00100	 EXPLANATIONS AND MODELS
00200	 The Nature of Explanation
00300		It is perhaps as difficult to explain explanation  itself  as
00400	it is to explain anything else. (Nothing, except everything, explains
00500	anything). The explanatory practices  of  different  sciences  differ
00600	widely but they all share the purpose of someone attempting to answer
00700	someone  else's  (or  his own) why-how-what-etc.    questions about a
00800	situation, event, episode, object  or  phenomenon.  Thus  explanation
00900	implies  a dialogue whose participants share some interests, beliefs,
01000	and values.   A consensus must exist about what  are  admissible  and
01100	appropriate  questions and answers.    The participants must agree on
01200	what is a sound and reasonable  question  and  what  is  a  relevant,
01300	intelligible,  and  (believed) correct answer. The explainer tries to
01400	satisfy  a  questioner's  curiosity  by  making  comprehensible   why
01500	something  is  the  way  it  is.   The answer may be a definition, an
01600	example, a synonym, a story, a theory, a model-description, etc.  The
01700	answer  attempts  to  satisfy  curiosity by settling belief, at least
01800	temporarily, since scientific beliefs are corrigible and revisable. A
01900	scientific  explanation aims at convergence of belief in the relevant
02000	expert community.
02100		Suppose a man dies and a questioner (Q) asks an explainer (E): 
02200		Q: Why  did  the  man  die?  
02300	One answer might be:
02400		E: Because he swallowed  cyanide.
02500	This explanation might be sufficient to satisfy Q's curiosity and  he
02600	and he stops asking further questions. Or he might continue:
02700	        Q. Why did the cyanide kill him?
02800	and E replies:
02900	        E: Anyone who swallows  cyanide dies.
03000	This explanation appeals to a universal generalization under which is
03100	subsumed  the  particular  fact  of  this  man's  death.  Subsumptive
03200	explanations  satisfy  some  questioners  but  not  others  who,  for
03300	example,  might  want  to  know  about  the  physiological mechanisms
03400	involved.                                                            
03500	        Q: How does cyanide work in causing death?
03600	        E: It stops respiration so the person dies from lack of oxygen.
03700		If Q has biochemical interests he might inquire further: 
03800		Q:What is cyanide's mechanism of drug action on the
03900		respiratory center?
04000		The last two questions refer to causes. When human action  is
04100	to  be  explained,  confusion  easily  arises  between  appealing  to
04200	physical, mechanical causes and appealing to  symbolic-level  reasons
04300	which   constitute  learned,  acquired  strategies  seemingly  of  an
04400	ontological order different from causes. (See Toulmin, 1971).
04500		It is established clinical knowledge that the phenomena of
04600	the paranoid mode can be found associated with a variety of  physical
04700	disorders.    For example, paranoid thinking can be found in patients
04800	with  head   injuries,   hyperthyroidism,   hypothyroidism,   uremia,
04900	pernicious   anemia,   cerebral  arteriosclerosis,  congestive  heart
05000	failure, malaria and epilepsy.      Also drug  intoxications  due  to
05100	alcohol,  amphetamines,  marihuana  and LSD can be accompanied by the
05200	paranoid mode. In these cases the paranoid mode is not a  first-order
05300	disorder  but  a  way  of  processing information in reaction to some
05400	other underlying disorder. To account for the association of paranoid
05500	thought  with  these  physical  states  of  illness,  a psychological
05600	theorist might be tempted to hypothesize that a  purposive  cognitive
05700	system would attempt to explain ill health by attributing it to other
05800	malevolent human agents. But before making such an explanatory  move,
05900	we must consider the at-times elusive distinction between reasons and
06000	causes in explanations of human behavior.
06100		One  view  of  the  association  of  the  paranoid  mode with
06200	physical disorders might be that the physical illness  simply  causes
06300	the  paranoia,  through  some  unknown mechanism, at a physical level
06400	beyond the influence of deliberate self-direction and self-control.
06500	That  is,  the  resultant  paranoid  mode  represents  something that
06600	happens to a person as victim, not  something  that  he  does  as  an
06700	active  agent.   Mechanical causes thus provide one type of reason in
06800	explaining behavior. Another view is that the paranoid  mode  can  be
06900	explained  in terms of symbolically represented reasons consisting of
07000	rules and patterns of rules which specify an agent's  intentions  and
07100	beliefs.   In  a given situation does a person as an agent recognize,
07200	monitor and control what he is doing or trying to do?    Or  does  it
07300	just happen to him automatically without conscious deliberation?
07400		This question raises a third view, namely  that  unrecognized
07500	reasons,  aspects of the symbolic representation which are sealed off
07600	from reflective deliberation, can function like mechanical causes  in
07700	that  they  are  inaccessible  to  voluntary  control. If they can be
07800	brought to consciousness, such  reasons  can  sometimes  be  modified
07900	voluntarily  by  the  agent,  who, using ordinary language as its own
08000	metalanguage, can reflexively talk to  and  instruct  himself.   This
08100	second-order  monitoring  and control through language contrasts with
08200	an agent's inability to modify mechanical causes or symbolic  reasons
08300	which    lie    beyond    the   influence   of   self-criticism   and
08400	self-emancipation  carried  out   through   linguistically   mediated
08500	argumentation.     Timeworn  conundrums  about concepts of free-will,
08600	determinism, responsibility, consciousness and the powers  of  mental
08700	action  here  plague  us  unless  we can take advantage of a computer
08800	analogy in which a clear and  useful  distinction  is  drawn  between
08900	levels  of mechanical hardware and symbolically represented programs.
09000	This important distinction will be elaborated shortly.
09100	
09200		Each of these three views provides a serviceable  perspective
09300	depending on how a disorder is to be explained and corrected.    When
09400	paranoid processes occur during amphetamine intoxication, they can be
09500	viewed  as  biochemically  caused and beyond the patient's ability to
09600	control volitionally through internal self-correcting dialogues  with
09700	himself.  When a paranoid moment occurs in a normal person, it can be
09800	viewed as involving a symbolic misinterpretation.   If  the  paranoid
09900	misinterpretation  is  recognized as unjustified, a normal person has
10000	the emancipatory power  to  revise  or  reject  it  through  internal
10100	debate.  Between  these  extremes of drug-induced paranoid states and
10200	the self-correctable paranoid moments of the normal person, lie cases
10300	of  paranoid  personalities, paranoid reactions and the paranoid mode
10400	associated   with   the   major    psychoses    (schizophrenic    and
10500	manic-depressive).
10600		One opinion has it that the major psychoses are a consequence
10700	of unknown  physical  causes  and  are  beyond  deliberate  voluntary
10800	control.   But  what  are we to conclude about paranoid personalities
10900	and paranoid reactions where no hardware disorder  is  detectable  or
11000	suspected?  Are  such  persons  to  be  considered  patients  to whom
11100	something is mechanically happening at the physical level or are they
11200	agents whose behavior is  a  consequence  of  what  they  do  at  the
11300	symbolic level?   Or are they both agent and patient depending on how
11400	one views the self-modifiability of their  symbolic  processing?   In
11500	these  perplexing  cases  we  shall take the position that in normal,
11600	neurotic and characterological  paranoid  modes,  the  psychopathlogy
11700	represents  something  that happens to a man as a consequence of what
11800	he has experientially undergone, of something he now does, and
11900	something he now undergoes.    Thus he is both agent and victim whose
12000	symbolic processes have powers to do and liabilities to undergo.  His
12100	liabilities  are  reflexive  in that he is victim to, and can succumb
12200	to, his own symbolic structures.
12300	
12400		From this standpoint I  would  postulate  a  duality  at  the
12500	symbolic  level  between  reasons  and causes. That is, a consciously
12600	unrecognized reason can operate like  a  mechanical  cause  in  being
12700	inaccessible  to voluntary modification by symbolic reprogramming. It
12800	is, of course, not reasons themselves which operate as causes but the
12900	execution  of  the  reason-rules  which  serves  as  a determinant of
13000	behavior.  Human symbolic behavior is non-determinate to  the  extent
13100	that  it  is  autonomously self-determinate. Thus the power to select
13200	among alternatives, to make some decisions freely and to change one's
13300	mind  is  non-illusory.  When a reason is recognized to function as a
13400	cause  and  is  accessible  to  self-monitoring  (the  monitoring  of
13500	monitoring),  emancipation  from  it  can  occur  through  change  or
13600	rejection of belief. In this sense an at least two-levelled system is
13700	self-changeable and self-emancipatory, within limits.
13800		Explanations  both  in  terms  of  causes  and reasons can be
13900	indefinitely extended and endless questions  can  be  asked  at  each
14000	level of analysis.  Just as the participants in explanatory dialogues
14100	decide what is taken to be problematic, so they  also  determine  the
14200	termini   of   questions   and  answers.   Each  discipline  has  its
14300	characteristic stopping points and boundaries.
14400		Underlying such explanatory dialogues are larger and  smaller
14500	constellations   of   concepts   which   are  taken  for  granted  as
14600	nonproblematic background.    Hence in considering the strategies  of
14700	the paranoid mode "it goes without saying" that any living teleonomic
14800	system ,as the larger constellation ,  strives  for  maintenance  and
14900	expansion  of life. Also it should go without saying that, at a lower
15000	level, ion transport takes place through nerve-cell membranes.  Every
15100	function  of  an  organism  can  be viewed as governing a subfunction
15200	beneath and depending on a transfunction above which  calls  it  into
15300	play for a purpose.
15400		Just as there are many alternative ways of describing, there
15500	are many alternative ways of explaining.  An explanation is geared to
15600	some  level  of  what  the  dialogue  participants  take  to  be  the
15700	fundamental  structures  and processes under consideration.  Since in
15800	psychiatry   we   cope   with   patients'   problems   using   mainly
15900	symbolic-conceptual  techniques,(it is true that the pill, the knife,
16000	and electricity are also available), we are interested in aspects  of
16100	human  conduct  which can be explained, understood, and modified at a
16200	symbol-processing  level.  Psychiatrists  need  theoretical  symbolic
16300	systems from which their clinical experience can be logically derived
16400	to interpret the case histories of their patients. Otherwise they are
16500	faced  with  mountains  of  indigestible  data  and  dross.  To quote
16600	Einstein: "Science is an attempt to make the chaotic diversity of our
16700	sense  experience correspond to a logically uniform system of thought
16800	by correlating single experiences with the theoretic structure."
16900	
17000	 The Symbol Processing Viewpoint
17100	
17200		Segments and sequences of human behavior can be studied  from
17300	many  perspectives.   In  this  monograph  I  shall view sequences of
17400	paranoid symbolic behavior from an information-processing  standpoint
17500	in  which  persons  are  viewed  as symbol users. For a more complete
17600	explication and justification of this perspective , see Newell (1973)
17700	and Newell and Simon (1972).
17800		In brief, from this vantage point we  define  information  as
17900	knowledge   in   a  symbolic  code.  Symbols  are  considered  to  be
18000	representations  of  experience  classified   as   objects,   events,
18100	situations and relations. A symbolic process is a symbol-manipulating
18200	activity posited to account for observable symbolic behavior such  as
18300	linguistic  interaction. Under the term "symbol-processing" I include
18400	the seeking, manipulating and generating of symbols.
18500		Symbol-processing   explanations   postulate   an  underlying
18600	structure  of  hypothetical  processes,  functions,  strategies,   or
18700	directed  symbol-manipulating procedures, having the power to produce
18800	and being responsible for observable patterns of  phenomena.  Such  a
18900	structure  offers an ethogenic (ethos = conduct or character, genic =
19000	generating)  explanation  for  sequences  or  segments  of   symbolic
19100	behavior.  (See Harre and Secord,1972).  From an ethogenic viewpoint,
19200	we can posit processes, functions, procedures and strategies as being
19300	responsible  for  and  having  the  power  to  generate  the symbolic
19400	patterns  and  sequences  characteristic  of   the   paranoid   mode.
19500	"Strategies"  is  perhaps the best general term since it implies ways
19600	of obtaining an objective - ways which have suppleness and pliability
19700	since choice of application depends on circumstances.        However,
19800	I shall use all these terms interchangeably.
19900	
20000	 Symbolic Models
20100		Theories and  models  share  many  functions  and  are  often
20200	considered  equivalent.   One  important distinction lies in the fact
20300	that a theory states a subject has a certain structure but  does  not
20400	exhibit  that  structure in itself. (See Kaplan,1964). In the case of
20500	computer simulation models there exists a further useful distinction.
20600	Computer  simulation  models  which  have  the ability to converse in
20700	natural language using teletypes, actualize or realize  a  theory  in
20800	the  form of a dialogue algorithm. In contrast to a verbal, pictorial
20900	or  mathematical  representation,  such  a  model,  as  a  result  of
21000	interaction,  changes  its  states  over  time and ends up in a state
21100	different from its initial state.
21200		Einstein once remarked, in contrasting the act of description
21300	with  what  is  described,  that it is not the function of science to
21400	give the taste of the soup.  Today  this  view  would  be  considered
21500	unnecessarily  restrictive.  For  example, a major test for synthetic
21600	insulin is whether it reproduces the effects, or at least some of the
21700	effects  (such as lowering blood sugar), shown by natural insulin. To
21800	test whether a simulation is successful, its effects must be compared
21900	with  the  effects produced by the naturally-occuring subject-process
22000	being modelled.     An interactive simulation model which attempts to
22100	reproduce  sequences of experienceable reality, offers an interviewer
22200	a first-hand experience with  a  concrete  case.  In  constructing  a
22300	computer  simulation, a theory is modelled to discover a sufficiently
22400	rich  structure  of  hypotheses  and  assumptions  to  generate   the
22500	observable  subject-behavior  under  study.      A dialogue algorithm
22600	allows an observer to interact with a concrete specimen of a class in
22700	detail. In the case of our model, the level of detail is the level of
22800	the symbolic behavior of  conversational  language.   This  level  is
22900	satisfying  to  a clinician since he can compare the model's behavior
23000	with its natural human counterparts using familiar skills of clinical
23100	dialogue. Communicating with the paranoid model by means of teletype,
23200	an interviewer can directly experience for himself a  sample  of  the
23300	type  of  impaired social relationship which develops with someone in
23400	a paranoid mode.
23500		An algorithm composed of  symbolic  computational  procedures
23600	converts  input  symbolic  structures into output symbolic structures
23700	according to certain  principles.   The  modus  operandi  of  such  a
23800	symbolic  model  is simply the workings of an algorithm when run on a
23900	computer. At this level of explanation, to  answer  `why?'  means  to
24000	provide  an  algorithm  which  makes explicit how symbolic structures
24100	collaborate, interplay  and  interlock  -  in  short,  how  they  are
24200	organized to generate patterns of manifest phenomena.
24300	
24400		To simulate the sequential input-output behavior of a  system
24500	using  symbolic  computational  procedures,  one writes  an algorithm
24600	which, when run on a computer, produces symbolic behavior  resembling
24700	that  of  the  subject  system  being  simulated.    (Colby,1973) The
24800	resemblance is achieved through the  workings  of  an  inner  posited
24900	structure   in   the   form  of  an  algorithm,  an  organization  of
25000	symbol-manipulating procedures which  are  ethogenically  responsible
25100	for the characteristic observable behavior at the input-output level.
25200	Since we do not know the structure of the "real" simulative processes
25300	used  by  the mind-brain, our posited structure stands as an imagined
25400	theoretical  analogue,  a  possible  and  plausible  organization  of
25500	processes  analogous  to  the  unknown  processes  and  serving as an
25600	attempt to explain the  workings  of  the  system  under  study.    A
25700	simulation  model  is  thus  deeper  than  a  structureless black-box
25800	explanation because it postulates functionally  equivalent  processes
25900	inside  the  box  to  account  for  outwardly  observable patterns of
26000	behavior.    A   simulation   model   constitutes   an   interpretive
26100	explanation  in  that  it  makes intelligible the connections between
26200	external input, internal states and output  by  positing  intervening
26300	symbol-processing  procedures  operating  between  symbolic input and
26400	symbolic output. To be  illuminating,  a  description  of  the  model
26500	should  make  clear  why  and  how it reacts as it does under various
26600	circumstances.
26700		Citing a universal generalization to explain an  individual's
26800	behavior  is unsatisfactory to a questioner who is interested in what
26900	powers and liabilities are latent behind manifest phenomena.  To  say
27000	"x is nasty because x is paranoid and all paranoids are nasty" may be
27100	relevant, intelligible and correct. But another type  of  explanation
27200	is  possible:  a model-explanation referring to a structure which can
27300	account for "nasty" behavior as a consequence of input  and  internal
27400	states  of  a  system.     A  model  explanation specifies particular
27500	antecedents and processes  through  which  antecedents  generate  the
27600	phenomena.   An ethogenic approach to explanation assumes perceptible
27700	phenomena display the regularities and nonrandom irregularities  they
27800	do  because  of  the  nature  of  an  imperceptible  and inaccessible
27900	underlying structure.    The  posited  theoretical  structure  is  an
28000	idealization,  an  imagined analogue to the unobservable structure in
28100	persons.
28200		When attempts are made to explain human behavior,  principles
28300	in  addition  to  those accounting for the natural order are invoked.
28400	"Nature entertains no opinions about us", said Nietzsche.  But  human
28500	natures  do,    and  therein  lies  a  source  of  complexity for the	
28600	understanding of human conduct. Until the first quarter of  the  20th
28700	century,  natural  sciences  were  guided  by  the Newtonian ideal of
28800	perfect process knowledge  about  inanimate  objects  whose  behavior
28900	could  be  subsumed  under lawlike generalizations.  When a deviation
29000	from a law was  noticed,  it  was  the  law  which  was  subsequently
29100	modified, since by definition physical objects did not have the power
29200	to break laws. When the planet Mercury was observed to  deviate  from
29300	the orbit predicted by Newtonian theory, no one accused the planet of
29400	being an intentional agent disobeying a law. Instead it was suspected
29500	that something was incorrect about the theory.
29600		Subsumptive explanation is the acceptable norm in many fields
29700	but  it is seldom satisfactory in accounting for particular sequences
29800	of behavior in living purposive systems.  When physical  bodies  fall
29900	in the  macroscopic world, few find it scientifically useful to posit
30000	that bodies have an intention to fall .  But in the  case  of  living
30100	systems,  especially  ourselves,  our  ideal  explanatory practice is
30200	teleonomically Aristotelian in utilizing a concept of intention. (For
30300	a thorough discussion of purpose and intentionality see Boden,1972).
30400		Consider  a  man participating in a high-diving contest.   In
30500	falling towards the water he accelerates at the rate of 32  feet  per
30600	second. Viewing the man simply as a falling body, we explain his rate
30700	of fall by appealing to a physical law.  Viewing the man as  a  human
30800	intentionalistic  agent,  we  explain  his  dive  as the result of an
30900	intention to dive in a  certain  way  in  order  to  win  the  diving
31000	contest.  His  conduct  (in  contrast  to  mere movement) involves an
31100	intended following of certain conventional rules for what  is  judged
31200	by  humans  to constitute, say, a swan dive. Suppose part-way down he
31300	chooses to change  his  position  in  mid-air  and  enter  the  water
31400	thumbing his nose at the judges. He cannot disobey the law of falling
31500	bodies but he can disobey or ignore the rules of diving. He can  also
31600	make  a gesture which expresses disrespect and which he believes will
31700	be interpreted as such by the onlookers.   Our diver  breaks  a  rule
31800	for  diving but follows another rule which prescribes gestural action
31900	for insulting behavior.    To  explain  the  actions  of  diving  and
32000	nose-thumbing,  we would appeal, not to laws of natural order, but to
32100	an additional order, to principles of  human  order.  This  order  is
32200	superimposed  on  laws  of  natural  order  and  takes  into  account
32300	(1)standards of appropriate action in certain situations and (2)  the
32400	agent's  inner considerations of intention, belief and value which he
32500	finds compelling from his point of view. In this type of  explanation
32600	the  explanandum,  that  which  is  being  explained,  is the agent's
32700	informed actions, not  simply  his  movements.  When  a  human  agent
32800	performs  an  action  in  a  situation,  we  can  ask:  is the action
32900	appropriate to that situation and if not, why did the  agent  believe
33000	his action to be called for?
33100		Symbol-processing explanations of human conduct rely on
33200	concepts of intention, belief, action, affect, etc. These  terms  are
33300	close to the terms of ordinary language as is characteristic of early
33400	stages of explanations. It is also important to note that such  terms
33500	are  commonly utilized in describing computer algorithms which follow
33600	rules in striving to achieve goals. In an  algorithm  these  ordinary
33700	language terms can be explicitly defined and represented.
33800		Psychiatry deals with the practical concerns of inappropriate
33900	action, belief, etc. on the part of a patient. His  behavior  may  be
34000	inappropriate  to  onlookers  since  it  represents  a lapse from the
34100	expected, a contravention of the human order. It may even appear this
34200	way  to  the  patient  in  monitoring  and  directing  himself.   But
34300	sometimes, as in severe cases of the  paranoid  mode,  the  patient's
34400	behavior  does  not  appear  anomalous to himself.  He maintains that
34500	anyone  who  understands  his  point  of  view,  who   conceptualizes
34600	situations  as  he  does  from the inside, would consider his outward
34700	behavior appropriate and justified. What he does  not  understand  or
34800	accept is that his inner conceptualization is mistaken and represents
34900	a misinterpretation of the events of his experience.
35000		The  model  to  be  presented  in  the  sequel constitutes an
35100	attempt to explain some regularities and  particular  occurrences  of
35200	symbolic   (conversational)   paranoid  behavior  observable  in  the
35300	clinical situation of a psychiatric interview.   The  explanation  is
35400	at the symbol-processing level of linguistically communicating agents
35500	and  is  cast  in  the  form  of  a  dialogue  algorithm.   Like  all
35600	explanations,  it  is  tentative,  incomplete,  and does not claim to
35700	represent the only conceivable structure of processes .
35800	
35900	The Nature of Algorithms
36000	
36100		Theories  can  be  presented  in various forms: prose essays,
36200	mathematical  equations  and  computer  programs.    To   date   most
36300	theoretical  explanations in psychiatry and psychology have consisted
36400	of natural language essays with all their  well-known  vagueness  and
36500	ambiguities.  Many  of  these  formulations have been untestable, not
36600	because relevant observations were lacking but because it was unclear
36700	what  the  essay  was really saying.  Clarity is needed.  Science may
36800	begin with metaphors but it should end up with algorithms.
36900		An  alternative  way of formulating psychological theories is
37000	now available in the form of symbol-processing  algorithms,  computer
37100	programs,   which   have  the  virtue  of  being  explicit  in  their
37200	articulation and which can be run on  a  computer  to  test  internal
37300	consistency and external correspondence with the data of observation.
37400	The subject-matter or subject of a model is what it is  a  model  of;
37500	the  source of a model is what it is based upon. Since we do not know
37600	the "real" algorithms used by  people,  we  construct  a  theoretical
37700	model, based upon computer algorithms.     This  model  represents  a
37800	partial  analogy. (Harre, 1970).   The partial analogy is made at the
37900	symbol-processing  level,  not  at  the  hardware  level.           A
38000	functional,   computational   or   procedural  equivalence  is  being
38100	postulated.     The question then becomes  one  of  categorizing  the
38200	extent  of the equivalence.         A beginning (first-approximation)
38300	functional equivalence might be defined  as  indistinguishability  at
38400	the  level  of  observable  I-O  pairs.  A stronger equivalence would
38500	consist of indistinguishability at inner I-O levels.  That is,  there
38600	exists  a  correspondence  between  what  is being done and how it is
38700	being done at a given operational level.
38800		An  algorithm represents an organization of symbol-processing
38900	strategies or functions which represent an "effective procedure".  An
39000	effective procedure consists of three components:
39100	
39200		(1) A programming language in which procedural rules of
39300		    behavior can be rigorously and unambiguously specified.
39400		(2) An organization of procedural rules which constitute 
39500		    the algorithm.
39600		(3) A machine processor which can rapidly and reliably carry
39700		    out the processes specified by the procedural rules.
39800	The  specifications  of  (2),  written  in   the   formally   defined
39900	programming  language  of  (1),  is  termed  an  algorithm or program
40000	whereas (3) involves a computer as the machine processor,  a  set  of
40100	deterministic  physical  mechanisms  which can perform the operations
40200	specified in the algorithm.   The  algorithm  is  called  "effective"
40300	because  it  actually works, performing as intended and producing the
40400	effects desired bt  the  model  builders  when  run  on  the  machine
40500	processor.
40600		A simulation model is composed  of  procedures  taken  to  be
40700	analogous  to  the  imperceptible and inaccessible procedures.     We
40800	are not claiming they ARE analogous,  we  are  MAKING  them  so.  The
40900	analogy  being  drawn  here  is between specified processes and their
41000	generating  systems.  Thus,  in   comparing   mental   processes   to
41100	computational processes, we might assert:
41200	
41300	      mental process    computational process
41400	      --------------:: ----------------------
41500	      brain hardware      computer hardware and
41600	      and programs           programs
41700	
41800		Many of the classical mind-brain problems arose because there
41900	did not exist a familiar,  well-understood  analogy  to  help  people
42000	imagine how a system could work having a clear separation between its
42100	hardware descriptions and its program descriptions.  With the  advent
42200	of  computers  and  programs  some mind-brain perplexities disappear.
42300	(Colby,1971).  The analogy is not simply  between  computer  hardware
42400	and  brain  wetware.   We  are not comparing the structure of neurons
42500	with the structure of transistors; we are comparing the
42600	organization of symbol-processing procedures  in  an  algorithm  with
42700	symbol-processing  procedures of the mind-brain.  The central nervous
42800	system contains a representation of the experience of its holder.   A
42900	model  builder has a conceptual representation of that representation
43000	which he demonstrates in the form of a model.  Thus the  model  is  a
43100	demonstration of a representation of a representation.
43200		An  algorithm  can  be  run  on  a  computer  in two forms, a
43300	compiled version and an interpreted version. In the compiled  version
43400	a  preliminary  translation  has  been  made  from  the  higher-level
43500	programming  language  (source  language)  into  lower-level  machine
43600	language  (object  language)  which  controls  the  on-off  state  of
43700	hardware switching devices. When the compiled  version  is  run,  the
43800	instructions  of  the machine-language code are directly executed. In
43900	the interpreted version each high-level language instruction is first
44000	translated  into  machine language, executed, and then the process is
44100	repeated with the next instruction.   One  important  aspect  of  the
44200	distinction  between  compiled  and  interpreted versions is that the
44300	compiled version, now written in  machine  language,  is  not  easily
44400	accessible  to  change  using  the higher-level language. In order to
44500	change the program, the interpreted version must be modified  in  the
44600	source  language  and then re-compiled into the object language.  The
44700	rough analogy with ever-changing  human  symbolic  behavior  lies  in
44800	suggesting  that  modifications require change at the source-language
44900	level. Otherwise compiled algorithms are inaccessible to second order
45000	monitoring and modification.
45100		Since we are taking running computer programs as a source  of
45200	analogy for a paranoid model, logical errors or pathological behavior
45300	on  the  part   of   such   programs   are   of   interest   to   the
45400	psychopathologist.   These  errors  can  be  ascribed to the hardware
45500	level, to the interpreter, or to the programs  which  the  interpreter
45600	executes.    Different  remedies are required at different levels. If
45700	the analogy  is  to  be  clinically  useful  in  the  case  of  human
45800	pathological  behavior,  it  will  become  a  matter  of  influencing
45900	symbolic behavior with the appropriate techniques.
46000		Since  the algorithm is written in a programming language, it
46100	is hermetic except to a few people,  who  in  general  do  not  enjoy
46200	reading   other  people's  code.     Hence  the  intelligibility  and
46300	scrutability requirement for explanations must be met in other  ways.
46400	In  an attempt to open the algorithm to scrutiny I shall describe the
46500	model in detail using diagrams and interview examples profusely.
46600	
46700	
46800	Analogy
46900	
47000		I  have  stated  that  an  interactive  simulation  model  of
47100	symbol-manipulating  processes  reproduces  sequences   of   symbolic
47200	behavior  at the level of linguistic communication.  The reproduction
47300	is achieved through the operations of an algorithm consisting  of  an
47400	organization   of   hypothetical   symbol-processing   strategies  or
47500	procedures which can  generate  the  I-O  behavior  of  the  subject-
47600	processes   under  investigation. The   algorithm  is  an  "effective
47700	procedure" in the sense it really works in the manner intended by the
47800	model-builders.  In the model to be described, the paranoid algorithm
47900	generates  linguistic  I-O  behavior  typical   of   patients   whose
48000	symbol-processing  is dominated by the paranoid mode. Comparisons can
48100	be made between samples of the I-O behaviors of patients  and  model.
48200	But  the  analogy is not to be drawn at this level.   Mynah birds and
48300	tape recorders also reproduce human linguistic behavior  but  no  one
48400	believes  the  reproduction  is achieved by powers analogous to human
48500	powers.   Given that the manifest outermost I-O behavior of the model
48600	is  indistinguishable  from  the  manifest  outward  I-O  behavior of
48700	paranoid patients, does this imply that the  hypothetical  underlying
48800	processes  used  by  the  model are analogous to (or perhaps the same
48900	as?) the underlying processes used by persons in the  paranoid  mode?
49000	This deep and far-reaching question should be approached with caution
49100	and only when we are  first  armed  with  some  clear  notions  about
49200	analogy, similarity, faithful reproduction, indistinguishability  and
49300	functional equivalence.
49400		In comparing two things (objects, systems or processes )  one
49500	can   cite   properties  they  have  in  common  (positive  analogy),
49600	properties they do not share (negative analogy) and properties  which
49700	we do not yet know whether they are positive or negative (neutral
49800	analogy). (See Hesse,1966). No two things are exactly alike in  every
49900	detail.   If  they  were identical in respect to all their properties
50000	then they would be copies. If they were identical  in  every  respect
50100	including  their  spatio-temporal  location we would say we have only
50200	one thing instead of two. Everything  resembles  something  else  and
50300	maybe everything else, depending upon how one cites properties.
50400		In an analogy a similarity relation is  evoked.  "Newton  did
50500	not  show  the  cause of the apple falling but he showed a similitude
50600	between the apple and the stars."(D`Arcy Thompson). Huygens suggested
50700	an analogy between sound waves and light waves in order to understand
50800	something less well-understood (light) in terms of  something  better
50900	understood   (sound).   To  account  for  species  variation,  Darwin
51000	postulated a  process  of  natural  selection.    He  constructed  an
51100	analogy  from two sources, one from artificial selection as practiced
51200	by domestic breeders of animals and one from  Malthus'  theory  of  a
51300	competition  for  existence  in a population increasing geometrically
51400	while its resources increase arithmetically. Bohr's model of the atom
51500	offered  an  analogy  between solar system and atom. These well-known
51600	historical examples should be sufficient here to illustrate the  role
51700	of analogies in theory construction.    Analogies are made in respect
51800	to  those  properties  which  constitute  the  positive  and  neutral
51900	analogy.     The  negative analogy is ignored.   Thus Bohr's model of
52000	the atom as a miniature planetary system was not intended to  suggest
52100	that  electrons  possessed  color or that planets jumped out of their
52200	orbits. 
52300	
52400	Functional Equivalence
52500	
52600		When human symbolic processes are the subject of a simulation
52700	model, we draw the analogy from two sources, symbolic computation and
52800	psychology.  The  analogy  made  is between systems known to have the
52900	power to process symbols, namely,  persons  and  computers.       The
53000	properties  compared  in  the  analogy  are obviously not physical or
53100	substantive such as blood and wires, but functional  and  procedural.
53200	We  want  to  assume  that not-well-understood mental procedures in a
53300	person are similar to  the  more  accessible  and  better  understood
53400	procedures of symbol-processing which take place in a computer.   The
53500	analogy is one  of  functional  or  procedural  equivalence.  (For  a
53600	further    account   of   functional   analysis   see  Hempel, 1965).
53700	Mousetraps are functionally equivalent.    There exists a  large  set
53800	of  physical  mechanisms for catching mice. The term "mousetrap" says
53900	what each member of the set has in common.  Each  takes  as  input  a
54000	live  mouse  and yields as output a dead one. Systems equivalent from
54100	one point of view may not be equivalent from another (Fodor,1968).
54200		If  model  and  human  are  indistinguishable at the manifest
54300	level of linguistic I-O pairs, then they can be considered equivalent
54400	at  that  level.      If they can be shown to be indistinguishable at
54500	more internal symbolic levels, then a  stronger  equivalence  exists.
54600	How  stringent  and  how extensive are the demands for equivalence to
54700	be? Must the correspondence be point-to-point or can it be  the  more
54800	global system-to-system? Must there be point-to-point correspondences
54900	at every level? What is to count as a point and what are the  levels?
55000	Procedures  can  be  specified  and  ostensively  pointed  to  in  an
55100	algorithm, but how can we point to unobservable symbolic processes in
55200	a  person?    There  is  an  inevitable  limit  to  scrutinizing  the
55300	"underlying"  processes  of  the  world.     Einstein  likened   this
55400	situation to a man explaining the behavior of a watch without opening
55500	it: "He will never be able to  compare  his  picture  with  the  real
55600	mechanism  and  he  cannot even imagine the possibility or meaning of
55700	such a comparison".
55800		In  constructing  an   algorithm   one   puts   together   an
55900	organization  of  collaborating  functions or procedures.  A function
56000	takes some symbolic structure  as  input  and  yields  some  symbolic
56100	structure as output. Two computationally equivalent functions, having
56200	the same input and yielding the same output, can differ "inside"  the
56300	function at the instruction level.
56400		Consider  an elementary programming problem which students in
56500	symbolic computation are often asked to solve.  Given  a  list  L  of
56600	symbols,  L=(A  B  C  D), as input, construct a function or procedure
56700	which will convert this list to the list RL in which the order of the
56800	symbols  is  reversed,  i.e.   RL=(D  C B A).  There are many ways of
56900	solving this problem and the code of one student may  differ  greatly
57000	from that of another at the level of individual instructions. But the
57100	differences of such details are irrelevant. What  is  significant  is
57200	that  the  solutions  make  the required conversion from L to RL. The
57300	correct solutions will  all  be  computationally  equivalent  at  the
57400	input-output  level  since  they take the same symbolic structures as
57500	input and produce the same symbolic output.
57600		If  we  propose  that  an  algorithm  we  have constructed is
57700	functionally equivalent to what goes on in humans when  they  process
57800	symbolic   structures,   how   can   we   justify   this  position  ?
57900	Indistinguishability tests at,  say,  the  linguistic  level  provide
58000	evidence  only for beginning equivalence. We would like to be able to
58100	have access to the underlying processes in humans the way we can with
58200	algorithms.  (Admittedly, we do not directly observe processes at all
58300	levels but only  the  products  of  some).  The  difficulty  lies  in
58400	identifying,  making  accessible,  and  counting  processes  in human
58500	heads.    Many symbol-processing experiments are now  being  designed
58600	and  carried  out.  We  must  have  great  patience with this type of
58700	experimental information-processing psychology.
58800		In  the meantime, besides first-approximation I-O equivalence
58900	and plausibility arguments,  one  might  appeal  to  extra-evidential
59000	support  offering  parallelisms  from neighboring scientific domains.
59100	One can offer analogies between what is known to go on at a molecular
59200	level  in  the  cells  of  living  organisms  and  what goes on in an
59300	algorithm. For example, a DNA molecule  in  the  nucleus  of  a  cell
59400	consists  of an ordered sequence (list) of nucleotide bases (symbols)
59500	coded in triplets termed codons (words). Each element  of  the  codon
59600	specifies  which  amino acid during protein synthesis is to be linked
59700	into the chain of polypeptides making up the  protein.    The  codons
59800	function like instructions in a programming language. Some codons are
59900	known to operate as terminal  symbols  analogous  to  symbols  in  an
60000	algorithm  which  terminate  the  end of a list. If, as a result of a
60100	mutation, a nucleotide base is changed, the usual protein will not be
60200	synthesized.  The  polypeptide  chain  resulting  may  have lethal or
60300	trivial consequences for the  organism  depending  on  what  must  be
60400	passed  on to other processes which require polypeptides to be handed
60500	over to them. Similarly in an algorithm. If a symbol  or  word  in  a
60600	procedure  is incorrect, the procedure cannot operate in its intended
60700	manner.   Such a result may be lethal or  trivial  to  the  algorithm
60800	depending  on  what  information the faulty procedure must pass on at
60900	its interface with other procedures in the overall organization. Each
61000	procedure   in  an  algorithm  is  embedded  in  an  organization  of
61100	collaborating procedures just as are functions in  living  organisms.
61200	We  know that at the molecular level of living organisms there exists
61300	a process such as serial progression  along  a  nucleotide  sequence,
61400	which is analogous to stepping down a list in an algorithm.   Further
61500	analogies can be made between point mutations in which DNA bases  can
61600	be   inserted,   deleted,   substituted  or  reordered  and  symbolic
61700	computation in which the same operations are commonly carried out  on
61800	symbolic    structures.     Such   analogies   are   interesting   as
61900	extra-evidential support but obviously  closer  linkages  are  needed
62000	between  the macro-level of symbolic processes and the micro-level of
62100	molecular information-processing within cells.
62200		To obtain evidence  for  the  acceptability  of  a  model  as
62300	faithful  or  authentic,  empirical  tests are utilized as validation
62400	procedures. Such tests should also tell us which is  the  best  among
62500	alternative  versions  of  a  family  of  models  and,  indeed  among
62600	alternative families of models.  Scientific explanations do not stand
62700	alone  in  isolation. They are evaluated relative to rival contenders
62800	for the position of "best available".  Once we  accept  a  theory  or
62900	model  as  the  best available, can we be sure it is correct or true?
63000	We can never know with certainty. Theories and models are provisional
63100	and partial approximations to nature destined to become superseded by
63200	better ones.