Decision Making (DM) via Artificial Intelligence (AI) 

INDEX:

“Revolutionary” + evolutionary APProach

Human Decision Making

“CHESS” tree

Human search vs “electronic brain”

SELection criteria

MINIMAX & Alpha'Beta algorythms

NEGAMAX & Iterative algorythms

TRANSPOSITION grids

HIgh-PERFOrmance “Chain Rules” !!

..................................................................................................................

“Revolutionary” + evolutionary APProach:

Science-Fiction legend “human vs machine”,

supposing threat in a future era where “silicon monsters”

would be superior to humans regarding practical results

within analysis OR Decision Making (DM)... is here!!

As in other history leaps,

Science-Fiction speculations TREND towards Science...

Since 1930's, STandarD MODel for Artificial Intelligence (AI) R&D studies

has been CHESS... Why?

Being a “Science-GAMe” framed within

a LIMited space with infinite possibilities;

OPTimum MODel for any multifunctional and multi-parametre analysis.

Taking into account CHESS programs such as IBM's Deep Blue

towards the end of 20^th. century, it has convincingly DEMOnstrated

its “practical superiority” compared to “human super-specialists”.

The “revolutionary” APProach is precisely

to accept such “practical superiority”

and “TRANSLATE” programming criteria to human terms-methods

in a MINimalistic plus practical way.

Let us see the logic chain of such “eclectic-multidisciplinary” method:

1^st.) Problem 

2^nd.) Human APProach 

3^rd.) Decision Tree analysis following Artificial Intelligence methods

4^th.) 3 final OPTimum branches SELection via MECHanical criteria

5^th.) Balancing step 2 vs 5 (seeking analogy OR intersection)

6^th.) Results reinterpretation & EVALuation at human LeVeL

7^th.) Final Decision Making (DM)

8^th.) Decision Making (DM) work-out within case-to-case basis

9^th.) Specific case RECord for INFOrmation's sake (REC INFO)

10^th.) Logic chain restart...

Human Decision Making is based on the following “natural” function type:

F = f(x, y, z) ; being F the función and rest 3 dependent parametres (para-m).

We humans are LIMited to 3 parametres

on account of our CALCulus incapacity 

(50 cycles human calc. vs 100 million each 3 min. machine's),

our graph outline need 

(2D OR 3D patterns)

as well as by Physics laws 

(1^st. Parametre = SurFaCe < 2^nd. para-m= Space < 3^rd. para-m= time).

Human Decision Making (DM) tree (human fork) would be the following:

Yes vs No vs ? = Fork MODel = Survey MODel...

“CHESS” Decision Making (DM) tree would draw only via SELected branches.

Let us see such “CHESS” tree at NUMerical LeVeL (NUM LVL):

1^st. MOVE = 20 possibilities or branches OR decisions OR parametres OR...

2^nd. MOVE = 20² = 400 possibilities

3^rd. MOVE = 20⁴ = 160.000 possibilities

4^th. MOVE = 20⁶ = 64 million possibilities

5^th. MOVE = 20⁸ = 25.600 million possibilities!!

... and so on and so forth...

1^st. OPENing phase = 10 to 15 MOVEs = 500 million x 10²⁵ possibilities approx.!!

2^nd. Mid GAMe phase = 15 to 30 MOVEs = 1000 milliones x 10⁵⁰ possibilities approx.!!

As you can see, Decision (DM) tree GROWs EXPonentially

up to unCALCulable quantity; programing TarGeTing focuses upon algorithms which do 1^st. SELect good MOVEs so as to only CALCulate meaningful lines.

Main difference () between both methods would show VISually.

The Key Performance Indicator (KPI) of our novelty

would be to apply the strongpoints of both search methods

so as to be able to combine and interpret them correctly!!

Balancing & analogy between human search vs “electronic brain”

Retracing our steps to human search tree, we do have:

  Yes OR 100% OR Present OR 1^st. ...

Problem No  OR 50% OR Past OR 2^nd. ...

? OR 0% OR Future OR 3^rd. ...

As you see, our daily (d) decisions, realistically framed,

are based upon STandarD criteria “survey” TYPe OR “percent” (%)

OR “history cycle” OR priorities "podium-like” OR “ABC” STOck Ctrl. method

OR “industrial dispute” MODel (Co. vs mediator vs workers)

OR “Function vs VARiable vs result” OR... etc.

Such human “fork”, DEMOnstrates perfectly our mind pattern

in front of any kind of Problem Solving.

Doubtless, such pattern works properly in most cases;

nevertheless, as a SYStem OR situation gains complexity,

such method becomes random and we end by applying dubious “intuition”

(intuition natural tempo = 2'' < 3' = MINimum analysis tempo).

So as to overcome the apparently unsurmountable barrier

of ever increasing complexity, let us go back to “machine” tree analysis:

 Problem

Omnidirectional exhaustive search (360 “brute force”)

vs

EXPonential complexity (infinite TREND = )

vs

“Pruning” TECHniques

(cut “branches” for OPTimum sequence SELection)

vs

Final SELection

If we draw our attention to “machine” final SELection

vs human Decision Making (DM) tree; we detect symbolic analogy

between 3 final MECHanical “branches” and human “fork”.

In both cases we focus 3 final decisions...

in common life or in daily situations human Decision Making (DM)

proves to be valid and applicable, but...

how to achieve clear-cut criteria in a scenario of enormous complexity?

Simply “translating” “pruning” TECHniques to human method;

brilliant idea, no doubt, easy to say

though critical to Research & Develop (R&D) a reliable method...

Nevertheless, if we think about it carefully,

we are really swaping the normal PRocess:

1^st. We MODel ideas and after we program;

such trascendental NOVELTY allows using programing heuristic criteria

to our LeVeL as well as allowing to “embrace infinite”

with finite SYStems.

It is also useful to verify IF a program or method or SYStem

does NOT forward practical results, being quite common nowadays,

considering that most so NAMed “multimedia generation”

accepts a final result within display with an “Act of God”...?!

So let us Design & Develop (D&D) STandarD “pruning” TECHniques

translatable to human method:

1^st. MINIMAX algorythm:

In descriptive terms following “GAMe Theory”:

MAXimum own profit linked to 3^rd. Party MINimum profit.

Representing “CHESS” Decision Making (DM) tree:

MAX

VALue

vs

MIN

Value

In LeVeL 1 (1^st. Branching) we do SELect our Decision Making,

taking into account the less possible Threat of LeVeL 2 OR “environment”

(alien OR 3^rd. Party OR MarKeT OR competition OR opponent OR RISK

OR negative issues...).

“Translating” MINIMAX to human method:

1^st. SELect alien MINima and 2^nd. SELF SELection

2^nd. Alfa-Beta (α-β) algorythm:

In descriptive terms:

Abandon branches with less value so as to save energy.

In LeVeL 3 (3^rd. branching) we do SELect our Decision Making (DM),

choosing our MAXimum VALue so as to save enery and CALCulus.

“Translating” α-β to human method:

2^nd. SELect highest value branch and abandon the rest

3^rd. NEGAMAX algorythm:

In descriptive terms:

Forget the values' sign (+ or -) so as to associate them to a unique analysis.

In LeVeL 5 (5^th. Branching) associating all VALues forgeting about their sign

(applying “ABSolute value”).

“Translating” NEGAMAX to human method:

3^rd. associating all values to a pondered average for EVALuation

4^th. ITERATIVE algorythm:

In descriptive terms:

ITERATIVE search algorythm based in EVALuating possibilities

to a FIX depth balancing available time to avoid so NAMed

“horizon effect”, just viewing:

-∞ ---------- CALCulus LIMit ---------- 0 ---------- CALCulus LIMit ---------- ∞

Representing “CHESS” Decision Making (DM) tree:

CALCulus

with

LIMited time

“Translating” ITERATIVE to human method:

4^th. SELective search LIMited to available time and capacity

5^th. TRANSPOSITION grids:

In descriptive terms:

Pattern recognition...

And so on and so forth until our “thought MODel” is reinforced

using such Artificial Intelligence (AI) criteria

once translated to human terms and used as Decision Making tooling...

HIgh-PERFOrmance “Chain Rules” !!

(AZ ref.)

“Chain Rules”, as HIgh-PERFOrmance tooling,

are a very powerful MEMory plus logic support

to any kind of step-by-step criteria, PRocess, PROCedure and so on...

It is also HIgh-PERFOrmance Fast Forward (FF) reading OR studying,

for a single A4 page of “Chain Rules” = 20 pages written text !!

It is a common tool used in HIgh-TECHnology sectors

where complexity and too much length is overwhelming indeed,

so that in a single PAGe, via “flash effect”,

you VISualize relevant INFOrmation.

Of course, we will only sample most relevant to GAMing:

(the reader may SELect those which seem better for him)

Chain Rule

ATTACK = Approach To Threshold Anticipating Counter Knock-out

CALC = CALCulus = Calculate As Limit Check (calculus limits)

CHECK = Cancel Harmony Evolvement Checking King (unbalancing)

CHESS = Choose Highest Evaluation Strongest Set (MAXIMAX)

COUNTER = Counter Only Upon New Target Expecting Results

DECIDE = Detect Estimate Choose Identify Do Evaluate (DM steps)

DEMO = Do Each Measured Option (trial & ERRor criterion)

DM = Decision Making

DRAW = Do Relax After War (Emotional Intelligence [EQ] DRAW)

ENDING = Endgame Next Decision If Newfound Goal (forcing ending)

ERR = ERRor

EVAL = EVALuation = Estimate Values And Limits

GAM = GAMing = Go And Measure (experimental metrology)

GROW = Goals Reality Options Will (Decision Making parametres)

KO = Knock Out

LOSE = Limit Options Solving Errors (Problem Solving criterion)

MOVE = Move Options View Expectations (forecasting criteria)

NO = No Options

OPEN = Own Possibilities Evolvement Net (SELF SWOT)

OPT = OPTimum = Options Preferences Trend

ORG = Organisation Resources Goals (horizon analysis)

PAX = Peace At X-point

PERFO = PERFOrmance = Plan Each Role For Outcome

PDCA = Plan Do Check Act (Quality cycle)

PIN = Pin If Nuisance

PWR = PoWeR = Potential Weight Resources (potential judgement)

RISK = Risk If Sure Know-how (RISK analysis criterion)

...