00100	EVALUATION 
00200	
00300		The primary aim in constructing this model  was  to  explore,
00400	clarify, develop, test and improve -all with a model- a theory having
00500	explanatory value. To satisfy this aim, the model must meet norms  of
00600	internal    consistency   (systemicity)   and   norms   of   external
00700	correspondence with observation (testability). A secondary aim  would
00800	involve pragmatic norms of application.  These aims are not unrelated
00900	but the primary one is more  fundamental  since  useful  applications
01000	require some degree of consistency and correspondence to observation.
01100		As emphasized in Chapter  2,  a  model  in  the  form  of  an
01200	algorithm  consists  of  a structure of functions or procedures whose
01300	inner workings are  sufficient  to  reproduce  the  outward  symbolic
01400	behavior  under  consideration.   The theory embodied in the model is
01500	revealed  in  the  set of statements which illuminate the connections
01600	betweeen input and output, i.e.  which  describe  how  the  structure
01700	reacts under various circumstances.
01800		What constitutes a satisfactory explanation has been  treated
01900	in  2.1.     The "fit" or correspondence with facts of observation as
02000	indicated by measurements and empirical tests indicating  the  degree
02100	of  faithfulness  of the  reproduction,  were described in Chapter 6.
02200	Given that the model has met the above criteria, what does it  as  an
02300	artifact tell us about naturally-occurring paranoid processes?
02400		First, the model attempts to revisualize  or  reconceptualize
02500	the  phenomena  of  paranoid disorders. It draws attention to factors
02600	(such as the scan for malevolence )  which  might  not  otherwise  be
02700	attended  to.     Paranoid  disorders  are  not viewed as first-order
02800	"diseases" but as a mode of processing symbols secondary to a primary
02900	disturbance.    The patterns of linguistic paranoid behavior observed
03000	in an interview are produced by an underlying organized structure  of
03100	rules  and  not  by  a  variety  of random and unconnected mechanical
03200	failures.  Second, the underlying structure is posited to consist  of an algorithm,
03300	an organization of symbol-processing strategies or procedures. Third,
03400	the model as an analogy indicates that to change this structure,  its
03500	procedures  must  be  accessible to reprogramming in the higher-level
03600	language of the algorithm. Finally, as a conceptual reform, the model
03700	suggests that other types of psychopathologies might be viewed from a
03800	symbol-processing standpoint.
03900		Decision  procedures  for  consensus acceptability of a model
04000	sometimes depend not so much  on  truth,  an  elusive  state,  as  on
04100	whether  a  majority  of  the  relevant expert community believes the
04200	theory or model to approximate truth to some unknown  and  unknowable
04300	degree  and  to be better than promising available alternatives, that
04400	is, to be the best we can do for the time being.  A model is  tenable
04500	as  long  as  it is worth working with by improving it, extending it,
04600	devising experiments and tests  to  probe  it,  and  applying  it  in
04700	contexts  of  practical  action.  Validation  is ultimately a private
04800	experience of the individual. Empirical truth or  falsity  cannot  be
04900	proven with certainty, but their presence can be assayed by some sort
05000	of critical assessment and deliberation.  We can forgive  models  for
05100	being  only  nearly  true.   A theory or model may bring cognitive or
05200	pragmatic comfort, not because  it  is  TRUE  but  because  it  is  a
05300	workable  and  exciting  possibility  which represents an improvement
05400	over its contending rivals.
05500		Cognitive comfort is a  type  of  intellectual  satisfaction.
05600	Pragmatic comfort accrues from technically exploitable knowledge,i.e.
05700	applications which make things work the way humans want them to  work
05800	efficiently  in  practical  contexts of technological action. For the
05900	pragmatist, a model is a means to an end; for  the  theoretician,  an
06000	explanatory model is an end in itself. It is hoped that this paranoid
06100	model can contribute to understanding one of the mysteries  of  human
06200	conduct,  the paranoid mode. There remains the enigma of the paranoid
06300	"streak" which  renders  whole  nations  susceptible  to  ideological
06400	convictions  in  which  Elsewhereans  are  believed  to be malevolent
06500	oppressors.
06600		It  is  a  truism  of  methodology textbooks that an infinite
06700	number of theories or  models  can  account  for  the  same  data  of
06800	observation.        Without   questioning  whether  "infinite"  means
06900	indefinitely large or just more than one, we  must  allow  for  rival
07000	explanations.  For a rival to be a live and tenable option, it should
07100	be truly alternative (i.e., not just a family version saying the same
07200	thing  in  a  different way), and be confirmable or disconfirmable by
07300	tests.
07400		Prediction of new facts from a theory not only test a  theory
07500	but  provide  useful information. Not all acceptable theories predict
07600	new facts, e.g. Copernicus'. Although I would maintain that  faithful
07700	reproduction  (fidelity  as  measured  by  indistinguishability along
07800	specific dimensions) is a proper and major test for the  adequacy  of
07900	simulation models, it would be a bonus if our model could satisfy the
08000	function of making possible new knowledge through prediction.     The
08100	term  "prediction"  has a spectrum of meanings ranging from forecasts
08200	to prognoses to prophecies to precise point-predictions in time.   To
08300	predict  (and  to  postdict)  from a theory or model is to derive and
08400	announce  a  fact  prior  to  knowledge  of  its  actual  occurrence.
08500	However  one  needs  knowledge  of  the  kind  of  fact expected, the
08600	conditions which produce it and the circumstances under which it will
08700	occur.  The  interest  in  prediction  may  stem from a desire (1) to
08800	confirm or disconfirm a theory or  model  or  (2)  to  obtain  useful
08900	information  about  the future, as in weather forecasting.  Celestial
09000	mechanics provides the  ideal  of  accurate  long-range  predictions.
09100	But  even  astronomers,  with  the advantage of studying isolated and
09200	repetitive systems, have their troubles.    In  1759  Halley's  comet
09300	arrived  four  days later than predicted.    In spite of our advanced
09400	20th century knowledge, a prediction made in 1962 was  off  by  eight
09500	days, that is, twice as bad. (In fairness we must make allowances for
09600	the fact that great masses, distances and  velocities  are  involved.
09700	Also comets defy Newton's law of gravity).
09800		Predictions of individual human behavior are severely limited
09900	by  our  restrictions  of  knowledge.    For  example, (1) sufficient
10000	knowledge of initial conditions may require that we  know  the  whole
10100	past  history of an individual (something not yet achieved for even a
10200	single person), (2) individuals do not remain isolated over the  time
10300	stretch of the prediction; they interact with other individuals of an
10400	unknown nature, (3)  since  life  is  a  fortuitous  flux  of  chance
10500	intersections of independent causal chains, one would also have to be
10600	able to foresee events of the physical environment and  its  changes,
10700	(4)  the  process  of  observation  needed  to obtain information for
10800	predictions may have non-negligible and unforeseeable effects on  the
10900	observed.
11000		In  one  sense  our  paranoid  model  makes  moment-to-moment
11100	predictions and asserts  new  counterfactuals  about  behavior  in  a
11200	psychiatric  interview.  That  is,  if  an  interviewer  says X under
11300	conditions Y, then the model's  response  will  be  characterized  by
11400	z1...zn,   and   the   same   holds   true   for  paranoid  patients.
11500	Counterfactual  prediction  means  that  on  the  basis  of  observed
11600	behavior  we  are willing, with  an  inductive  risk,  to  assume the
11700	presence of unobserved behavior potentials in a model's or  patient's
11800	repertoire of capabilities.
11900		Predicting  new  kinds  of  events  or properties, instead of
12000	kinds we are already familiar with, would represent a genuine  bonus,
12100	indicating  the  model is more than ad hoc and has excess content. It
12200	would give both clinicians and investigators something to  look  for.
12300	This  novelty  could  arise  in  two  ways.    First, the model might
12400	demonstrate a property  of  the  paranoid  mode  hitherto  unobserved
12500	clinically.      In  principle  this could come about because the I-O
12600	behavior of  the  model  is  a  consequence  of  a  large  number  of
12700	interacting  hypotheses  and  assumptions chosen initially to explain
12800	frequently observed phenomena.    When the elements of such a complex
12900	conjunction  interact  with  highly  variable  inputs  they  generate
13000	consequences in addition to those  they  were  designed  to  explain.
13100	Whether  any  of these consequences are significant or characteristic
13200	of the paranoid mode remains a subject for future study.
13300		It  is  also  possible that a new property of paranoia may be
13400	discovered in the clinical  interview,  although  perhaps  everything
13500	that  can  be  said about paranoid dialogues has been said.  If a new
13600	property were found, a search  for  it  might  be  conducted  in  the
13700	model's  behavior. If successful, this again would add to the model's
13800	acceptability.
13900		A second novelty might arise in the behavior of the model  in
14000	some  new situation.   Since it is designed to simulate communicative
14100	behavior in an interview situation, the "new" circumstance would have
14200	to involve some new type of linguistic interaction to which the model
14300	is capable of responding. From its behavior one  might  then  predict
14400	how paranoid patients would behave under similar circumstances.   The
14500	requisite empirical tests and  measures  would  show  the  degree  of
14600	correspondence between patient and model behaviors.
14700		This possibility is of importance in considering emancipatory
14800	therapies for patients  entangled  in  the quandaries of the paranoid
14900	mode.    Since the model operates at a symbol processing level  using
15000	natural  language,  it  is      this  level  at  which linguistic and
15100	conceptual skills of clinicians can be  applied.   Language-based  or
15200	semantic  techniques  do not seem very effective in the psychoses but
15300	they are useful in states of lesser severity. A  wide  range  of  new
15400	semantic  techniques, including extremes, could be tried first on the
15500	model without subjecting patients to blind experimentation.
15600		While we have used the model principally to explore a  theory
15700	and  to study psychiatric judgements, its potential use as a training
15800	device has  not  escaped  our  reflections.    Medical  students  and
15900	psychiatric  residents  need  "disposable  patients"  to  practice on
16000	without jeopardy (to either). A version of  the  paranoid  model  can
16100	display  the  changes  in  its  inner  states  during  an  interview.
16200	Whether  the  optimal  goal  of  interviewing   (gathering   relevant
16300	information  without  upsetting  the patient)  has been achieved, can
16400	thus be estimated.     A  beginning  interviewer  could  practice  in
16500	private or with a supervisor present. Many interviewers have reported
16600	that the model has a definite effect on them.    The student can  get
16700	the  feel  of  the  paranoid mode long before he interviews an actual
16800	patient.     The effect  of  various  interviewing  styles  might  be
16900	studied and compared.
17000	
17100		Although this simulation of  paranoia  covers  a  variety  of
17200	phenomena,  it  is  circumscribed in what it attempts to explain. The
17300	proffered explanation is local and restricted in that it accounts for
17400	only  one  type  of  symbol-processing mode.  Past attempts at grand-
17500	scale explanations of all  mental  processes  in  all  contexts  have
17600	failed.   A  preferable strategy, successful in other sciences, is to
17700	build one circumscribed and tested theory or model at a time so  that
17800	the  field  can  gradually  move  forward a step at a time, each step
17900	gaining consensus before attempting the next.