human mind modelling

8/9/2019 Human Mind Modelling

1/6

Modeling Human MindTohru Ni t t a , Toshio Taiiaka, Kenji Nishida, Hiroaki Iiiayoshi

Minds Function Wloclels Group, Coiiiputer Science Division, Electrotechnical Laboratory1-1-4 Uniezono, Tsukuba Science City, Ibaraki, 305-8568 .JapanEiiiail: tnittaiQetl.go.jp

ABSTRACTIn this paper, we propose a computational modelof personality (called personali ty model) for thepurpose of implementing non-intellectual fiinc-tions of human mind on computer systems. Thepersonality model will be formulated based onpsychoanalysis, assuming that defensive mecha-nism plays an essential role in a personality. In-ductive probability will be employed for modelingdefense mechanism. The personality model is use-ful for the expression of feelings, and will be usedin virtual reality, computer game characters, agentsecretaries, and robotics.

1. INTRODUCTIONThis paper proposes a computational modelof personality, called personali ty model , which isbased on concepts, principles, and structures ofhuman mind [13.Human mind may be classified into in tel lectual

activities, such a s learning, reasoning, judging, es-timating, memorizing and associating, and non-in tel lectual functions, such as emotion and un-consciousness. Various att emp ts have been madeto implement the intelligence, such as studies onexpert systems and neural networks[3], but non-intellectual activities have scarcely been investi-gated. This paper models non-intellectual func-tions of human mind. The personality modelis formulated based on psychoanalysis, assumingth at defensive mechanism plays a n essential rolein a personality. Inductive probability is employedfor modeling defense mechanism.

2. THE PERSONALITYMODELIn this section, we formulate a computationalmodel of personality based on the knowledge ofpsychoanalysis.

0-7803-5731-0/91k$10.001999 EEE

2.1 Personality in psychoanalysisThis section describes briefly how human person-ality is understood in psychoanalysis. S. Freudproposed that it was possible to analyze person-ality in terms of topography, structure, economy.dynamics and development [4].

In terms of topography, Freud pointed out thatmost human activities were regulated by uncon-sciousness rathe r than conscious volition. Struc-turally, personality consists of ego, super-ego andid , which are associated with reality. morality,and pleasure principles, respectively. Id is asso-ciated with the drives of w a n t t o d o someth ing .and super-ego is associated with the drives of m u s tdo something. Ego adjusts between the drives ofid and those of super-ego. The economic theorystates that the mental process is a result of thequantita tive distribution of energy. Mental activ-ities involve a certain kind of energy, whose quan-tity and distribution determine which mental pro-cess occurs. The dynamic theory states that psy-chological phenomena are the result of conflictsamong ego, super-ego and id. Such conflicts causeanxiety (feeling of the collapse of ego). Ego usesdefensive mechanism to escape from the anxiety.Defense is an adaptive mechanism, which main-tains ego by avoiding danger, anxiety, or dis-pleasure. Th e mechanism functions automaticallywhen the energy of ego drops (or when ego be-comes weak). Defense mechanism involves repres-sion, denial, projection, and reaction formation,among which repression is the most important.Repression withholds from recall and encloses un-der unconsciousness the memories, ideas, feelings.or desires tha t are painful or dangerous to accept.2.2 General framework of the person-

ality modelThe rest of this section formulates a personal-ity as a computational model. The psychoana-lytical structural, economic, and dynamic theo-ries described in Section 2 .1 are used for modelingpersonality. The model focuses on the defensivemechanism as the essentials of personality. Th at

II -342
http://tnittaiqetl.go.jp/http://tnittaiqetl.go.jp/


2/6

M >e,2 .eg>e3 > eS1 ..

External WorldSuper-ego83super-ego

Eg o ( NP (

ego events

\a 2 ). . .

Idesires to attain honorbut does not desire to

p ( a ~ A B l langry but not angry(ambivalent feelings) I . . .attain honorI a 1A '1 - 1 I (ambivalentdesire)a , A a ,I L L I

Figure 1: An image of the framework of the per-sonality model

is, we model the following mechanism:Trigger --+ Conflict+ nxiety +-+ efenseFor exaniple, a person may have a conflicting feel-ing toward a subject he/she sees or hears: I likeit, bu t I hate i t . The person becomes anxiouswhen such a conflict occurs. Usually, the defen-sive mechanism works to avoid anxiety, but thereare cases in which the mechanism does not prop-erly work, anxiety grows, and a vicious cycle isformed. This paper focuses on repression, whichis a typical defensive mechanism.

This section gives the general framework of thepersonality model. An iniage of the framework isshown in Figure 1. Personality is constituted bythree modules: ego, super-ego. and id, for each ofwhich events occur.(1 ) Actualization of super-ego eventsA finite set is given in advance for the eventsof super-ego (hereinafter called super-ego events);S = { S I , . . . , I } . An event of these super-egoevents is actualized according to a given probabil-ity distribution which depends on the person.

clef

(2) Actualization of id eventsSimilarly, a finite set for the events of id ( i deven ts ) is given in advance; U = ( ~ 1 , . .;U,}.An event of these id events is actualized according

def

to a given probability distribution which dependson the person.(3) Ego events

A finite set is given in advance for the eventsof ego (ego even ts ) ; K = { a 1 ; ~ ~ , a p } .Egoevents include feelings, memories, ideas. and de-sires. Feelings. for example. are further classi-fied into fear, anger, sadness, and pleasure. Egoevents are defined by the inductive probability(described later) that is calculated from the an-ticipated events.

de f

(4) Anticipated eventsWe assume tha t a person anticipates in his/hermind what will occur next when he/she receivesinformation of some kind from the external world.A finite set of the anticipated events is given in

advance; E = { e l , . . e n } . The person stochas-tically ant icipates which of these events will occur(we assume that only one event will occur). Here,let 2 be a random variable for expressing the ac-tualized anticipated event, whose probability dis-tribution is as follows:

de f

{P(Z e i ) I i = 1,. . ,n}.This probability distribut ion depends on infornia-tion given and the personality. Here, the orderon strength is given for the set of the anticipatedevents { e l , . . - , e , } in advance. An event witha stronger value has a larger effect on super-egoevents. For example, the order of strength maybe e12 > eg > e3 > e31 > . . - .(5) Law of energy of conservation

Energy is defined for each of these three mod-ules of personality. Th e sum of the energy ofsuper-ego M , the energy of ego N , and the en-ergy of id L is always constant: M +N +L = 1.0,provided tha t M ,N ,L 2 0.(6 ) Conditions for the defensive mechanismto work

Th e energy of ego N determines the intensity ofdefense (for example, repression). A largeN valuesignifies strong ego (reality principle)' and that theperson can solve a conflict by facing reality andnot by using defensive mechanism. Such a per-son neglects the effect of the super-ego event thatcauses a conflict (e.g. sg), and the correspondingid event (e.g. ~ 3 1 )ecomes dominant.On the other hand, a small N value signifiesthat ego does not well function and thus the de-fensive mechanism works. Defense is an adaptivemechanism that works when ego is weak.

II -333


3/6

(7) Rules for determining conditional prob-ability valuesWhen a super-ego event s j and an id event~ 3 1ccur, the conditional probability {q(aj I e i ) }between anticipated events { e l , . . , ,} and egoevents {a 1, . .up} s determined in the followingmanner.An arbitrary a3 ( j= 1 , * . p ) is chosen. a3 istreated as follows.The process bescribed below is performed foranticipated events { e l . . . . e ,} one by one in theorder of strength (from the strongest one to theweakest one). We assume tha t e l > . . > e , with-out losing generality. Let the target anticipatedevent be e,. In this case, the conditional probabil-ity { q (a I e l ) , . .,q(a, I e , - l ) } for the anticipatedevents ( e l , . . , z - . l } is already determined. Theconditional probability q(aj I e , ) is thus stochasti-cally determined according to the following rules.

0

0

2.3

The strengths of super-ego energy M and idenergy L determine which of the super-egoevent s j or the id event ~ 3 1as a strongereffect (the greater the energy is, the strongerthe effect is).The strength of anticipated events (e1 >. . > e,) determines the effect of the super-ego event s5 on e i . This effect is smallerthan the effect of the super-ego event sg on( e l , . . ., i - I ) .

ConflictThis section defines conflict based on the inductiveprobability P ( a n 5 ) of ego event a . Inductiveprobability was proposed by Otsu [2], which wasa natural extension of the usual probability. Thefollowing characteristic, which is not seen in usualprobability, is inherent in inductive probability.Proposition 1

P ( An A )2 . (2 )Tha t i s , law of contradict ion does not hold . 0

Equation (2 ) shows that the probability of Aand not A can,be a positive value.We can easily find that inductive probability

has the following characteristics:Proposition 2 The induct ive probabil i ty P ( AnA ) kfE,[q(An A I b,)] take s a value within therange of

0 5 P ( A n A )5 -. (3)14

T h e m a x i m u m v a l u e 1 /4 is real ized when q(A Ib,) = 1 /2 f o r a n y n E N , an,d t h e m i n i m u m v a l u e0 is realized w he n q(A I b,) = 0 OT q(A I b,) = 1f o r a n y n E N where N is th,e set of all naturalnumbers .

We then define conflict using the inductiveprobability.Definition 1 (conflict) Using the in ,duct iveprobabil i ty , for any ego event %, the degree ofconfl ict is def ined as

The la rger the va lue of equation (4 ) is7 thestronger the confl ict is . A s shown in Proposi t ion2 , the va lues of equation (4 ) ar e 0 OT larger b,utnever la rger tha n 114. W h e n th i s v a lu e i s 0 , therei s n o c o n fl i ct , a n d t h e c o n fl ic t i s m a x i m u m w h e nthe va lue i s 114. 02.4 AnxietyIn this section, anxiety is formulated.Definition 2 (anxiety) L e t a be a real numbersuch tha t 0 5 CY 5 1/4, which is cal led anxietyparameter . Th en feel ing anxious is def ined as as ta te whe n there ex i s t s an ego even t a E K whichsatisfies

P(un a ) 2 C Y . ( 5 )Th e degree of anxiety UE(a ) s defined by

where K = { a ~ , ~ ~ ~ , a p }s a f in i t e se t of egoeven ts , IKI is the cardinari ty of K , and 4/IKI i sa coe f fi c ien t f o r normal i za tion . 02.5 RepressionRepression does not work immediately after theperson feels anxious. If the anxiety is not severe,the person can support himself/herself even un-der an anxious state . However, if the anxiety isvery strong, the person cannot support i t , and re-pression works to defend himself/herself. We callthe anxiety limit over which repression works cri t -ical anxiety. This section formulates critical anxi-ety a nd repression using the degree of anxiety UE(equation (6)).

II -3344


4/6

Definition 3 (critical anxiety) L e t P be a realnumbe r such tha t 0 5 P 5 1, which is called crit-ical anxiety parameter . The n, cri t ical anxiety isdef ined as a s ta te when(7)0

Definition 4 (repression) Repression occursby the mechanism described below when the per-son confronts th,e critical anxiety.(1)Ego produces a n art i f icia l ant icipated event 61which satisfies

ord a el) = 1 (9)

for any ego event a , E K s u ch t h at P ( a n 3 )2(art i f icia l anticipated events are expressed withon let ters) . Tha t is , an arti f icia l ant icipated event61 i s added to the se t of anticipated events E =

T h e n , t h e n u m b e r of anticipated events in-creases f ro m n t o n+ 1, und t he probability distri-bution expressed by equation ( l ) ,which is{ e l , . * I e n > .

{ P ( Z = e l ) , ..., P ( Z = e , ) } (10)changes in to(yp(Z = e l ) , - - . yP (Z = e n ) , P ( Z = 6 1 ) = I-y}.Here, the real number 0 5 y < 1 s called repres-s ion parameter .

(11)

Arti f icia l ant icipated events 62 . 63 , . . are suc-cessively produced unti l the person escapes fr omthe cr i t ical anxiety, in other words, UE(cy)< p .( 2 ) For each artificial anticipated event & , a self-acceptance degree ADI , is produced. Th e anxietyparameter a and the cr i t ical anxiety parameter Pare updated based on the sum of self-acceptancedegrees:

AD Gf ADk. (12 )k

0

Th e following two theorems can be easily proved(see [l] or the proofs).Theorem 1 T h e c o nf li c t P ( a n ii) strictlymonotonously decreases as artificial anticipated

0vents occur accordirzg t o Defin ition 4 (1).

Theorem 2 Ei t h e r P(u) or P(7L) strictlymonotonously decreases , and the other s tr ict lymonotonously increases as artificial anticipatedevents occur according t o Defini t ion 4 (1). 0

Definition 4 (1) signifies that a person artifi-cially produces an anticipated event 61 for eas-ing the anxiety caused by ego events satisfyingP ( a n a ) 2 , and reduces the conflict P(an 7 i )(see Theorem 1). The person attempts to avoidthe critical anxiety by reducing P(una ) below a.The rnechanisni of Definition 4 s called repres-s ion due to the following reason. By Theorem 2,an artificial anticipated event 61 causes a drop ineither P ( a ) or P(zi) and an increase of the other.Therefore, one of the two is suppressed. This isrepression.Definition 4 ( 2 ) assumes that the self-

acceptance degree ADk takes a small negativevalue and that the resulting changes in anxietyparameter a and critical anxiety parameter ,8 arealso small negative values. In other words, cy andP decrease. When CY decreases, the number of egoevents that satisfy P(ana)2 y increases, thu s theanxiety degree UE a )becomes larger (generally,UE ( a )7 CY \ 0)). A s a result, repression eas-ily occurs. When the critical anxiety parameterdrops further, repression occurs more easily. Re-pression takes place in order to ease anxiety, andthe conflict seems to disappear. However, anxi-ety actually increases, and repression further ad-vances (in other words, the person becomes un-stable).

Definition 5 (behavior of a person whenhe/she is not at the critical anxiety) W h e na pers on does not feel th e cr i t ical anxiety, t he fol-lowing me chan ism works:Self-acceptance degree AD is produced,which changes the anxiety parameter CYand the cr i t ical anxiety parameter P .

0

Definition 5 assumes that the self-acceptancedegree is positive value (i.e., AD > 0) and thatchanges in a and ,B are small and positive values.In other words, Q and P increase. This signifiestha t the person accepts himself/herself bacause re-pression does not take place, and the possibilitytha t the person feels anxious, suffers critical anx-iety, and uses repression, is somewhat reduced.

n -345


5/6


6/6

a result of repression.(11) Changes in parameters of various kindsAnxiety parameter a = 0.1 and critical anxi-ety parameter p = 0.7 change according to thesum of self-acceptance degrees AD = &ADk.Since AD1 = -0.01 is the only self-acceptancedegree produced in this case study, AD = -0.01.Assume that, the magnitude of change in crit-ical anxiety parameter ( A p ) is determined bya hyperbolic tangent tanh and that the mag-nitude of change in anxiety parameter ( A a ) s

dLfdetermined by (1 / 4 ) tanh. Tha t is. Aa -(1 /4) tanli(AD),A@ Lf tanh(AD). Since AD =-0.01, Aa = -0.0025 and A/3 = -0.01, whichlead a = 0.0975 and /3 = 0.69. Such changes inparameters cause repression to occur slightly moreeasily.

4. CONCLUSIONSThis paper proposed a computational modelcalled personali ty model based on the psychoan-alytical knowledge? whose purpose was to im-plement the non-intellec ual activities of humanmind on computer systems. Inductive probabil-ity played an important role in the formulation ofthe personality model. In the future, we will eval-uate? analyze and improve the personality modelvia computer simulations.

AcknowledgementsThe authors would like to express their thanks toDr.K.Ohmaki, Director of the Computer ScienceDivision, for having an opportunity to do thisstudy and his continual encouragement.References[l]Nitta. T. (1999) A Computational Model ofPersonality, ETL Technical Report , TR-99-5,p.19.[2] Otsu, N. (1989) Toward Soft Logic for theFoundation of Flexible Information Process-ing, Bu l l e t i n of the Electrotechnical Labora-tory. Vo1.53, No.10, pp.75-95.[3] Russell, S.J., Norvig, P. (1995) Arti f icia l In-telligence - A Modern Approach, , Prentice-Hall Inc.[4]Tyson, A . , Strachey, J. (1956)A Chronological Hand-List of Freuds Works,In t . J. P s y c h o - A n d . . 37 . 1-2.

IT -347

human mind modelling

Documents