STATEMENT OF(Sanitized), DEPUTY DIRECTOR, OFFICE OF TRAINING, CENTRAL INTELLIGENCE AGENCY

• Document Type: CREST
• Collection: General CIA Records
• Document Number (FOIA) /ESDN (CREST): CIA-RDP81-00896R000100170002-8
• Release Decision: RIPPUB
• Original Classification: S
• Document Page Count: 8
• Document Creation Date: December 12, 2016
• Document Release Date: September 21, 2001
• Sequence Number: 2
• Content Type: TRANS

25X1A
courses for intelligence analysts= a
feel that the training conducted in those courses is
( A rL . 4J~, .-
applicable for indications analysis as! for other types of
intelligence analysis. And I think the statement that is
underlined on page 6 of your report points this out very
nicely.
We offer three different intelligence analysis courses/
.4 AJ
1 1 .e-s-?conduct,"training
Approved For Rise 2001/11/23 :
O W100170002-8
i
10 Director of the Office of Training at CIA, which
i:s,s ecifically for analysts, although there are other
.arza d that p t
P, r~NU6y
trainingt,(courses`` 'e-^'analysts as well. In these courses we
try to cover the very things that I have heard you talking
to the Defense representatives here today. For example, we
cover such things as methodologies that are available to and
help clarify analysis.and "'uo aL have one very
Approved For Release 2001/11/23 : CIA-RDP81-00$96R0Q0100170002-8
is based upon )3a\ analysis, which is an effort to teach
exercise akao run on a computer, a PDP-1145, which.
1
interesting
f3!J.y e `> t r. rs
 UR d'
1" 1
12
Approved For Ruse 2001/11/23 : CIA-RDP81-00896R0g61'100170002-8
people to change the values of their predicti ns based upon
receipt of successive data. This is ;sot.ig.,.tl~a analysts
and I am an old intelligence - nalyst to hang on to the hypothesis that start out with, and
as you get new data, y -do t change yoour hypothesis. Well,
this is an attempt to teach people how to do that.
Another area that we teach m-?n is what is called
subjective probability assessment, to sharpen up their judg-
If I could explain it like this, they are asked'to
make judgments a'beut --- and this is in exercises} factual
place a probability upon the
statements or predictio~andto
likelihood of these things, and then having done so, to give
an indication of the confidence that they have in their own
judgment. mod.ou can train)( people through this technique
14
.17
to improve their judgments and
/Y,0/l /MtM x
the confidence
-
thttiC they have i -; and to do a little bit better job of
doing that and in communicating it to the people that use. the//L.
intelligence product th stems..~sQU> his_.
Finally, in this area, we have worked on the question of
bias, address the question of institutional bias which is
pretty easily identified, unit bias`ifferent parts of
organizations have-different biases,
but above all, the really
difficult one is the bias that resides in each individual. It
13
is harder to identify and it is harder to do something about
and
~&r AS1ePie'S:F&1Mffi
~i e$ 0$ 96 00 100~'~ v0 2
 Approved For
se 2001/11/23 : CIA-RDP81-00896R0QW100170002-8
experience/ perhaps cognitive bias, you might say.
And finally, one of the kinds of things that we do in
25X1A
I /V
hypotheses, tdelress-what. In that connection we have got
these programs -- we have three different courses here -- is
Y.f/ 1dr ~,e .l
in the area of creativity, of helping to t-1 rx1--'dp new
of what makes people thin- (creatively,, 1/ihey have been very
cooperative with us and have given us the benefit of a lot
of their research and of training programs that they have
conducted.
We originally sent people down there for training, and
15
subsequently, one of their principal training programs was
transferred in-house and is conducted now at CIA six times a
year with the assistance of these people.
We have three different courses for different levels of
people. These are all intelligence analysis cours9 and they
25X1A
20 vary in length
11 o Ji a4)t'Ji t A 4J r wi
vr-ryaptzrrt~
21 20.
22
about 300 people a year.
Ie statement said, in
total capacity
of trainingrfo
We do not train pecifically, as the
indications and warning, arn1 we do sind
2d people;over to Captai. course at the Defense Intelligence
A mrd For Release 2001/11/23 : CIA-RDP81-00896R000100170002-8
from two weeks to four weeks, and the numbers
 - ) E b 11YI , 7+ 1.
Approved For Rase 2001/11/23 CIA-RDP81-00896RQQ100170002-8 14
1 ~) Mr. Rose. Do all analysts at the CIA benefit from these
,' course?
25X14 I
3
5
6
9
Mr. Rose. How do you to t them to make sure that they can
I'
l4 do that when you hire them?
25X1A
15 !
Not all analysts have taken the courses,
particularly people who have been around for a while. Some of
these are fairly new programs that have not been under way for
a long time. I guess we would like to think1c e are doing t-h-e.
it3f/P/y ~f~
training, that if trytake them, they would certainly benefit
from them.
I think we start from the point of view that we expect
people to be able to analyze and to think when they are hired.f
That is one of the things that we don't.6kpect to have to teach
them after they get there.
Well, there's a variety of ways. One is we
.16 f do give a variety of psychological tests to them in the first
171 place to establish their IQ. I am not an expert on this, sir,
13 and I can't go into it in great depth. We do a lot of inter-
19 viewing of them at that time by people who are experienced in
20-t th-i-9-'field. We would not rely on test batteries by themselves
21 And we obviously look at their previous. experience, education,
22 academic transcripts, and above all ontact wi=th the people t-1(
they have worked for or in their communities.I think of
Iti4st f < Sa r0,t ~J~eP,~" f1~r~
interest to--u-s-;. contact ,/'basically lookiviwj at their background
"~ a re Cti &ord ledse 8 ~/2$o C D~18~~0 0~ kf1~1 ~ 1 e of
x acrd.
 25X1A Approved For Release 2001/11/23 : CIA-RDP81-00896R000100170002-8
Next 1 Page(s) In Document Exempt
Approved For Release 2001/11/23 : CIA-RDP81-00896R000100170002-8
 9
Approved For Re e 2001/111.23: CIA-RDP81h00896RG4910017Q902-8
Mr. Rose. ou sai you aye ree courses.
Yes, sir.
Mr. Rose. What are the-.titles of those courses? Do you
have them?
Yes, indeed. -.Seminar on Intelligence
Analysis, which runs four times each year for three weeks,
Introduction to Analysis for Career Trainees -- those are
entry level people -- which is two weeks and
7Z, 16,-t / V
izt run four times a year. -h c ae~ag called the
10
.11
Intelligence Process Course, which is conducted twice a year,
four weeks in length.
Mr. Rose. Does the Intelligence Process Course, which
teaches analysts to make finished intelligence meaningful
to executive-level policy makers, does it deal specifically
25X1A 15
with warning?
Yes, but only inl.a very short fashion. There
is not!~much coverage on warning in it.
17
11
Mr. Rose. The National Intelligence Daily, which serves at
{
high-level audience, could be considered to have a warning
.9
-function. Are the articles in the NID evaluated for their
21 it predictive accuracy?
~y
22 25X1 A
25X1A
No, sir, they are not. We ran an experiment
in this, not with the NID because it didn't exist at the
Approved For Release 2001/11/23 CIA-RDP81-00896R000100170002-8
i1
I had better come up to the table.
 Approved For Ruse 2001/11/23 : CIA- - 8 1 1.70002-8 1~ 18
time, but with its predecessor, the National Intelligence
2 Bulletin, and we found that it tas not a fruitful exercise,
25X1A
3 11 that you ran into too many the-sun-will-rise--tomorrow predic-
tions. If the analysts knew we were grading them on this, we
S would-.-..run into a great many more of them, and there were any
6 number of pitfalls in it that made the statistics that we
came up with quite meaningless.
8 Mr. Rose. Some procedures have been developed to
9 reduce or compensate for the effects of perceptual bias in
10 human mental processes, as you mentioned. For example,
Z algebra. The other is a subjective, 1 ~ probability test where ones
22 is able to get instant feedback.
':3 I think these are in that general ballpark. I think they:
24 are very limited and they may be somewhat primitive, but I do
Approved For Re e M1/1 1/23 : CIA-RDP81-00896R000100170002-8
think that t'hie-ft apropos as far as intelligence analysis
20`1 receipt of new data. It is based on Bayes'work, or Bayesian
repetitive training with constant feedback allows airline
12 pilots to compensate for illusions such as things appearing
to be closer on clear days thain0on hazy days.
iy Does the CIA provide any training to its analysts which
l5 will actually help them. overcome cognitive biases?
16 I referred a moment ago, sir, to two extercises
17 which are done real 1i at,._say., a computer terminal keyboard,
S
la ? which provides instant feedback. One of those exercises is
;g 'this business of adjusting one's hypothesis based upon the
 d :J
Approved For Reiewse 2001/11/23: CIA-94 U8URLQ&A01 70002-8
work is concerned.
2 Mr. Rose. All right, thank you very much.
3 j (Whereupon, at 3:14 o'clock p.m., the Subcommittee
4 proceeded to the consideration of other matters.)
Approved For Release 2001/11/23 : CIA-RDP81-00896R000100170002-8
19