=Paper=
{{Paper
|id=Vol-2919/paper11
|storemode=property
|title=On the Formal Description of Decision-Making in an Intelligent System with Technology of the Direct Imposition of Knowledge
|pdfUrl=https://ceur-ws.org/Vol-2919/paper11.pdf
|volume=Vol-2919
|authors=Gennady Bronfeld,Dmitrii Kirov
}}
==On the Formal Description of Decision-Making in an Intelligent System with Technology of the Direct Imposition of Knowledge==
https://ceur-ws.org/Vol-2919/paper11.pdf
On the Formal Description of Decision-Making in an
Intelligent System with Technology of the Direct
Imposition of Knowledge
Gennady Bronfeld 1[0000-0003-0164-8303] and Dmitrii Kirov 2[0000-0003-2333-5826]
1
Nizhny Novgorod State Technical University,
Nizhny Novgorod, Minin St., 24, 603950 Russia
stolem1985@gmail.com
2
Nizhny Novgorod State Technical University,
Nizhny Novgorod, Minin St., 24, 603950 Russia
tmdagger2000@gmail.com
Abstract. In this paper several approaches to decision making in intelligent sys-
tems of various kinds are considered. The most commonly used decision-
making approaches are considered - engineering approach, purely scientific ap-
proach, system analysis, logical solution of problems, heuristic search, "the the-
ory of inventive problem solving", direct search for solutions, method of “ra-
tional” decision-making. For a new intelligent system based on technology of
direct imposition of knowledge with knowledge models in the form of
molingas, a rarely used approach to solving problems for such systems is ap-
plied, as once put together by the famous mathematician Pólya - direct search
for solutions. This reinforces the applicability of representation of such systems
as Ackoff-Emery purposeful systems. The experience of use of a formal de-
scription of decision-making processes in such systems from the point of view
of system analysis is applied. Several of comparative characteristics of deci-
sion-making systems related to intelligent systems are considered in comparison
with the new direction of intelligent systems based on technology of direct im-
position of knowledge. The analysis revealed a number of significant ad-
vantages of such a system compared to a number of approaches used for deci-
sion-making solutions, which is important in practice.
Keywords: knowledge, decision making, logical inference, problem solving.
1 Introduction
Throughout recent decades, active effort has been set off in different countries
around the world to create various decision-making (DM) systems using artificial
intelligence (AI) as the means [1-8]. Within the framework of systems based on AI
methods currently in use, a limited set of options from known approaches to problem
solving get applied [9-13]. There is a certain degree of tradition to a formal descrip-
tion of DM processes in such systems (from the point of view of system analysis)
[2,11, 14-18]. The approaches are very different from each other due to the complexi-
Copyright © 2021 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0
International (CC BY 4.0).
Proceedings of the of the XXIII International Conference "Enterprise Engineering and Knowledge Management"
(EEKM 2020), Moscow, Russia, December 8-9, 2020.
ty of the problem. Let us consider some comparative characteristics of modern DM
systems related to intelligent systems (IS) in comparison with the new direction of IS
based on the technology of direct imposition of knowledge (TDIK) [19-22].
Many of them belong to the category of purposeful systems [9], wherein IS “can
change its tasks under constant environmental conditions: it chooses the means of
their implementation. In so doing, it displays its’ own will.” This is due to the fact that
in their case an active role is played by a person, which brings about, in some cases,
certain features of structural makeup of algorithms and, as is typical for all ISs,
change (more or less gradual) of knowledge bases used in ISs. These structural fea-
tures of purposeful systems are particularly clear in IS with TDIK [19-22].
Let’s consider the following main ISs that allow you to find solutions, for example,
expert systems (ES) [1,7], IS Watson [6], in the dialogue mode with the decision
maker (DM) in the IS with TDIK [19-22]. Then let’s compare them with the more
standard approaches - the engineering approach [11, 12,14,17,23], work with the In-
ternet [3,4].
2 Analysis of existing approaches to solving problems /
tasks
DM by humans is a relatively routine everyday activity. How a person (or other living
creatures) does it is investigated in sufficient detail by psychologists [13], neurophys-
iologists [24], applied engineering specialists [11,12,14,23], and even mathematicians
[2,10,16] and, specifically, AI specialists [1,7,15].
Psychologists (whose opinion is considered cornerstone) have yet to find a relatively
acceptable solution on how a person thinks, and this understanding constantly slides
further towards ever distant future [13, 24].
The mathematician Pólya believes “... a universal and infallible method for solving
problems, unfortunately, does not exist: strict rules applicable to any situation have
not yet been found and, in all probability, will never be found” [10].
“As the solver moves along towards solution, the appearance of the task is con-
stantly changing. At each stage of the problem, the solver encounters a new situation,
and once again faces the question of choosing the correct intermediate solution: what
should be done in such a situation, what should be the next step? If he knows the per-
fect method, the infallible strategy for solving problems, then he can choose the next
step via reasoning alone, based on the current situation and guided by a set of clearly
defined laws [10]. In various embodiments, this is emphasized in [11, 14,17,23].
Over a long while, a set of approaches has developed that allows us to find solu-
tions to problems / tasks:
engineering approach;
a purely scientific approach, in particular, solving problems in mathematics;
system analysis;
logical problem solving;
heuristic search;
“the theory of inventive problem solving” (TIPS);
“brainstorm”;
direct search for solutions;
Let’s consider some of the most well-developed technologies in science and technol-
ogy for solving tasks / problems.
2.1 The engineering approach
The engineering approach to finding solutions is based on the widespread use of tech-
niques, technologies, standards, research, development, in use sometimes for millen-
nia. They are not always scientifically substantiated, but give good results in practice
[11,12,14,23].
Disadvantages:
these techniques, technologies, standards for each of the sub-directions with thou-
sands of design options and technologies, there are quite literally thousands. No
single person can know all of this; existing IT technologies are also insufficient.
Therefore, either outdated solutions are made use of, or insufficiently effective, or
decisions get not made at all;
not finding the right solution, the decision maker is not able, as a rule, to quickly
figure out and conduct at least some analysis. This leads to the fact that the deci-
sion maker cannot perform the task, or leads to gross errors in DM.
2.2 A “purely” scientific approach for finding solutions
A “purely” scientific approach for finding solutions, in particular, solving problems in
mathematics, has also been developing for thousands of years and is based on strict
methods for conducting research, experiments, processing results, checking results
and discussing them. It does not always lead to quick results though. [2,10,15].
The engineering approach gives a quick answer, you can find solutions to tasks /
problems and implement what you need, or you can’t do it at the current level of de-
velopment. The scientific approach, if it gives a positive answer, is inaccurate (alt-
hough it may seem precise). It must be checked and implemented often with respect
to a point of view of engineering approach. The scientific approach will either not
give any answer, or the answer will be incorrect (something is not taken into account),
or the answer will be no. But it’s not certain that there is no decent solution. For ex-
ample, a decision maker can find (invent) new solutions that cannot be reached using
a purely scientific approach.
2.3 System analysis
System analysis developed as a kind of symbiosis of the scientific approach and the
engineering approach to finding solutions. System analysis is based on the division of
great uncertainty into more manageable parts, identifying relationships, relying on the
mathematical theory of systems, clarifying the structure of goals, identifying criteria
for their achievement, using not only formal, but also qualitative research methods,
while ensuring the participation of specialists from different areas of knowledge
[2,9,14,18].
System analysis alone is not able, as a rule, to find a satisfactory solution without
the use of engineering approaches, in particular, due to its abstractness.
2.4 The logical solving of tasks and problems
The logical approach to solving tasks and problems has been considered one of the
key and of most inherent in man approaches for quite a while, and is widely used
[1,13,15].
Disadvantages:
in order to perform logical inference (LI), objective knowledge of the relationship
of individual concepts and facts with each other is necessary. This is possible only
with scientific or engineering approaches, i.e. logical thinking does not have inde-
pendent significance for solving problems;
even with relatively small volume of initial data and concepts, the volume of logi-
cal connections begins to tend to infinity and a person without assistance of special
technologies is unable to adequately draw conclusions.
2.5 Heuristic search
Heuristic search is a way to solve problems, based on special methods to reduce the
enumeration of options and solutions called “heuristics” [1, 13]. One of the simplest
options for heuristic search is actively used in the direction of AI based on heuristic
evaluation functions [1, 19].
Disadvantages: insufficient validity of a number of applied techniques; lacking ef-
fectiveness.
2.6 “The theory of inventive problem solving”
The theory of inventive problem solving (TIPS) and on its’ basis the algorithm of
inventive problem solving (AIPS) was developed by G.S. Altshuller and his support-
ers during the 40-80s of the twentieth century [5,11-13]. It is based on the initial iden-
tification of contradictions. And through a sequential procedure, a situation is manu-
factured that is assessed by psychologists as "insight” [13,20], leading to the neces-
sary solution. AIPS was implemented as a software package on a PC called "Invent-
ing Machine" back in the days of the USSR [12,20].
Unfortunately, TIPS could not be used to solve most of the practical problems,
which was partially corrected after Altshuller’s death by his followers [5,11,12].
2.7 Direct search of solutions
Direct search for solutions is a special approach to the perception of solving a certain
problem. It was described at time by the famous Hungarian mathematician G. Pólya
[10] – “To find a solution to a problem means to establish a connection between pre-
differentiated objects or ideas ... like ... a chain - perhaps it will be a long chain - of
conclusions. The whole chain ... is not more durable than its weakest link ... ".
This is a rarely proposed and used method. The drawback (in the past) is the lack
of clarity on how to practically implement it.
2.8 Method of “rational” problem solving
“Rational” problem solving is an approach that has been taught to managers and ex-
ecutives around the world in recent decades, gradually being refined in formulations.
Making a rational decision is not an elementary act. It is carried out in several stag-
es [17]:
1. Formulation of the problem.
2. Setting goals and formulating the problem of choice.
3. Finding alternatives and determining their properties.
4. Design of selection model (DSM).
5. Solving the selection problem based on the selection model.
6. Assessment of the consequences of the choice and quality of the solution.
7. If unsatisfactory, then return to 3 (or 2, or 1).
This is more realistic in practical situations. Moreover, the degree of approximation to
optimality is determined by the level of pretentiousness of the decision maker, which
poses the question of whether he can achieve a solution to the problem at certain costs
and what consequences are associated with this.
3 General formalized decision-making procedure
Now we shall provide a formalized procedure for finding solutions (based on
[2,10,11,16]):
1 i i 1 R
P(W , G, C , I , Q, Z ) ... Pi (Wi , Gi , Ci , I i , Qi , Zi ) ... R(WR , GR , CR , I R , QR , Z R )
(1)
Where P(W, G, C, I, Q, Z) - description of a certain initial state, in which
─ W – a set of descriptions of the problem to be solved;
─ G – a set of desired states;
─ С – a set of goals and criteria;
─ I – a set of current information in possession of decision maker;
─ Q – a certain set of different descriptions in possession of decision maker;
─ Z – a certain set of knowledge of decision-makers’ (in the initial state of the sys-
tem).
In (1) Φ – some transformation operator that helps decision makers move from one
state to another. It can be a purely computer system with or without dialogue with a
person, a purely organizational system, etc., and i – transformation stage number.
In (1) R(WR,GR,CR,IR,QR,ZR) - the resulting solution to the problem after the last
transformation operator ΦR.
It should be noted that in (1) the concept of “state” (especially intermediate) corre-
sponds to the concept of “state at some point in time” [9], interpreted as "the set of
essential properties that the system possesses at this point in time." The “set of desired
states” in (1) then can be interpreted as a lot of variants of “events” that are desirable
in terms of achieving a “result” [9]. The “result” of the action of a decision maker
with or without IS is a change in this decision-maker (for example, obtaining infor-
mation or knowledge), or IS, or in their environment, produced by this action [9].
"Result" is the product of the action of the decision-maker or IS.
Now let’s consider sequentially different options for approaches to DM and vari-
ous ISs, with the IS in question equipped with TDIK in the variant of the elinga,
which was described in [19-22].
Let’s present a formal procedure for finding solutions in elinga as (2) [19-22]:
1El iEl iEl1
P(W , G, C , I , Q, Z ) ... Pi (Wi , Gi , Ci , I i , Qi , Zi ) ...
El El El El El El El
ElR (2)
... R(WREl , GREl , CREl , I REl , QREl , Z REl )
Where P(WiEl,GiEl,CiEl,IiEl,QiEl,ZiEl) - description of a certain initial state, in which
─ WiEl – a set of descriptions of the problem to be solved in the KB of elinga;
─ GiEl – a set of desired conditions given in the KB of elinga;
─ CiEl – a set of goals and criteria given in the KB of elinga;
─ IiEl – a set of current information in possession of decision maker;
─ QiEl – a certain set of different descriptions in possession of decision maker;
─ ZiEl – a certain set of knowledge of decision-maker.
In (2) R(WiEl,GiEl,CiEl,IiEl,QiEl,ZiEl) is the resulting solution by elinga of the problem
after the last transformation operator ΦREL. Let us make a comparison with other ap-
proaches based on some qualitative assessments.
4 Comparative review of formal decision-making
procedures
4.1 Comparison with the possibilities of the Internet
Most often, the capabilities of elinga are compared with the capabilities of the Inter-
net. Imagine a formal procedure as part of our methodology for finding solutions via
the Internet.
1Int iInt iInt1
P(W , G, C , I , Q, Z ) ... Pi (Wi Int , GiInt , CiInt , I iInt , QiInt , ZiInt ) ...
Int
R (3)
... R(WRInt , GRInt , CRInt , I RInt , QRInt , Z RInt )
Please, note, you usually cannot get to ΦRInt – the final stage of obtaining a solution
without making stops at the preliminary stages, drowning in vast volumes of interme-
diate information IiInt, the volume of which in thousands, tens of thousands, hundreds
of thousands of links is much more than the amount of information that will be used
in the process of problem solving by the elinga, i.e. |IiInt|>>|IiEl|.
Moreover, in most cases you simply can’t get to the stage ΦRInt – the final stage of
obtaining a solution or the necessary intermediate algorithms for processing infor-
mation at the final stages, because they simply are not clearly represented on the In-
ternet, or are not singled out properly.
4.2 Comparison with capabilities of engineering approach
Let’s compare the capabilities of the elinga with those of the “engineering ap-
proach”. Imagine a formal procedure within the framework of our methodology to
find solutions in the “engineering approach”:
1Eng iEng iEng
1
P(W Eng , G, C , I , Q, Z ) ... Pi (Wi Eng , GiEng , CiEng , I iEng , QiEng , ZiEng ) ...
Eng
R (4)
... R(WREng , GREng , CREng , I REng , QREng , Z REng )
Where R(WREng,GREng,CREng,IREng,QREng,ZREng) - is achievable.
However, the initial volume of the problems being solved WEng, is much less then
volume of solvable problems WEl for the elinga, i.e. |WEl|>>|WEng|. In addition, the
final volume of intermediate and resulting concepts GREng, CREng, IREng, QREng, ZREng,
created using the engineering approach is smaller than GREl, CREl, IREl, QREl, ZREl for
the elinga. This is due to the fact that the elinga uses the same capabilities of the engi-
neering approach, but additionally there are other possibilities due to the much wider
set of sources that are fit to be entered into its’ KB. In addition, the elinga, due to the
LI from different sources allows you to link together elements that are practically
impossible for a particular specialist to quickly detect and use, i.e. the capabilities of
systems implementing engineering approach are much less than the capabilities of
elinga.
4.3 Comparison with capabilities of expert systems
In ES, due to the principle of construction, you get to Φ R – final stage of decision
making:
1ES iES iES1
P(W ES , G, C , I , Q, Z ) ... Pi (Wi ES , GiES , CiES , I iES , QiES , ZiES ) ...
ES
R (5)
... R(WRES , GRES , CRES , I RES , QRES , Z RES )
Where R(WRES,GRES,CRES,IRES,QRES,ZRES) - is achievable.
However, the set of the problems being solved by ES WES is much less than a set of
problems solvable with the help of elinga WEl, maybe by tens or even hundreds of
times |WEl|>>|WES|. Thus, the possibilities of elinga are much more serious in terms of
the breadth of the tasks / problems being solved.
4.4 Comparison with capabilities of Watson System
It is interesting to compare the technology of elinga with the latest development of
IBM - IS Watson [6]. For one of the first Watson demos in the competition with hu-
mans, a KB that was used contained results of processing sources in form of about
200 million pages. IS Watson rapidly automatically responded to questions and won
the competition over a person. The IS was built on a very powerful computing com-
plex with 10 servers with a total of 2880 processor cores. The competition demon-
strated the limitation of memory of people in remembering large volumes of specific
information and the speed of its processing. Here is another important point - the
Watson system made mistakes as well during the game. The level shown by Watson
IS in the game, in practice with automatic DM, is unacceptable and in narrow specific
areas people give better results. In the DM process, a specific specialist uses either
special literature to clarify the situation, or connects a different type of computer sys-
tem, or consults with another specialist. This is a form of dialogue between the deci-
sion-maker and something (someone) in an explicit or veiled form, before making a
decision. In elinga, the mechanism of dialogue-associative search for a solution is
built in from the start [19-22], it is used in the process of finding a solution, and not at
the end with formulation of new request, if the solution does not satisfy, as is current-
ly the case for IS Watson. Over the years, IBM has been able to bring Watson IS to an
acceptable DM results for some medical applications where solution stability is val-
ued.
Thus, the Watson IP IS procedure has the form of:
1W Wi Wi1
P(W , G, C , I , Q, Z ) ...
W
Pi (Wi , Gi , Ci , I i , Qi , Zi ) ...
W W W W W W
WR (6)
... R(WRW , GRW , CRW , I RW , QRW , Z RW )
Where R(WRW,GRW,CRW,IRW,QRW,ZRW) - achieved only in the long run.
The fact is that Watson is becoming a hostage to its approach. It makes an attempt
to automatically create a KB with a volume much larger initially than in the elinga,
but it does not use the TDIK mode, which drastically reduces the volume used by KB
of elinga [19-22]. The TDIK functions with the participation of expert editors who
initially solve many linguistic problems in the texts [25] while creating a KB of
elinga, which then sharply accelerates the LI [21]. As a result, IS Watson is forced to
use a much more powerful computing complex than elinga (a laptop is enough), and
developers are forced to debug the resulting solutions for a long time in narrow sub-
ject areas, as it happens in ES [1, 7].
Thus, the elingas’ capabilities are more suitable for mass use in varied subject are-
as, creating relatively quickly necessary knowledge bases. But the elingas still allow
for a wide range of calculations, including intermediate, and find rare or new solu-
tions. The latter is especially problematic for the Watson IS, because in the process of
working with giant knowledge bases, they simply “drown” in the total volume and do
not get a pass from the final statistical models that highlight the most “reliable solu-
tions” based on the highest frequency of their application.
5 Conclusion
The practice of developing and using IS for DM has been demonstrating uncondition-
al success in certain areas and for individual methodologies for decades. For example,
this applies to ES, but it also revealed their limited capabilities. On other hand, the use
of the Internet and the IS Watson, despite the enthusiasm for them, also revealed fun-
damental shortcomings and limitations of the development and impossibility of solv-
ing many simple tasks / problems that arise before the decision-maker, compared with
ISs based on the use of TDIK.
References
1. Russel, S., Norvig, P.: Artificial Intelligence: A Modern Approach. Publishing House
“Williams”, Moscow (2018).
2. Klir, G., Elias, D.: Architecture of Systems Problem Solving. Radio and telecommunica-
tions, Moscow (1990).
3. Zeleny, M.: The IEBM Handbook of Information Technology in Business. Thomson. Lon-
don (2000).
4. Khoroshevsky, V.F. Knowledge Spaces on the Internet and Semantic Web, Part 1. Artifi-
cial Intelligence and Decision Making, vol. 1 issue 1, pp. 80-97, (2008); vol. 2 issue 4, pp.
15-36, (2009); vol. 3 issue 1, pp. 3-38, (2012).
5. IHS Goldfire. Accelerating Decisions. Powering Innovation. White Paper, IHS Inc. (2013),
https://ihsmarkit.com/pdf/IHS-Goldfire-Platform-Whitepaper_140823110915517432.pdf,
last accessed 2019/04/02.
6. High, R.: Epoch of cognitive systems: The principle of construction and operation of IBM
Watson. IBM (2013), http://www.olap.ru/home.asp?artId=2507, last accessed 2019/04/02.
7. Gavrilova, T.A., Kudryavtsev, D.V., Muromtsev, D.I.: Knowledge Engineering. Models
and methods: a textbook. Publishing House “Lan”. St. Petersburg (2016).
8. Eremenko, Yu.I.: Intelligent systems of decision making and management. Publishing
House “TNT”, Staryi Oskol (2017).
9. Ackoff, R., Emery, F.: On Purposeful Systems. Aldine-Atherton Press, vol. 7, Chicago and
New York (1972).
10. Pólya, G.: Mathematical Discovery: On Understanding, Learning and Teaching Problem
Solving. Publishing House “Nauka”, Moscow (1976).
11. Altshuller, G.S.: To Find an Idea. Introduction into the theory of inventive problem solv-
ing. Publishing House “Nauka”, Novosibirsk (1986).
12. Orlov, M.A.: Foundations of classical TIPS. Practical guide for inventive thinking. Solon-
Press, Moscow (2005).
13. Spiridonov, V.F.: Psychology of thought: Solving tasks and problems. Publishing House
“Genesis”, Moscow (2006).
14. Polovinkin, A.I.: Fundamentals of Engineering. Publishing House “Mashinostroeniye”,
Moscow (1988).
15. Pospelov, D.A.: Modeling reasoning. Experiment in the analysis of mental acts. Radio and
telecommunications, Moscow (1989).
16. Larichev, O.I.: Theory and decision-making methods, as well as a chronicle of events in
magical realms. Publishing House “Logos”, Moscow (2002).
17. Mikoni, S.V.: Multicriteria choice on a finite set of alternatives: a tutorial. Publishing
House “Lan”, St. Petersburg (2009).
18. Volkova, V.N., Denisov, A.A.: Theory of systems and systems analysis. Publishing House
“Yuright”, Moscow (2014).
19. Bronfeld, G.B.: Basics of artificial intelligence: a textbook. NNSTU R.E.Alekseeva, Nizh-
ny Novgorod (2014).
20. Bronfeld, G.B.: Direct imposition of knowledge and its capabilities. Analysis, methodolo-
gy, new model of knowledge, algorithms, the possibility of "impossibility". LAP
LAMBERT Academic Publishing, Saarbrucken, Deutschland (2014).
21. Bronfeld, G.B.: Opportunities for a sharp acceleration of logical inference for intelligent
systems with a knowledge base based on the technology of direct imposition of
knowledge. Control systems & Informational Technologies, vol. 1(71), pp. 28-32 (2018).
22. Bronfeld, G.B., Kirov, D.I., Kondratyev, V.V.: The prototype of an intelligent e-book
based on technology of knowledge direct imposition. Programmnye Produkty I Sistemy
(Software & Systems), vol. 3, pp. 403-410 (2019).
23. Brooks, Jr., F.P.: The Design of Design: Essays from a Computer Scientist. Addison-
Wesley, New York (2010).
24. Baars, B., Gage, N.: Cognition, Brain and Consciousness. Introduction to Cognitive Neu-
roscience, 2nd edition. Academic Press (2010).
Nickolaev, I.S. Applied and Computational Linguistics. In: Nickolaev, I.S., Mitrenina,
O.I., Lando, T.M. (eds.). Publishing House “Leland”, Moscow (2016).