00100	.SEC EXPLANATIONS AND MODELS
00200	.SS The Nature of Explanation
00300		It is perhaps as difficult to explain explanation  itself  as
00400	it  is  to  explain  anything  else.     The explanatory practices of
00500	different sciences differ widely but they all share  the  purpose  of
00600	someone   attempting   to   answer   someone   else's  (or  his  own)
00700	why-how-what-etc. questions about a situation, event, episode, object
00800	or phenomenon. Thus explanation implies a dialogue whose participants
00900	share some interests, beliefs, and values.    A consensus must  exist
01000	about   admissable  and  appropriate  questions  and  answers.    The
01100	participants must agree on what is a sound  and  reasonable  question
01200	and  what is a relevant, intelligible, and (believed) correct answer.
01300	The explainer tries to satisfy a  questioner's  curiosity  by  making
01400	comprehensible  why  something  is the way it is. The answer may be a
01500	definition,  an  example,  a  synonym,   a   story,   a   theory,   a
01600	model-description,  etc.  The answer attempts to satisfy curiosity by
01700	settling belief.
01800	.V
01900	    Suppose a man dies and a questioner (Q) asks an explainer (E):
02000	.END CONTINUE
02100	       Q: Why did the man die?
02200	One answer might be:
02300	.V
02400	       E: Because he took cyanide.
02500	.END CONTINUE
02600	This explanation might be sufficient to satisfy Q's curiosity and he
02700	stops asking further questions. Or he might continue:
02800	.V
02900	       Q: Why did the cyanide kill him?
03000	.END CONTINUE
03100	and E replies:
03200	.V
03300	      E: Anyone who ingests cyanide dies.
03400	.END CONTINUE
03500	This explanation appeals to a universal generalization under which is
03600	subsumed  the  particular  fact  of  this  man's  death.  Subsumptive
03700	explanations  satisfy  some  questioners  but  not  others  who,  for
03800	example,  might  want  to  know  about  the  physiological mechanisms
03900	involved.
04000	.V
04100	       Q: How does cyanide work in causing death?
04200	       E: It stops respiration so the person dies from lack of oxygen.
04300	.END CONTINUE
04400		If Q has biochemical interests he might inquire further:
04500	.V
04600		Q:What is  cyanide's  mechanism  of  drug  action  on  the 
04700		  respiratory center?
04800	center?
04900	.END CONTINUE
05000		The last two questions refers to causes. When human action is
05100	to  be  explained,  confusion  easily  arises  between  appealing  to
05200	physical,  mechanical causes and appealing to symbolic-level reasons,
05300	that is, learned, acquired procedures or strategies (Toulmin, 1971).
05400		It  is  established  clinical knowledge that the phenomena of
05500	the paranoid mode can be found associated with a variety of  physical
05600	disorders.    For example, paranoid thinking can be found in patients
05700	with  head   injuries,   hyperthyroidism,   hypothyroidism,   uremia,
05800	pernicious   anemia,   cerebral  arteriosclerosis,  congestive  heart
05900	failure, malaria and epilepsy.      Also drug  intoxications  due  to
06000	alcohol,  amphetamines,  marihuana  and LSD can be accompanied by the
06100	paranoid mode. In these cases the paranoid mode is not a  first-order
06200	disease but a way of processing information in reaction to some other
06300	underlying disorder. To  account  for  the  association  of  paranoid
06400	thought  with  these  physical  states  of  illness,  a psychological
06500	theorist might be tempted to hypothesize that a  purposive  cognitive
06600	system   would  attempt  to  explain  a  physical  illness  state  by
06700	constructing persecutory beliefs blaming other human agents  for  the
06800	ill-being of the disease state. But before making such an explanatory
06900	move, we must consider the elusive distinction  between  reasons  and
07000	causes in explanations of human behavior.
07100		One view  of  the  association  of  the  paranoid  mode  with
07200	physical  disorders  might be that the physical illness simply causes
07300	the paranoia ,through some unknown mechanism, at a  "hardware"  level
07400	beyond the influence of deliberate reprogramming and beyond voluntary
07500	self-control.   That  is,  the  resultant  paranoid  mode  represents
07600	something  that  happens to a person as victim, not something that he
07700	does as an active agent. Another view is that the paranoid  mode  can
07800	be  explained  in  terms of reasons, justifications which describe an
07900	agent's  intentions  and  beliefs.    Does  a  person  as  an   agent
08000	recognize,  monitor  and control what he is doing or trying to do? Or
08100	does  it  just  happen  to  him   automatically   without   conscious
08200	deliberation?   This  question  raises  a  third  view,  namely  that
08300	unrecognized  (but  potentially recognizable) reasons, aspects of the
08400	program which are sealed off and inacessible  to  voluntary  control,
08500	can function like causes.  Once brought to consciousness such reasons
08600	can  be  modified  voluntarily  by  the agent, as a language user, by
08700	reflexively talking to and  instructing  himself.  This  second-order
08800	monitoring  and  control  through  language contrasts with an agent's
08900	inability  to  modify  causes  which  lie  beyond  the  influence  of
09000	self-criticism  and  change  through internal linguistically mediated
09100	argumentation.   Timeworn conundrums  about  concepts  of  free-will,
09200	determinism,  responsibility,  consciousness and the powers of mental
09300	action here plague us unless we stick closely to a  computer  analogy
09400	which  makes  a  clear  and  useful  distinction  between  levels  of
09500	hardware, interpreter and programs. (See p. 000 in Chap 2)
09600	
09700		Each  of these three views provides a serviceable perspective
09800	depending on how a disorder is to be explained and corrected.    When
09900	paranoid  processes  occur during amphetamine intoxication they might
10000	be viewed as biochemically caused and beyond the patient's ability to
10100	control  volitionally through internal self-correcting dialogues with
10200	himself.  When a paranoid moment occurs in a normal person, it can be
10300	viewed  as  having a mistinterpretation as a reason.  If the paranoid
10400	misinterpretation is recognized as such, a person has  the  power  to
10500	revise  or reject it through internal debate.  Between these extremes
10600	of drug-induced paranoid processes and the self-correctible  paranoid
10700	moments  of  the  normal  person, lie cases of paranoid personalities
10800	paranoid reactions and the paranoid mode associated  with  the  major
10900	psychoses (schizophrenic and manic-depressive).
11000		One opinion has it that the major psychoses are a consequence
11100	of  unknown  physical  "hardware"  causes  and  are beyond deliberate
11200	voluntary control.  But  what  are  we  to  conclude  about  paranoid
11300	personalities  and  paranoid  reactions where no hardware disorder is
11400	detectable or suspected? Are such persons to be  considered  patients
11500	to  whom something is mechanically happening or are they agents whose
11600	behavior is a consequence of what they do?   Or are they  both  agent
11700	and  patient  depending on on how one views the self-modifiability of
11800	their symbolic processing?  In these perplexing cases we  shall  take
11900	the  position that in normal, neurotic and characterological paranoid
12000	modes, the psychopathlogy represents something that happens to a  man
12100	as  a  consequence  of  what  he  has  experientially  undergone,  of
12200	something he now does, and something he now undergoes.    Thus he  is
12300	both  agent and victim whose symbolic processes have powers to do and
12400	liabilities to undergo.    His liabilities are reflexive in  that  he
12500	is victim to, and can succumb to, his own symbolic structures.
12600	
12700		From this standpoint I  would  postulate  a  duality  between
12800	reasons  and causes. That is, a reason can operate as an unrecognized
12900	cause in one context and be offered as a recognized justification  in
13000	another.   It  is, of course, not the reason itself which serves as a
13100	cause  but  having   the   reason.   Human   symbolic   behavior   is
13200	non-determinate  to  the extent that it is self-determinate. Thus the
13300	power to make some decisions freely  and  to  change  one's  mind  is
13400	non-illusory.    When  a  reason is recognized to function as a cause
13500	and is accessible to self-monitoring, it may be changed or  rejected.
13600	In  this sense a two-levelled system involving an interpreter and its
13700	programs is self- changeable and self-correcting, within limits.
13800	.END
13900		Explanations  both  in  terms  of  causes  and reasons can be
14000	indefinitely extended and endless questions  can  be  asked  at  each
14100	level of analysis.  Just as the participants in explanatory dialogues
14200	decide what is taken to be problematic, so they  also  determine  the
14300	termini   of   questions   and  answers.   Each  discipline  has  its
14400	characteristic stopping points and boundaries.
14500		In  the  background  of  explanatory dialogues are larger and
14600	smaller constellations of concepts which are  taken  for  granted  as
14700	nonproblematic  background.   Hence  in considering the strategies of
14800	the paranoid mode `it goes without  saying',  that  is,  transcending
14900	this  particular  mode  of  functioning  is  the fact that any living
15000	teleonomic  system  ,as  the  larger  constellation  ,  strives   for
15100	maintenance  and  expansion of life. Also it should go without saying
15200	that, at a lower level, ion transport takes place through  nerve-cell
15300	membranes.  Every function of an organism can be viewed a governing a
15400	subfunction beneath and depending  on  a  transfunction  above  which
15500	calls it into play for a purpose.
15600		Just as there are many alternative ways of describing,  there
15700	are many alternative ways of explaining.  An explanation is geared to
15800	some  level  of  what  the  dialogue  participants  take  to  be  the
15900	fundamental  structures  and processes under consideration.  Since in
16000	psychiatry   we   cope   with   patients'   problems   using   mainly
16100	symbolic-conceptual  techniques,(although  it  is true that the pill,
16200	the knife, and electricity are also available.), we are interested in
16300	aspects  of  human  conduct  which  can be explained, understood, and
16400	modified at a symbol-processing level. Psychiatrists need theoretical
16500	symbolic   systems  from  which  their  clinical  experience  can  be
16600	logically derived to interpret the case histories of their  patients.
16700	Otherwise  they  are  faced  with mountains of dross and indigestible
16800	data.   "Science is an attempt to make the chaotic diversity  of  our
16900	sense  experience correspond to a logically uniform system of thought
17000	by correlating single experiences  with  the  theoretic  structure."-
17100	Einstein.
17200	.SS The Symbol Processing Viewpoint
17300	
17400		Segments  and  sequences  of  human behavior can be looked at
17500	from many standpoints.   In this monograph I shall view sequences  of
17600	paranoid symbolic behavior from an information processing standpoint.
17700	For  a  more  complete  explication   and   justification   of   this
17800	symbol-processing view, see Newell (1973) and Newell and Simon (1972).
17900		In brief, information is defined as knowledge in  a  symbolic
18000	code.    A symbolic process is a symbol-manipulating activity posited
18100	to account  for  observable  symbolic  behavior  such  as  linguistic
18200	interaction.  Symbols  are  defined  as representations of experience
18300	classified as objects, events, situations, and  relations.
18400		Symbol-processing  explanations   postulate   an   underlying
18500	structure   of  hypothetical  processes,  functions,  strategies,  or
18600	directed symbol-manipulating procedures, having the power to  produce
18700	and  being  responsible  for the manifest phenomena. Such a structure
18800	offers  an  ethogenic  (ethos  =  conduct  or  character,   genic   =
18900	generating)   explanation  for  sequences  or  segments  of  symbolic
19000	behavior.  In  adopting  an  ethogenic  viewpoint,  I   shall   posit
19100	processes,  functions, procedures and strategies as being responsible
19200	for and having the  power  to  generate  the  symbolic  patterns  and
19300	sequences  characteristic  of  the  paranoid  mode.   "Strategies" is
19400	perhaps the best general term since it implies ways of  obtaining  an
19500	objective  which have suppleness and pliability since their choice of
19600	application depends on circumstances.     However  I  shall  use  all
19700	these terms interchangeably.
19800	
19900	.SS Symbolic Models
20000		Theories and  models  share  many  functions  and  are  often
20100	considered  equivalent.  One  important  distinction lies in the fact
20200	that a theory states a subject has a certain structure but  does  not
20300	exhibit  that  structure in itself. (See Kaplan,1964). In the case of
20400	interactive simulation models,  such  as  will  be  described,  there
20500	exists  a  further  distinction.      Interactive  simulation models,
20600	having the ability to converse in natural language  using  teletypes,
20700	actualize or realize a theory in the form of a dialogue algorithm. In
20800	contrast to a verbal, pictorial or mathematical representation,  such
20900	a model changes its states over time and ends up in a state different
21000	from its initial state.
21100		In  contrasting  description from what is described, Einstein
21200	remarked  that it is not the function of science to give the taste of
21300	the soup.  But an interactive simulation  model  which  reproduces  a
21400	segment  of  reality does just that, since it offers an interviewer a
21500	first-hand  experience  with  a  concrete  case.  In  constructing  a
21600	computer  simulation  a theory is modelled to discover a sufficiently
21700	rich structure of assumptions to  generate  the  observable  behavior
21800	under  study.    A  dialogue algorithm allows an observer to interact
21900	with a concrete specimen of a class in detail. In  the  case  of  our
22000	model,  the  level of detail is the level of the symbolic behavior of
22100	conversational language which is satisfying to a  clinician  who  can
22200	compare  the  model  with human counterparts at his familiar level of
22300	clinical dialogue. Communicating with the paranoid model by means  of
22400	teletype, an interviewer can directly experience for himself the type
22500	of impaired  social  relationship  which  develops  with  someone  in
22600	paranoid mode.
22700		An algorithm composed of  symbolic  computational  procedures
22800	converts  input  symbolic  structures into output symbolic structures
22900	according to certain principles.  The modus operandi  of  a  symbolic
23000	model  is simply the workings of an algorithm when run on a computer.
23100	At this level of explanation, to answer `why?' means  to  provide  an
23200	algorithm  which  makes explicit how symbolic structures collaborate,
23300	interplay and interlock  -  in  short,  how  they  are  organized  to
23400	generate patterns of manifest phenomena.
23500	
23600		To  simulate the sequential input-output behavior of a system
23700	using symbolic  computational  procedures,  we  write  an  alogorithm
23800	which,  when run on a computer, produces symbolic behavior resembling
23900	that  of  the  subject  system  being  simulated.   (Colby,1973)  The
24000	resemblance  is  achieved  through  the  workings of an inner posited
24100	structure  in  the  form  of  an  algorithm,   an   organization   of
24200	symbol-manipulating   procedures   which   are  responsible  for  the
24300	characteristic observable behavior at the input-output level.   Since
24400	we  do not know the structure of the `real' simulative processes used
24500	by the mind-brain,  our  posited  structure  stands  as  an  imagined
24600	theoretical  analogue,  a  possible  and  plausible  organization  of
24700	processes analogous to  the  unknown  processes  and  serving  as  an
24800	attempt  to  explain  the  workings  of  the  system  under study.  A
24900	simulation model is thus deeper than  a  pure  black-box  explanation
25000	because  it  postulates  functionally equivalent processes inside the
25100	box to account for observable patterns  of  behavior.   A  simulation
25200	model  constitutes  an  interpretive  explanation  in  that  it makes
25300	intelligible the connections between external input, internal  states
25400	and  output  by  positing  intervening  symbol-processing  procedures
25500	operating  between  symbolic  input  and   symbolic   output.      An
25600	intelligible  description  of the model should make clear why and how
25700	it reacts as it does under various circumstances.
25800		Citing a universal generalization to explain an  individual's
25900	behavior  is unsatisfactory to a questioner who is interested in what
26000	powers and liabilities are latent behind manifest phenomena.  To  say
26100	`x is nasty because x is paranoid and all paranoids are nasty' may be
26200	relevant, intelligible and correct. But another type  of  explanation
26300	is  possible,  a model-explanation referring to a structure which can
26400	account for `nasty' behavior as a consequence of input  and  internal
26500	states  of  a  system.   A  model  explanation  specifies  particular
26600	antecedants  and  processes  through  which  antecedants generate the
26700	phenomena.   An ethogenic approach to explanation assumes perceptible
26800	phenomena  display the regularities and nonrandom irregularities they
26900	do  because  of  the  nature  of  a  imperceptible  and  inaccessible
27000	underlying  structure.   The  posited  theoretical  structure  is  an
27100	idealization, unobservable in human heads,  not  because  it  is  too
27200	small, but because it is imaginary.
27300		When attempts are made to explain human behavior,  principles
27400	in  addition  to  those accounting for the natural order are invoked.
27500	"Nature entertains no opinions about us", said Nietzsche,  but  human
27600	natures  do  ,  and  therein  lies  a  source  of  complexity for the
27700	understanding of human conduct. Until the first quarter of  the  20th
27800	century,  natural sciences have been guided by the Newtonian ideal of
27900	perfect process knowledge about inanimate objects whose behavior  can
28000	be  subsumed  under lawlike generalizations.  When a deviation from a
28100	law  was  noticed,it  was  the  law  which  was  modified,  since  by
28200	definition physical objects do not have the power to break laws. When
28300	the planet Mercury was observed to deviate from the  orbit  predicted
28400	by  Newtonian  theory,  no  one  accused  the  planet  of  being   an
28500	intentional agent breaking the law; something was incorrect about the
28600	theory.    Subsumptive explanation is the acceptable norm in  physics
28700	but  it  is  seldom  satisfactory  in  accounting for the behavior of
28800	living purposive systems.     In considering the behavior  of  bodies
28900	falling   in  a  macroscopic  world,  no  one  nowadays  follows  the
29000	Aristotelian pattern of attributing to them intentions to fall .  But
29100	in  the  case  of  living  systems,  especially  ourselves, our ideal
29200	explanatory practice remains Aristotelian in utilizing a  concept  of
29300	intention.  Aristotle's misconception in physics was to extend to the
29400	macroscopic non-living world an intentionalistic concept  of  purpose
29500	appropriate  to  the  living world as a principle of intelligibility.
29600	(See Ayala,1972).
29700		Consider  a  man participating in a high-diving contest.   In
29800	falling towards the water he accelerates at the rate of 32  feet  per
29900	second. Viewing the man simply as a falling body, we explain his rate
30000	of fall by appealing to a physical law.  Viewing the man as  a  human
30100	intentionalistic  agent,  we  explain  his  dive  as the result of an
30200	intention to dive in a cetain way in order to win the diving contest.
30300	His  conduct  (in  contrast  to  mere  movement) involves an intended
30400	following of certain conventional rules for what is judged by  humans
30500	to  constitute, say, a swan dive. Suppose part way down he chooses to
30600	change his position in mid-air and enter the water thumbing his  nose
30700	at  the  judges. He cannot break the law of falling bodies but he can
30800	break the  rules  of  diving  and  make  a  gesture  which  expresses
30900	disrespect  and  which he believes will be interpreted as such by the
31000	onlookers.   Our diver breaks a rule for diving but  follows  another
31100	rule  which  prescribes  gestural action for insulting behavior.   To
31200	explain the actions of diving and nose-thumbing, we would appeal, not
31300	to  laws  of natural order, but to an additional order, to principles
31400	of human order, superimposed on laws of natural order and which  take
31500	into account (1)standards of appropriate action in certain situations
31600	and (2) the agent's inner considerations  of  intention,  belief  and
31700	value  which he finds compelling from his point of view. In this type
31800	of explanation the explanandum, that which is being explained is  the
31900	agent's  informed  actions,  not  simply his movements.  When a human
32000	agent performs an action in a situation, we can ask:  is  the  action
32100	appropriate  to  that situation and if not, why did the agent believe
32200	his action to be called for.
32300		As  will  be  shown,  symbol-processing  explanations rely on
32400	concepts of intention, belief, action, affect, etc. These  terms  are
32500	close to the terms of ordinary language as is characteristic of early
32600	stages of explanations. It is also important to note that such  terms
32700	are  commonly utilized in describing computer algorithms which strive
32800	to achieve goals.  In  an  algorithm  these  ordinary  terms  can  be
32900	explicitly defined and represented.
33000		Psychiatry deals with the practical concerns of inappropriate
33100	action, belief, etc. on the part of a patient. His  behavior  may  be
33200	inappropriate  to  the  onlooker since it represents a lapse from the
33300	expected, a contravention of the human order. It may even appear this
33400	way  to  the  patient  in  monitoring  and  directing  himself.   But
33500	sometimes, as in severe cases of  the paranoid  mode,  the  patient's
33600	behavior  does  not  appear  anomalous to himself.  He maintains that
33700	anyone  who  understands  his  point  of  view,  who   conceptualizes
33800	situations  as  he  does  from the inside, would consider his outward
33900	behavior appropriate and justified. What he does  not  understand  or
34000	accept is that his inner conceptualization is mistaken and represents
34100	a misinterpretation of the events of his experience.
34200		The  model  to  be  presented  in  the  sequel constitutes an
34300	attempt to explain some regularities and  particular  occurrences  of
34400	symbolic   (conversational)   paranoid  behavior  observable  in  the
34500	clinical situation of a psychiatric interview.   The  explanation  is
34600	at the symbol-processing level of linguistically communicating agents
34700	and  is  cast  in  the  form  of  a  dialogue  algorithm.  Like   all
34800	explanations  it  is  incomplete  and does not claim to represent the
34900	only conceivable structure of processes .
35000	.SS The Nature of Algorithms
35100	
35200		Theories can be presented in various forms  such  as  essays,
35300	mathematical   equations   and   computer   programs.  To  date  most
35400	theoretical explanations in psychiatry and psychology have  consisted
35500	of  natural  language  essays with all their well-known vagueness and
35600	ambiguities. Many of these formulations  have  been  untestable,  not
35700	because relevant observations were lacking but because it was unclear
35800	what the essay was really saying. Clarity is needed.
35900		An  alternative  way of formulating psychological theories is
36000	now available in the form of symbol-processing  algorithms,  computer
36100	programs,  which have the virtue of being clear and explicit in their
36200	articulation and which can be run on  a  computer  to  test  internal
36300	consistency and external correspondence with the data of observation.
36400	The subject of a model is what it is a model  of;  the  source  of  a
36500	model  is  what  it  is  based  upon. Since we do not know the `real'
36600	mind-brain algorithms, we construct a theoretical model,  bades  upon
36700	computer    algorithms,   which   represents   a   partial   analogy.
36800	(Harre,1970). (Harre, 1970).  The analogy  is  made  at  the  symbol-
36900	processing   level,   not  at  the  hardware  level.   A  functional,
37000	computational or procedural equivalence  is  being  postulated.   The
37100	question   then  becomes  one  of  categorizing  the  extent  of  the
37200	equivalence.      A   beginning   (first-approximation)    functional
37300	equivalence  might be defined as indistinguishability at the level of
37400	observable I-O pairs.    A  stronger  equivalence  would  consist  of
37500	indistinguishability  at  inner  I-O  levels. That is, there exists a
37600	correspondence between what is being done and how it is being done at
37700	a given level of operations.
37800		An algorithm represents an organization of  symbol-processing
37900	strategies  or functions which represent an `effective procedure'. It
38000	is  essential  to  grasp  this  fundamental   concept   of   computer
38100	simulation. An effective procedure consists of two compoments:
38200	.V
38300		(1) A programming language in which procedural rules of
38400		    behavior can be rigorously and unambiguously specified.
38500	
38600		(2) A machine processor which can rapidly and reliably carry
38700		    out the processes specified by the procedural rules.
38800	.END
38900	The specifications of (1), written in a formally defined  programming
39000	language,  is  termed  an  algorithm  or program while (2) involves a
39100	computer as the machine processor, a set  of  deterministic  physical
39200	mechanisms   which  can  perform  the  operations  specified  in  the
39300	algorithm. The algorithm is called `effective'  because  it  actually
39400	works, performing as intended when run on the machine processor.
39500		A simulation model is composed of procedures analogous to the
39600	real   and  unknown  procedures.    We  are  not  claiming  they  ARE
39700	analogous, we are MAKING them so. The analogy  being  drawn  here  is
39800	between specified processes and their generating systems. Thus
39900	
40000	.V
40100	      mental process    computational process
40200	      --------------:: ----------------------
40300	      brain hardware      computer hardware and
40400	      and programs           programs
40500	.END
40600	
40700	Many of the classiclal mind-brain problems arose  because  there  was
40800	nothing else in the world at the time to serve as a familiar analogy.
40900	With  the  advent  of  computers   and   programs   some   mind-brain
41000	perplexities  disappear.  (Colby,1973).  The  analogy  is  not simply
41100	between computer hardware and brain wetware.  We  are  not  comparing
41200	the  structure  of neurons with the structure of transisitors; we are
41300	comparing the organization  of  symbol-processing  procedures  in  an
41400	algorithm  with  symbol-processing  procedures of the mind-brain. The
41500	central nervous system contains a representation of the experience of
41600	its  holder.  A model builder has a conceptual representation of that
41700	representation which he demonstrates in the form of a model. Thus the
41800	model is a demonstration of a representation of a representation.
41900		Since we are taking  running computer programs as a source of analogy for  a
42000	paranoid model,  errors or pathological behavior on the part of such  programs 
42100	are of  interest  to  the  psychopathologist.  These  errors  can  be
42200	ascribed to the hardware level, to the interpreter or to the programs
42300	which the interpreter executes. Different remedies  are  required  at
42400	different levels. If the analogy is to be useful in the case of human
42500	pathological behavior, it becomes a matter  of  influencing  symbolic
42600	behavior with the appropriate techniques.
42700		Since the algoritm is written in a programming  language,  it
42800	is  hermetic  except  to  a  few  people, who in general do not enjoy
42900	reading  other  people's  code.    Hence  the   intelligibility   and
43000	scrutability  requirement for explanations must be met in other ways.
43100	In an attempt to open the model to  scrutiny  I  shall  describe  the
43200	model in detail using diagrams and interview examples profusely.
43300	
43400	
43500	.SS Analogy
43600		I  have  stated  that  an  interactive  simulation  model  of
43700	symbol-manipulating   processes   reproduces  sequences  of  symbolic
43800	behavior at the level of linguistic communication.  The  reproduction
43900	is  achieved  through the operations of an algorithm which represents
44000	an  organization  of  hypothetical  symbol-processing  strategies  or
44100	procedures  which  can  generate  the  I-O  behavior  of the subject-
44200	processes under  investigation.The  algorithm  is  be  an  "effective
44300	procedure" in the sense it really works in the manner intended by the
44400	model-builders. In the model to be described, the paranoid  algorithm
44500	generates   linguistic   I/O   behavior  typical  of  patients  whose
44600	symbol-processing is dominated by the paranoid mode. Comparisons  can
44700	be  made  between samples of the I/O behaviors of patients and model.
44800	But the analogy is not to be drawn at this level.   Mynah  birds  and
44900	tape  recorders  also  reproduce human linguistic behavior but no one
45000	believes the reproduction is achieved by powers  analogous  to  human
45100	powers.   Given that the manifest outermost I/O behavior of the model
45200	is indistinguishable  from  the  manifest  outward  I/O  behavior  of
45300	paranoid  patients,  does this imply that the hypothetical underlying
45400	processes used by the model are analogous  to  or  the  same  as  the
45500	underlying processes used by persons in the paranoid mode?  This deep
45600	and far-reaching question should be approached with caution and  only
45700	when  we  are  first  armed  with  some  clear notions about analogy,
45800	similarity,   faithful   reproduction,    indistinguishability    and
45900	functional equivalence.
46000		In comparing two things (objects, systems or processes )  one
46100	can   cite   properties   they  have  in  common,(positive  analogy),
46200	properties they do not share (negative analogy) and properties  which
46300	we  do  not  yet  know whether they are positive or negative (neutral
46400	analogy). (See Hesse,1966). No two things are exactly alike in  every
46500	detail.   If  they  were identical in respect to all their properties
46600	then they would be copies. If they were identical  in  every  respect
46700	including  their  spatio-temporal  location we would say we have only
46800	one thing instead of two. Everything  resembles  something  else  and
46900	maybe everything else, depending upon how one cites properties.
47000		In an analogy a similarity relation is  evoked.  "Newton  did
47100	not  show  the  cause of the apple falling but he showed a similitude
47200	between the apple and the stars."(D`Arcy Thompson). Huygens suggested
47300	an analogy between sound waves and light waves in order to understand
47400	something less well-understood (light) in terms of  something  better
47500	understood   (sound).   To  account  for  species  variation,  Darwin
47600	postulated a  process  of  natural  selection.    He  constructed  an
47700	analogy  from two sources, one from artificial selection as practiced
47800	by domestic breeders of animals and one from  Malthus'  theory  of  a
47900	competition  for  existence  in a population increasing geometrically
48000	while its resources increase arithmetically. Bohr's model of the atom
48100	offered  an  analogy  between solar system and atom. These well-known
48200	historical examples should be sufficient here to illustrate the  role
48300	of analogies in theory construction.    Analogies are made in respect
48400	to  those  properties  which  constitute  the  positive  and  neutral
48500	analogy.     The  negative analogy is ignored.   Thus Bohr's model of
48600	the atom as a miniature planetary system was not intended to  suggest
48700	that  electrons  possessed  color or that planets jumped out of their
48800	orbits. .SS Functional Equivalence
48900		When human symbolic processes are the subject of a simulation
49000	model, we draw from two sources, symbolic computation and psychology,
49100	to  construct  an  analogy between systems known to have the power to
49200	process symbols, namely, persons and  computers.      The  properties
49300	compared  in  the  analogy  are obviously not physical or substantive
49400	such as blood and wires, but functional and procedural.   We want  to
49500	assume  that  the  not  well-  understood  procedures of thought in a
49600	person are similar to the somewhat better  understood  procedures  of
49700	symbol-processing  which  take  place in a computer.   The analogy is
49800	one of functional or procedural equivalence. (For a  further  account
49900	of   functional   analysis   see   Hempel  (1965)).   Mousetraps  are
50000	functionally equivalent.   There  exists  a  large  set  of  physical
50100	mechanisms  for  catching mice. The term "mousetrap" says what all of
50200	the set has in common.   They take as input a live mouse and yield as
50300	output  a dead one. Systems equivalent from one point of view may not
50400	be equivalent from another. (Fodor,1968).
50500		If  model  and  human  are  indistinguishable at the manifest
50600	level of linguistic I-O pairs, then they can be considered equivalent
50700	at  that  level.     If  they can be shown to be indistinguishable at
50800	more internal symbolic levels, then a  stronger  equivalence  becomes
50900	achieved.    How  stringent  and  how  extensive  are the demands for
51000	equivalence to be?    Must there be point-to-point correspondences at
51100	every  level?   What  is to count as a point and what are the levels?
51200	Procedures  can  be  specified  and  ostensively  pointed  to  in  an
51300	algorithm  but  how can we point to and observe inaccessible symbolic
51400	processes in a person's head?  There is  a  limit  to  studying  real
51500	"underlying  processes".     Einstein likened this situation to a man
51600	explaining the behavior of a watch without  opening  it:    "He  will
51700	never  be  able to compare his picture with the real mechanism and he
51800	cannot even imagine the possibility or meaning of such a comparison".
51900	(Evolution of Physics).
52000		In  constructing  an   algorithm   one   puts   together   an
52100	organization  of  collaborating  functions or procedures.  A function
52200	takes some symbolic structure  as  input  and  yields  some  symbolic
52300	structure as output. Two computationally equivalent functions, having
52400	the same input and yielding the same output, can differ `inside'  the
52500	function at the instruction level.
52600		Consider an elementary programming problem which students  in
52700	symbolic  computation  are commonly asked to solve. Given a list L of
52800	symbols, L=(A B C D), as input, construct  a  function  or  procedure
52900	which will convert this list to the list RL in which the order of the
53000	symbols is reversed, i.e.  RL=(D C B  A).  There  are  many  ways  of
53100	solving  this  problem. The correct solutions will be computationally
53200	equivalent at  the  input-output  level  since  they  take  the  same
53300	symbolic structures as input and produce the same symbolic output.
53400		If we propose  that  an  algorithm  we  have  constructed  is
53500	functionally  equivalent  to what goes on in humans when they process
53600	symbolic  structures,  how   can   we   justify   this   position   ?
53700	Indistinguishability  tests  at,  say,  the  linguistic level provide
53800	evidence only for beginning equivalence. We would like to be able  to
53900	get  inside the underlying processes in humans the way we can with an
54000	algorithm by inspecting its instructional code. (Admittedly we do not
54100	observe  processes  directly but only their products). The difficulty
54200	lies in identifying, making tangible and counting processes in  human
54300	heads.   Many  experiments  must be designed and carried out. We must
54400	have great patience with this type of experimental psychology.
54500		In  the  meantime,  besides weak equivalence and plausibility
54600	arguments, one might appeal  to  extra-evidential  support  offering
54700	parallelisms  from  other  relevant  domains. One can offer analogies
54800	between what is known to  go  on  at  a  molecular  level  in  living
54900	organisms  and  what  goes  on  in  an  algorithm. For example, a DNA
55000	molecule in the nucleus of a cell consists  of  an  ordered  sequence
55100	(list)  of nucleotide bases (symbols) coded in triplets termed codons
55200	(words). Each element of the codon specifies which amino acid  during
55300	protein  synthesis  is  to  be  linked into the chain of polypeptides
55400	making up the protein.  The codons function like  instructions  in  a
55500	programming  language.  One  codon  is known to operate as a terminal
55600	symbol analogous to symbols in an algorithm which terminate  the  end
55700	of a list. If a stop codon appears in the middle of a sequence rather
55800	than at its normal terminal position, as in a point mutation, further
55900	protein  synthesis  is  prevented. The polypeptide chain resulting is
56000	abnormal and may have lethal or trivial consequences for the organism
56100	depending  on what other collaborating processes require to be handed
56200	over to them. Similarly in an algorithm. If a terminating  symbol  is
56300	incorrect  in  a  procedure,  the  procedure  cannot function. Such a
56400	result may be lethal or trivial to the algorithm  depending  on  what
56500	role  the  faulty  procedure  plays  in theoverall organization. Each
56600	function  in  an  algorithm  is  embedded  in  an   organization   of
56700	collaborating  functions  just as is the case in living organisms. We
56800	know that at the molecular level  of  living  organisms  there  exist
56900	rules  for  processes  such  as serial progression along a nucleotide
57000	sequence which are analogous to stepping down a list in an algorithm.
57100	Further  analogies  can  be made between point mutations in which DNA
57200	codons  can  be  inserted,  deleted,  substituted  or  reordered  and
57300	symbolic  computation  in  which  the  same  operations  are commonly
57400	carried out.   Such  analogies  are  interesting  as  extraevidential
57500	support   but  obviously  closer  linkages  are  needed  between  the
57600	macro-level of symbolic processes and the  micro-level  of  molecular
57700	information-processing .
57800		To  obtain  evidence  for  the  acceptability  of  a   model,
57900	empirical  tests  are utilized in validation procedures.   Such tests
58000	should also tell us which is the best among alternative versions of a
58100	model  and  among  alternative models. Scientific explanations do not
58200	stand alone in isolation.   They  are  evaluated  relative  to  rival
58300	contenders  for  the  position  of "best available". Once we accept a
58400	theory or model as the best available, can we be sure it  is  correct
58500	or true?    We can never know with certainty. Theories and models are
58600	provisional approximations to nature destined to become superseded by
58700	better ones.