A PROPOSAL FOR THE DEVELOPMENT OF AUTO-INSTRUCTIONAL MATERIALS IN INFORMATION ANALYSIS AND CLASSIFICATION
Document Type:
Collection:
Document Number (FOIA) /ESDN (CREST):
CIA-RDP67-00896R000100130079-3
Release Decision:
RIPPUB
Original Classification:
K
Document Page Count:
9
Document Creation Date:
December 19, 2016
Sequence Number:
79
Case Number:
Content Type:
REGULATIONS/INSTRUCTIONS/NOTICES/FIELD NOTICES
File:
Attachment | Size |
---|---|
CIA-RDP67-00896R000100130079-3.pdf | 576.48 KB |
Body:
Approved For Release 2006/11/02 : CIA-RDP67-00896R000100130079-3
21Aa1
T&E/365 12 April 1963
A Proposal for the
DEVELOPMENT OF AUTO-INSTRUCTIONAL MATERIALS
IN INFORMATION ANALYSIS AND CLASSIFICATION
In. a retrieval system, information is indexed in the categories outlined by
a descriptor language such as the Intelligence Subject Code. This language, a
list of words and phrases, is used to classify information in categories represent-
ing areas of interest to the user of the system. Usually the categories of the
descriptor language are not determined by logically dividing the subject field,
but by analyzing the type of questions asked of the system and the type of material
indexed in it. For example, a logical analysis might suggest that the descriptor
.language should contain as many subtopics about one political party as another.
However, if the Communist Party were of primary interest to the user of the system,
information about its activities would be classified in more detail. Therefore,
the descriptor language of a retrieval system often seems to be a highly arbitrary
arrangement of topics; one which often conflicts with a common sense classification
of the same subject matter. An early goal of training personnel to index information
in a retrieval system is to have them abandon their own "common sense" concepts of
classification and respond in terms of the categories defined by the descriptor
language.
In addition to the difficulties that arise from the nature of the descriptor
language, a second training problem is caused by the nature of the information to
be indexed. The descriptor language of the system is a finite list of words and
phrases descriptive of the system's categories, but the information to be indexed
is expeessed in an infinite number of ways. Although two documents might be phrased
in completely different words, they could actually mean the same thing. Because
dissimilar documents can fall in the same descriptor language category, classifi-
cation of information in a retrieval system depends on the judgment and the accu-
mulated knowledge of experienced indexers. In spite of the automation of many
retrieval system operations, such as filing and storage, the input process has
remained a human, and consequently, expensive, time-consuming, and often unreliable
operation.
Approved For Release 2006/11/02 : CIA-RDP67-00896R000100130079-3
Approved For Release 2006/11/02 : CIA-RDP67-00896R000100130079-3
Although it is as yet impossible to automate initial classification of infor-
mation in a retrieval system, it is possible to enhance the training of the indi-
viduals who make the input judgments. Already the procedural operations involved
in indexing information for the Intellofax Retrieval System have been taught by
auto-instructional techniques. Although this program dealt to some extent with
both of the training problems mentioned here -- the categories of the descriptor
language and the relationship between document wording and descriptor language
inology -- it emphasized the procedural operations of a specific retrieval system.
To extend the usefulness of these materials and further enhance training, it
is proposed that auto-instructional materials be prepared to train judgment skills
in using a descriptor language, in this case the Intelligence Subject Code, without
reference to any particular retrieval system. This material would give new analysts,
ones who had completed the Intellofax Program, the needed additional practice in
making classification judgments. At present, this practice is provided by a lengthy
combination of experience and training in analyzing actual documents under the close
supervision of experienced indexers. Such supervision is costly and not necessarily
an efficient means for rapidly producing satisfactory classification judgment be-
cause documents are coded as they are received, regardless of their relevancy to
learning new skills. By providing guided and relevant classification experience,
an auto-instructional program can substantially reduce the cost of on-the-job su-
pervision of new personnel and decidedly accelerate the time when an analyst would
be a productive member of his organization.
Since the Intelligence Subject Code is a widely used descriptor language, an
instructional program which developed skill in classifying information according
to its categories is applicable to the training of document analysts working with
either automated or manual retrieval systems. In many of these systems, where
there is a lack of trained supervisory personnel, the replacement of on-the-job
experience by programmed instruction will mean a gain in the accuracy of indexing
and consequently more reliable retrieval. The proposed auto-instructional program
would provide training for both analysts who work with the Intellofax System and
for those who will use the less complex systems which are also based on the In-
telligence Subject Code.
Approved For Release 2006/11/02: CIA-RDP67-00896R000,1,00130079-3
Approved For Release 2006/11/02 : CIA-RDP67-00896R000100130079-3
The terminal behavior to be developed by the proposed program is the ability
to classify a group of typical documents or document summaries into their correct
Intelligence Subject Code categories. Thus, an adequate response in dealing with
a to-be-coded document is the listing of appropriate Intelligence Subject Code
categories in the form of subject codes, subject modifiers, area relationships,
and area codes. This response is, however, only the final act in a long behavioral
chain which begins when the analyst reads the document and ends only after many
responses, which are usually described as "thinking" or "analysis," have occurred.
In this situation, "thinking" implies that the words or context of the document
serve as stimuli for covert responses on the part of the analyst. These responses
are usually the names of some categories in the Intelligence Subject Code. For
example, a document concerning "aluminum" immediately evokes a response on the part
of the analyst of a possible classification category for the document. Other words
in the document such as "research on aluminum," "foreign sales of aluminum," or
"labor force in the aluminum industry" evoke other category responses. As the an-
alyst scans or reads the document, it continues to evoke responses expressed in
the categories of the Intelligence Subject Code. All of these responses, including
the rejection of tentative categories and the substitution of others, occur sub-
vocally and are not observable in the work of the experienced analyst. He may
not even realize that they occur. He "thinks" about the document.or "analyzes"
it. However, these descriptions of the process used to classify documents are of
tie aid to the new analyst because of the nature of the Intelligence Subject
Code and the abundance. of synonyms in the English language.
The proposed auto-instructional program will not only give the student rele-
vant practice in analyzing documents, but it will teach him the analysis process..
It will be designed to teach him to "think" in terms of the Intelligence Subject
Code. This will be done by shaping his responses to key words and phrases in docu-
ments. By breaking down the thought process that occurs. in classifying a document,
the program-will develop his skill in each phase of the process. Then it will
build his ability to perform the entire response chain beginning with the reading
of the document and ending with its correct classification in the descriptor lan-
guage
It is proposed that the research project and the development of the programmed
materials be carried out in the following six steps:
Approved __For,Release _2006/1.1/02 : CIA-RDP67-0089.68000100130079-3
Approved For Release 2006/11/02 : CIA-RDP67-00896R000100130079-3
1. Behavioral analysis. Since the staff of the proposed project is already
familiar with the contents of the Intelligence Subject Code, the initial development
of objectives will emphasize an examination of the actual behavior involved in
classifying documents in the descriptor language. During this period, the sponsor
will collect fifty representative documents and fifty document summaries. Two
skilled analysts will then classify these materials while being interviewed by two
members of the project staff. The interview questions will be designed to make
observable the thought processes involved in classification and to identify the
stimulus dimensions affecting category choice. The interviews will then be recorded.
When they are transcribed they will provide a list of behavioral patterns of skilled
analysts expressed in terms of significant stimulus conditions and appropriate clas-
sification responses (suggested possible categories, indexing rules, rejected cate-
gories, etc.).
The project staff will use this list of behavioral patterns as a guide to clas-
sifying an additional 50 documents and 50 summaries provided by the sponsor. Any
document which cannot be classified by reference to the list of behavior patterns
will be referred to the expert analysts. From this extensive observation of actual
classification behavior, a list of significant stimuli and appropriate responses
will be derived which will form the basic outline of the auto-instructional program.
It is estimated that an experienced analyst can probably classify the 50 documents
and 50 summaries in about three days. An additional day should be allowed to clas-
sify documents which the project staff are unable to categorize.
The collection of the documents and document summaries, their classification
by expert analysts, their transcription, the summary of the behaviors patterns, and
the classification of the additional documents by the project staff will take ap-
proximately one month.
2. Development of the criterion test. The sponsor will select approximately
fifteen documents and fifteen summaries from the two hundred that have been classi-
fied in Step 1. These will be chosen as representing typical and fairly difficult
classification problems. These examples will not be used in the training material,
but will form the criterion test to be given after the program to measure its effec-
tiveness. The time required for developing this test is included in the time allowed
for the first step above.
??P9d.or..elease 2006/11/02: CIA-RDP67-00896R000100130079-3
Approved For Release 2006/11/02 : CIA-RDP67-00896R000100130079-3
3. Development of the draft program. The programmed materials will be developed
from the list. of behavior patterns of skilled analysts described in Step 1 above.
These patterns will be organized into procedural generalizations of the analytic
process. These generalizations will form the outline of the program which will
be designed to teach new trainees to use the same analytic procedures used by skilled
analysts. The success of the program will be determined by the trainee's skill
and accuracy in indexing the fifteen documents and fifteen summaries comprising
the criterion test.
In the process of developing the program, draft versions will be tried out
on appropriately representative subjects. Twelve or more subjects will be admin-
istered the program at this stage in its development, and revisions will be made
in the program after tryout results are obtained from each individual subject or
small group of individuals. The program also will be revised on the basis of stu-
dent performance on the criterion test. The program will be designed so that no
previous experience in indexing is required. This will be made possible by in-
cluding an introductory unit. Students who have worked the Intellofax Program or
have some on-the-job experience with Intelligence Subject Code will be able to skip
this initial sequence and move ahead to advanced materials. By making the program
self-contained, it will be applicable to other indexing systems and usable with
tryout subjects who have not learned from the Intellofax Program.
Since there are few guidlines to use in preparing auto-instructional materials
in "thinking" or "analysis," the present project will, to a large extent, develop
new techniques to teach these skills. The final form of the program cannot be
definitely described at the present time for it will be determined by analysis of
the behavior patterns of the skilled analysts and by experience in student tryouts.
It is estimated that the final program will consist of 500 to 600 frames and take
from seven to ten hours of actual working time to complete. The final program will
be prepared in a format which is suitable for inexpensive reproduction and which
is suitable for general distribution to users of the Intelligence Subject Code.
It is estimated that program development will take approximately three months.
4. Twenty-five copies of the revised draft program will be submitted to the
sponsor for use in training document,.analysts. As each student completes the pro-
gram, he will be given the criterion test to measure the program's effectiveness.
A xcrvv or Re.lease.2.0061.1_1/02-:-CIA.-RD.P67-.00896 R000100130079-3
Approved For Release 2006/11/02 : CIA-RDP67-00896R000100130079-3
-6-
The completed programs and the test results will be returned to the American
Institute for Research for analysis.
5. Final revision. The final revision of the auto-instructional materials
will be based on the data from the first field trial. Probable time for revision
is one month, after which fifty copies of the final revision will be delivered
to the sponsor.
6. Final field trial and report. The sponsor will use the final version of
the program for training their own personnel and for training personnel from other
departments. In addition, copies of the material will be distributed to departments
which wish to train analysts at other locations. It is hoped that these departments
will make available to the sponsor and th both 25X1
response data and criterion test scores. The final report will summarize the stu-
dent data collected in both the first and final field trials. In addition, the
report will contain the list of behavior patterns developed in Step land a descrip-
tion of techniques used to train classification skills.
The proposed time schedule provides that the initial version of the program
will be delivered to the sponsor within four months of the initiation of the pro-
ject. Although it is difficult to determine the time required for the field trials,
it is estimated that the final report will be prepared within ten rnonths.of the
contract date.
Personnel
The successful completion of the proposed project will depend, to a large ex-
tent, on the combination of the experience and competence of the project staff.
The project will be directed b under the supervision of
25X1 25X1 25X1
has directed the current United States Government project con-
cerned with the programming of indexing and document analysis skills for use with 25X1
the Intellofax Retrieval System.l
September 1962. He has had extensive work in directing the writing, editing, re- 25X1
vision, and tryout of programmed materials in many subject areas for all education
levels. He has trained Air Force personnel in programming techniques and is cur-
rently a consultant to Air Training Command on training problems.
.-Approved-For Release 2006/1-4/02: CIA-RDP67-008968000100130079-3
Approved For Release 2006/11/02 : CIA-RDP67-00896R000100130079-3
staff of thel Individuals
s who participate on the project will be chosen
the basis of their relevant experience and training in psychological and educa-
tional research. The large staff of the 0 makes it possible to assign in- 25X1
dividuals to specific projects in such a way as to take maximum advantage of staff
capabilities and to meet the needs of the project as exactly as possible.
Additional research assistance will be provided from among the professional
--Ap ve eFR,edease-2Q06/1 a/02,.:.. CIA-R-DP67-00896RD00.100130079-3
Approved For Release 2006/11/02 : CIA-RDP67-00896R000100130079-3
Budget
This budget is for a period of ten months. The staff will be active on the
project for six months; the other four months are allowed for tryout of materials.
The budget is offered on a cost-reimbursable basis as an extension, or additive
task order, to Contract XG-2527.
Salaries, Wages, and Leave
Executive Research Scientist (part-time)
Research Scientist (1/2 time, 6 months)
Research Assistant (full-time, 6 months)
Secretarial Assistance (1/2 time, 6 months)
Overhead (58.07. of Salaries and Wages)*
Miscellaneous Expenses
.Paid Subject 20 x 10 hours @ $1.60
Communications and Postage
Reproduction of Reports and Materials
25X1
8 days per diem
Subtotal
Fixed-Fee (6% of subtotal)**
TOTAL
*
This is the suggested provisional rate for the period of the contract, and is
subject to redetermination to the actual rate. The cognizant government au ?ting
agency for the 25X1
25X1
This fee is intended to cover such expenses as research staff conferences; pro-
viding a fund for fundamental research essential to our overall program, but
which would not normally be included in specific contracts; providing funds
for the procurement of new scientific equipment and the replacement of obsolete
equipment; and establishing sufficient reserves to meet working capital require-
ments.
-Approved-For Release 2006/11/02 ?CIA-RDP67=00896RO00100130079-3
Approved For Release 2006/11/02 : CIA-RDP67-00896R000100130079-3
Administrative information
The person to be contacted for additional research details with respect to
this proposal is: 25X1
The person to be contacted for additional administrative details with respect
to this proposal is: 25X1
Submitted by,
12 April 1963
-AlprovEd"For Release -2> /11102: CIA-RDP67-00896R000100130079-3