SEMINAR ON EVALUATION OF THE INTELLIGENCE PRODUCT
Document Type:
Collection:
Document Number (FOIA) /ESDN (CREST):
CIA-RDP83M00171R001500070003-2
Release Decision:
RIPPUB
Original Classification:
C
Document Page Count:
13
Document Creation Date:
December 20, 2016
Document Release Date:
June 2, 2006
Sequence Number:
3
Case Number:
Publication Date:
April 21, 1980
Content Type:
REPORT
File:
Attachment | Size |
---|---|
![]() | 771.71 KB |
Body:
Approved For Rel se 2006/06/02 CIA-RDP83M00171 R001500070003-2
Seminar on Evaluation of the Intelligence Product
Center for the Study of Intelligence
Central Intelligence Agency
Washington, D.C. 20505
21 April 1980
MORI/CDF
Approved For Release 2006/06/02 : CIA-RDP83M00171 R00150
CONFIDENTIAL
C&itic ,sm .shou.Pd not be quenu,Louus and
wa4.i.ng, aU knise and toot-puUe , but
guiding, insttc.uc i.ve, inspi ng, a south
wind, not an east wind.
--Eme}uson, Jouhnatz
SEMINAR REPORT: EVALUATING THE INTELLIGENCE PRODUCT
Evaluating finished intelligence is not a new subject for
discussion among intelligence managers. Three major reports by
intelligence officers and outside groups written within the past
three years have characterized the absence of an evaluative
mechanism as a serious deficiency in the intelligence structure.
The studies disagreed on where the responsibility for this evalua-
tion should be assigned, but there was agreement that some method
of evaluation was needed to judge the utility and quality of
analysis, to assess the impact of intelligence on consumers, and
to provide lessons for intelligence managers to guide future
production.
On April 21, 1980, a group of 30 people--intelligence pro-
fessionals and Congressional staff officers--met under the auspices
of the Center for the Study of Intelligence to share their views
and to exchange ideas on the.evaluation of intelligence products.
The agenda included discussion of three major aspects of the issue:
1. Is there a continuing need for a systematic evaluation
of finished intelligence products? If so, who should be responsible
for it?
2. How should judgments about utility and quality be made?
Can the development of an adversary relationship between the intelli-
gence producers and judges be avoided?
3. How can the lessons learned from an evaluation process be
translated into guidelines for future production?
A Need to Institutionalize Intelligence Evaluation
In a December 1978 meeting held to discuss the evaluation of
human-source, intelligence collection, a group of intelligence collec-
tion managers concluded that it was imperative to establish a
systematic means for evaluating intelligence collection methods and
results. This experience suggests that production elements might
Approved For Release 2006/06/02 ': CIA-RDP83M00171 RQp1500070003-2
10
also have to devise an evaluative mechanism for their products.
One seminar participant commented that the "choice" apparently
made in the past by producers has been simply to ignore the whole
issue. Only recently, he felt, has the Intelligence Community
begun seriously to examine the feasibility of such a mechanism.
The necessity for and the desirability of developing some
procedure for evaluation was acknowledged by the participants.
Questions arose, however, in discussing what should be evaluated
and by whom it?should be done. To clarify thinking, it was
suggest hat the issue be visualized in terms of a three-sided
figure:
- --First, one must consider carefully what is to be evaluated.
There is in reality not one but a whole spectrum of "whats"
ranging from the analytic process itself to the publication of a
piece of finished intelligence. Whether referring to the product
or the process, it is important to distinguish the type of intelligence
being considered--national or departmental, current or. long-
range.
--Second, the criteria used to judge the product must be
clearly defined. I one is using direct feedback from the consumer
to judge product responsiveness to consumer needs, then one must
be aware of the cognitive problems associated with the users'
perceptions of their requirements. If one is using objective
norms of sound analysis, then such norms must be identified,
codified and agreed upon by the production judges.
--Third, it must be clear who is to perform the product
evaluation, whether that be the agency producing the product, an
outside group, the DCI staff or CIA's Senior Review Panel.
While there may be different perspectives associated with
each of these three elements, it was felt that a combination of
the three would provide a more comprehensive way of thinking
about the problems related to evaluation.
One participant felt that it was necessary to review the
criticism leveled at the Intelligence Community in regard to
product evaluation. He suggested that there are three possible
responses to such criticism:
--We can attempt to placate the critics by pointing out that
analysis is a creative process and that we are doing as well as
can be expected by bureaucratic standards.
-We can take the criticism to heart and strive to develop
a mechanism for assessing performance, if only by providing a
means of customer feedback.
Approved For Release 2006/06/02: CIA-RDP83MOOl 71 R001500070003-2
Q.
--We can develop some method for providing the top management
with a data base needed to judge how well it is doing and for
helping it to use the resources of the community more effectively.
There was some debate as to the cost-effectiveness of these
responses. One discussant felt that whatever evaluative system
is developed will use up considerable manpower over time. We
should try, therefore, to satisfy as many of these objectives as
we can simultaneously. Another felt that while these objectives
were not mutually exclusive, the mechanisms needed to deal with
them may be quite different. Nevertheless, against the whole
cost of the community's effort he argued that the cost of additional
staff to provide product evaluation would be trivial. Finally, a
third participant questioned the pertinence of this debate on cost.
We must take a closer look at how manpower would be used before
dealing with the cost problem in more detail.
Internal Versus External Evaluation Mechanism
Having agreed generally that a formal evaluation mechanism
for the intelligence product was needed, the discussion narrowed to
the question: To whom should this responsibility be given? Earlier
studies have usually avoided specific recommendations in this regard.
This prompted a participant from DIA to cite his own experiences
in describing one workable internal evaluative format. He acknow-
ledged his distrust of any outside evaluation of intelligence
estimates, especially when undertaken after a period of time had
elapsed. Such an external evaluating group may not see or be aware
of all the factors or data used at the time the estimate was
written. In his opinion, it would be more useful to have a National
Intelligence Officer bring together the original author(s) and any
others who had taken part in the discussion and final approval of the
draft. They could then go back and see where and why they were right
or wrong as a group and review how the final collective analysis was
made. An internal process such as this would not only be useful and
stimulating but would also provide valuable insights for updating
the existing estimates.
There is at least one precedent for an external as well as an
internal evaluative mechanism. In the mid-1970s the Intelligence
Community Staff (ICS) had as one of its functions the current
evaluation of interagency products, although this later developed
into a postmortem process before falling into disuse., When the
Director, NFAC, insisted in 1977 that the evaluative function
belonged within NFAC, the ICS dropped its efforts to revive the
mechanism. This recollection prompted one discussant to raise the
question of the role and charter of the NFAC Senior Review Panel
to judge intelligence products. One member of the panel present
.Approved For Release 2006/06/02 : CIA-RDP83M00171 R001500070003-2 -
?
I
CONFIDENTIAL
at the seminar commented that it was still largely in an experi-
mental stage; an assessment of its performance and effects of its
Judgments are not yet complete. He stressed, however, that the
Senior Review Panel was not a replacement for the ICS's community-
wide evaluation system.
Perhaps the most pertinent question to be considered in any
discussion of an internal evaluation mechanism.is whether or not
the people who-had been actively involved in the production of the
product are capable of judging it with sufficient objectivity.
Several participants thought that producers could indeed be accurate
judges. DIA has developed what it believes to be a useful format
for product evaluation as an internal process.' The DoD policy
relating to product evaluation obligates producers to carry out
systematic evaluations of their own products. A summarization of
the results is used as an input to future management and budget
programs. While perhaps less objective than an academic effort
performed in the abstract, the DIA system nevertheless offers a
means of improving the intelligence product within the bounds of
DoD resources.
Like artists, intelligence analysts are highly sensitive to
any criticism of their work. Reviewers must respect this while
persisting in their efforts to render constructive judgments. The
consensus of those present was that the product of an independent
review would not be as helpful to intelligence managers as one
resulting from an internal evaluation. 'The ideal solution would
be a combination of the two--a mechanism which reflected a know-
ledge and understanding of the intelligence system in combination
with a perceived ability to judge the utility of its-product without
bias. Some participants felt that a group of experienced insiders--
if they could develop the detached independence associated with
an external group--might be accepted as professional evaluators.
Thus, we could have our cake and eat it too.
Several participants pointed out that a form of evaluation
automatically takes place in many areas of production as supervisors
of analysts critique the paper while it is being produced. Some
felt that the best evaluation process is one that occurs before
publication of the product. Formulating the terms of reference and
participating in the review and coordination process should provide
an on-going evaluation of the quality of analysis. If one waits for
a year or more before doing an evaluation, the intelligence judgments
may be seen as incorrect in the light of subsequent information, even
though the analysis may have been as accurate as possible when it was
written.
There exists a general sense that an evaluative process which is
not imposed from above is essentially useless; there must be a
Approved For Release 2006/06/02 -
CONFIDENTIAL
commitment from the top in order to have any evaluation mechanism
function effectively. If the evaluation is performed by an inside
group and accepted by production managers, then its results may
?prove to be influential on future intelligence products. If the
evaluation is done by an external group which exerts no real
authority or direction on the production elements, then it will be
easier for intelligence managers to ignore the evaluative
recommendations.
The principal problem is not the potential adversary relation-
ship between intelligence producers and evaluators who judge the
products, but rather involves bureaucratic managerial resistance..
Any outside evaluation procedure will be viewed as an intrusion
on the prerogatives of intelligence managers, who want to keep
control of their product. One participant remarked that the
interviews he conducted on this subject encountered little resis-
tance until he reached the higher levels. One intelligence official
told him in effect, "I don't want my people to read about their
work; I want them to work." Another senior officer in the CIA
was quoted as saying that we ought not to spend so much time
worrying about where we have? been, but where we are going.
Consumer Judgment
Assuming that some type of evaluative mechanism can be estab-
lished, to what degree should it rely on consumer reaction? Several
participants felt that policy. consumers--the ultimate recipients of
intelligence information--were the ones most able to judge the
usefulness of intelligence products. The people in the best position
to judge the quality and utility of an IIM on naval readiness, for
example, would be the fleet commanders and the program officers and
implementers. Another discussant, however, argued by analogy that
the quality of medical care cannot be judged fairly by the patients
receiving it, nor can consumers judge intelligence products without
falling into self-serving distortions.
Since intelligence is ultimately the servant of the policy maker,
consumers ought to be consulted in any evaluation effort. If one
begins tinkering with an intelligence evaluation procedure, consumers
should be brought into the process in a more relevant way than the
rather ad hoc systems of the past. This is not meant to imply,
however, that consumer reaction should provide the sole evaluative
input.
There are times when excessive reliance on consumer reaction to
the usefulness of intelligence may prove to be misleading. In
interviewing consumers during the process of individual postmortems,
Approved For Rele se 2006/06/02: CIA-RDP83M00171 R001500070003-2
O
CONFIDENTIAL
5
nnnrnair> - ri, -
Approved For Rel se 2006/06/02 : CIA-RDP83M00171 RC~1500070003-2
CONFIDENTIAL
one intelligence officer uncovered responses such as "I didn't
read it" or "I read it but I didn't believe it" when asking about
intelligence product utility. Another common reaction to intelli-
gence analysis--probably resulting from a simultaneous reading of
common mail--was "I already know that." The merging of intelligence
into policy formulation at an early stage also makes consumer
responsiveness more difficult to weigh. What some policy makers
see may not be intelligence per se, but some "middle" product in
which intelligence inputs have been buried.
Judging Utility
One participant sensed a certain amount of "hand-wringing"
in this seminar in regard to the assessment of`consumer needs and
the ability of consumers to articulate those needs, especially in
the area of such "soft" measures as quality, timeliness and
accuracy. From his experiences with consumer surveys and ques-
tionnaires, he found that intelligence can and does get useful
feedback from its consumers. But surveys must ask specific,
detailed questions about particular products--not a vague "How do
you like...?" which may elicit such broad responses as "I like
it, it's free."
The experience of DIA has also shown that an internal
intelligence evaluation mechanism may usefully apply commercial
marketing techniques such as bartering or threats of product
elimination. For example, consumers were not at all reluctant to
articulate their needs when they were informed that only three of
five particular services would remain operable and were asked to
list their preferences. Although DIA has gone far in the develop-
ment and use of such management techniques, others in the community
are not aware of these efforts.
There was some discussion on whether it was possible to make
judgments on intelligence without looking at the larger policy.
framework. Some participants thought that it was possible to get
adequate feedback on whether intelligence was used, how it was
used and under'what circumstances it was used. Others thought
that such replies tended to be too simplistic. Consumer response
to the question "Did you use intelligence" has little to do with
the quality of the intelligence product, for there are many
reasons why they may not have used it. How the utility of intelli-
gence is judged depends strongly on which consumers are being
asked. This is a problem of consumer judgment--one must be aware
of the dangers in asking the wrong consumer whether or not a
particular report was worthwhile. There must be some effort at a
more systematic surveying of reactions across the entire known
community of consumers.
Approved For Release 2006/06/02: CIA-RDP83M00171 R001500070003-2
?
CONFIDENTIAL
Much time and effort is expended in asking consumers about the
utility of the intelligence product. But what is really meant by
"useful," and useful to whom? How can product utility be effec-
'tive?ly measured in terms of its degree of influence on policy -
action or its relative impact on the perceptions of policy makers?
Is an evaluation exercise designed only to find out how well we are
or are not doing?
It was sug.gested that there are in fact two important aspects
of this problem which must be clarified. First, the quality of
intelligence and its relevance or utility are not uniquely related
to resources, although allocation of resources is important to
quality. One can make an effective argument that a reduction in
resources could make an overall improvement in the intelligence
product. Increasing available resources also increases the tendency
to depend on someone else to do the necessary work. Working groups
whose relationship to themselves and the issues is uncertain will
result in products of limited quality. The resource-quality
relationship must be addressed as a separate issue.
Second, one must be more specific when examining the consumers
of interagency products. Members of CIA represent only one agency
opinion, and trying to evaluate interagency products from an
exclusively NFAC perspective results in a narrow, limited view
of the problem and its solutions. The objectives and principal
concerns of DIA, the State Department and the individual service
branches may be quite different. These underlying interagency
differences must be taken into account in evaluating the utility
of the finished intelligence product.
Evaluating Quality
On the issue of quality improvement, several questions were
raised. How do we know when we have succeeded in improving the
intelligence product? What does "better" intelligence mean? Is
there some intrinsic attribute divorced from the question of
how well the consumer likes the product which can be used as a
guideline? How should an evaluative process make quality judg-
ments about intelligence?
At a recent symposium, Professor Richard Pipes of Harvard
remarked that he would have given one intelligence estimate an
"F" if it had been turned in to him by one of his students because
of its lack of "academic rigor." Is the equating of intelligence
products with academic research, however, really a.fair way to
judge intelligence? Perhaps the difficulty lies in part in the
fact that quality and utility judgments at times may be at variance
with each other.
CONFIDENTIAL
nNN1 vvcv 1 vi i xci c wvvr vvr v~ vin i'vi vviviv J i r i i i vvvv r vvvv
Presently, there exists no generally accepted objective
criteria to judge intelligence. Subjective judgments of quality
such as "good" or "bad" will inevitably be contested by someone.
If we can define approximate standards for questioning the
utility of an intelligence product, we can also set questions
concerning quality:
--What are the assumptions being made?
--Is the analysis reasonable? Rigorous?
--Is the product timely and well written?
The process of evaluating "quality" is itself rather subjec-
tive in nature, analogous to judging a wine-tasting contest or a
figure skating event. For evaluating intelligence products, the
lack of objective guidelines is a hindrance to any group judgment.
This need not, however, invalidate the notion of developing a
systematized judgment mechanism. Evaluations made under such
circumstances are a common kind of experience. In the academic
community, for example, the evaluation of prospective students
and of student papers does not always follow objective criteria,
but involves personal judgments. Yet there is no wide margin of
disagreement about what constitutes a satisfactory university paper.
There was some discussion on the applicability of a right/
wrong dichotomy to act as a criterion for judging intelligence
products. One discussant felt that most consumers had built-in
appraisal or evaluation systems which work faster than any which
could be set up. For example, a person who uses intelligence
will let the producer know if he thinks the analysis is wrong.
The problem with this is that even if an internal evaluative unit
is set up and evaluates a product as "good", a user may shake his
head in resignation and say the evaluation is as "wrong" as the
product.
Other participants felt that "wrongness" was not a proper
standard for judgment. Can situations ever be depicted in such
black-and-white terms? For example, in 1973, the Intelligence
Community said that the Arabs would definitely not attack Israel.
While the community judgment was wrong, the more significant
question should be why was it wrong. Such issues should not be
dealt with simplistically. One can be right for the wrong reasons
and, particularly when deception is involved, draw a sound but
incorrect conclusion from the data available. A right/wrong
dichotomy, it was judged, is an inadequate framework for evaluating
intelligence components and their methods and products.
Approved For Release 2006/06/02: CIA-RDP83M00171 R041500070003-2
CONFIDENTIAL
One participant concluded the discussion on quality judgments
by suggesting that what was needed was a reasonable set of attributes
against which each product could be judged. Evaluators could
then be more confident of the results and intelligence managers
more comfortable with using this feedback. An acceptable evaluation
process may have one or all of the following benefits:
--justify budget and resource allocations;
--stimulate closer communication between intelligence
producer and consumer;
--improve product quality.
Postmortem Judgments
Although most of those present at the seminar felt that a
serious new effort at evaluating intelligence products should be
made, many believed that any process considered should be selective.
If we should try to evaluate production from a day-to-day perspective,
we will become bogged down by sheer. volume. It was generally
agreed that any evaluation process should be particular in scope,
dealing with specific issues or topics concerning Intelligence
Community services or failings. One must then be careful, however,
not to generalize too much from these specific cases.
For all the objectives about the unfairness implicit in a
postmortem's perfect hindsight, the passage of time does bring
useful new perspectives. Postmortems can produce fair judgments
when intelligence does everything right and still comes up with
the wrong answer; they can also show when intelligence has "lucked
out" for all the wrong reasons. There are lessons to be learned
in either case.
While the postmortem function of the Intelligence Community
Staff was potentially beneficial when the lessons learned were
internalized by the production elements, the process tended to
stimulate defensive attitudes rather than critical or creative
thoughts. Some seminar participants were bothered by the automatic
assumption of postmortem writers that something went wrong with
the intelligence system.
Another criticism was the lack of a postmortem process to
address intelligence successes. It would seem reasonable to
think that there are lessons to be learned from successful intelligence
estimates.
Evaluation of Collection
Since intelligence collectors have already established a
methodology for evaluating their services, the seminar discussion
turned its attention to the experiences of collectors in order to
see if any parallels could be drawn for production. According to
one intelligence-collector present, there is a continuum along
which one must operate. At one extreme is quality measurement
which involves looking at the product against certain accepted
criteria. At the other extreme is a measurement of cost effectiveness
which involves relating results to initial objectives and actual
resource input. The collection evaluation process is trying to
address itself to a wide range across this spectrum. Now the
collection community is trying to apply the lessons it has learned
from this evaluation process, both in terms of substance as well
as collection management.
In the opinion of one collector, any evaluation system ought
to be broad enough to touch both extremes of this continuum.
Another discussant, playing the role of a devil's advocate, saw
potential problems in the system and observed that some collectors
may be reluctant to report interesting but unevaluatable informa-
tion from the field out of fear that the system will give them a
"poor grade."
A Critique and a Conclusion
.One participant summed up his opinions on the feasibility of
an evaluative mechanism and the application of its findings as
follows:
First, he sensed that some of the other seminar members were
quite negative about using consumer reactions to intelligence
products as a basis for judging quality or utility. He disagreed
that consumers lacked the ability to articulate these concerns.
Where else would intelligence find out about the effectiveness of
its products if not from its users?
Second, people around the community have said that establishing
an evaluative system would be too demanding of resources. In his
view, less than 1% of existing resources would be required. Is
this too much to ask for quality control? If the problem is too
tough to address all at once, we ought to work at it incrementally.
It seemed to him that not much was going on as far as production
evaluation was concerned. This made him wonder if intelligence
managers really wanted to establish an evaluation system.
Nevertheless, the general feeling among the participants was
that there is a whole series of things we can do to make judgments
concerning intelligence products. We can look at the nature of
CONFIDENTIAL
10
AnnrnvaN Fnr RaI ca 2flfRlflRifl9 ? flIA_RfPR'ARAnnJ71Rfl(11Ffnnn7fn(1m_9
Approved For Release 2006/06/02 : CIA-RDP83M00171 8001 500070003-2
the product by itself, the rigor of its analysis, its assumptions
and resulting statements; we can look at the atmosphere or environ-
ment in which it was produced; we can examine consumer reaction to
the final product. It was also generally agreed that looking at the
stream of production rather than its individual products is more
likely to prove useful. The major unresolved issue remains: who
should perform the evaluation of the product?
To establish an evaluation system, the process must be started
with people who*are in a position to apply the lessons learned. Any
evaluative mechanism must be designed so that analysts and managers
can see something beneficial in it for them. Commitment to evalua-
tion from the top levels is important, but simply imposing it from
above will not work. Perhaps a good starting point would be to
survey how intelligence managers look at product evaluation.
In the'end, most managers are in the same position. If they
cannot demonstrate that they are producing something of value, they
will lose in the competition for production resources. Managers must
show that they are spending their resources effectively. Results
from an accepted evaluation program-would provide inputs for budget
considerations.
Approved For Release 2006/06/02 : CIA-RDP83M00171 R001500070003-2
Approved For Release 2006/06/02 : CIA-RDP83M00171 R001500070003-2