=3.5 |
| neutral | a=0.5; b=0; net=-0.5 |
a=0; b=0; net=0 |
a=0.5; b=0; net=-0.5 |
| victim | a= 0.5; b=-1.5; net=-2.5 |
a=0; b=-3; net=-3 |
a=0.5; b=-3; net=-3.5 |
| insure | neutral | invest |
A slightly more realistic scenario would be where rabbits are borne and thrive in an abundance of food and the reduced threat of predators (again avoiding the real-world dynamic complexities of attracting foxes and diminishing food supplies). Here, we would see a shift in the probabilities of the context [tables needs doing properly with r] whereby the chances of being a “victor” is increased while the chances of being a “victim” decrease. Note that the sum of probabilities for all of the available contexts remains constant at one – one of those things is going to happen. This simple model illustrates three available contexts, but the same would be true where there are many more providing that all the available contexts are considered.
| victor | a=1; b=0; net=-1 |
a=0; b=0; net=0 |
a=1; b=8; net=7 |
| neutral | a=1; b=0; net=-1 |
a=0; b=0; net=0 |
a=1; b=0; net=-1 |
| victim | a= 1; b=0; net=-1 |
a=0; b=-3; net=-3 |
a=1; b=-6; net=-7 |
| insure | neutral | invest |
These shifts would reflect the weight put on a particular choice.
choice of migration, mutation, mating strategems of COBRA
concurrence with noam framework (the algorithm) see photo from pub
to machine executable decision support with clearly defined and precicely measured parameters.
which to use is a subject for cobra as the context and resources available come into play
Perverse incentives the cobra effect
Interactive and state diagrams (negotiation)
Behavioural economics etc.
Relation of other framework elements
Examples
{law, science, sales (direct, mlm), PR, military, thermostat, environmentalism, the cat sat on the mat}
Sales (direct)
Application
Related Issues
Versioning and change-log
Origin and history
This started from a personal dissatisfaction with traditional CBA which seemed to lack essential components in simple (personal) decision making. Adding risk pushed the acronym to CBRA, which read cobra. The acronym took some thinking about and some massaging to hit its final COBRA form, but the idea seemed to better embrace ideas of behavioural science; those of context operant. Context replaced Cost in the acronym and Cost was replaced by Resource allocation in principle. the R in Resource allocation clashed with risk, and so was just reduced to Allocation. The “Analysis” was dropped from the acronym itself. Upon doing a COBRA Analysis, a retronym occurred to me – I was considering whether some project would be worth doing or could I be bothered, it became: Could One Be Really Arsed. Which summed the sentiment of the method.
The model is also a reference to the cobra effect which alludes to the concept of perverse incentives
COBR[A] is also the code name for the UK government’s emergency room
Inter-COBRA
your genes dont write your destiny
a species’ superpower
junk values making us mentally sick