Approved For Release 2006h fZUM6-00789ROO2100350011-9 QEC L. I SUN STREAK Ops/Tng Objectives January 1986 January 1988 Operations 900 sessions (4 sources) Utility Assessment 80% (720) Collection 20% (180) Training 450 sessions (2 trainees) SI SHI 66% (300) C! TX7 - 0'17T 33% (150) Approved For Release 2000/08/6;:EIORJZ-IW789ROO2100350011-9 %- I fO Approved For Release 20 sfri "M 96-00789ROO2100350011-9 TARGET CATEGORIES RV Intelligence Utility Assessment 1. (S/SK/WNINTEL) Attached hereto are the target categories selected for examination by RV in support of the planned RV Intelligence Utility Assessment. They have been carefully chosen to represent a broad spectrum of intelligence problems. The intelligence function and production codes of the DIA Register of Intelligence Publications (RIP) have been used so that results from the RV assessment study will be readily understood by intelligence users. Intelligence indicators will be identified for each category as target packages are developed. 2. (S/SK/WNINTEL) Plans are to conduct 20 RV examinations (5 each per source) of each of the 36 selected categories from January 1986 to January 1988 (20 X 36 = 720 sessions) . Care will be taken to maintain appropriate compartmentalization of information relative to actual sites to be RVed. These procedures will guarantee that sources are unburdened by preconceived notions of the sites as well as insure that the information reported by the sources is the product of RV and not more conventional analytic efforts% 3. (SISKIWNINTEL) The RV Intelligence Utility Assessment is not an attempt to scientifically prove or disprove RV through the use of statistical methodology. It is, as stated earlier, an attempt to identify those categories of intelligence targets which would best be serviced by RV. Therefore, each of the 720 PV examinations will be rated for accuracy and overall intelligence value. The 0 to 3 point accuracy rating scale developed by SRI International for GRILL FLAME will be used in conjunction with a 0 to 3 point intelligence value rating: Accuracy Scale 0 - Little correspondence 1 - Site contact with mixed results 2 - Good 3 - Excellent HANDLE VIA SKEET CHANNELS ONLY SPECIAL ACCESS REQUIRED WARNING NOTICE: SENSITIVE CLASSIFIED BY: DIA/DT INTELLIGENCE SOURCES AND DECLASSIFY ON: OADR METHODS INVOLVED Approved For Release 96-00789R002100350011-9 E/C8 REV IrN C-X Approved For Release 2006E8C "F~6-00789ROO2100350011-9 RE Value Scale 0 - No intelligence value 1 - Minimal intelligence value 2 - Moderate intelligence value 3 - High intelligence value The reason for this dual rating is based on the experience of CENTER LANE. Many times a RV session will yield very "accurate" data, but that data will be of little "value" to the intelligence community. The RV Intelligence Utility Assessment will identify those categories of intelligence targets for which both accurate and valuable information can be provided. Approved For Release 2000 00789ROO2100350011-9 AlEOR , 91?- SG1A Approved For Release 2000/08/08 : CIA-RDP96-00789ROO2100350011-9 Next 11 Page(s) In Document Exempt Approved For Release 2000/08/08 : CIA-RDP96-00789ROO2100350011-9 Approved For Release 2000/08/08 : CIA-RDP96-00789ROO2100350011-9 UNCLASSIFIED OVERVIEW OF SRI-INTERNATIONAL EVALUATION METHODS Judging on a Blind Basis In this method, SRI selected an analyst not otherwise associated with the series of sites to be judged to quantitatively assess the degree of correspondence of a given transcript of the remote viewer's data concerning each site and the actual targets. The judge was given a transcript including a list of specific concepts (e.g., "red" or "shady") for each site and was instructed to compare each transcript to each site (i.e., 6 sites = 6 transcripts/concept lists = 36 assessments per concept). Each concept was rated on a scale of 0 to 10 with "0" implying no correspondence between the concept and the target site and "10" implying complete correspondence. Upon completion of the rating, the judge then computed the average rating score for all concepts in each transcript matched against each target, displaying his results in a matrix with target displayed as rows and transcripts displayed as columns. The transcripts were then ranked 1-6 indicating which transcript most closely matched the actual site. This technique, although apparently excellent with regard to demonstrpting the presence or absence of a viable RV function, did not provide a uniform measure from transcript to transcript of the quality of RV functioning. It therefore became necessary to develop an objective method of evaluation. 0-to-7 Point Rating Scale This system was applied "nonblind:, post hoc to the evaluation of the transcripts by the blind-judging method. For no correspondence between the transcript and target site, 0 was assigned; for excellent correspondence, 7; and for intermediate correspondence an intermediate rating. Several methods of "targeting" the viewer were used, and an average of accuracy was obtained for each method. A comparison of the rating produced with this approach against the ratings produced by the blind-judging approach showed statistically significant positive correlation between the two techniques. This, combined with the finding that application of the 0-to-7 point scale to randomly matched transcripts and targets (blind-judging method) yielded chance results, established that application of the 0-to-7 point scale provided a reliable, objective measure of RV quality. UNAPPmedfor Release 2000/08/08 : CIA-RDP96-00789ROO2100350011-9 Try C_3 Approved For Release 2000/08/08 : CIA-RDP96-00789ROO2100350011-9 UNCLASSIFIED 0-to-3 Point Accuracy Rating Scale The FY 1984 SRI Evaluation project further refined RV evaluation procedures by developing the 0-to-3 Point Accuracy Rating Scale. This was used to provide a "bottom-line" assessment of the accuracy/quality of individual elements in a transcript, or of the transcript as a whole. 0 was considered a "miss", progressing throuqh some correspondence (1), good correspondence (2), and excellent correspondence (3). While this reduced the number of factors needed to determine a viewer's accuracy based on objective measures, it still did not meet the criteria for scientific standards. Princeton Evaluation Procedure (PEP) The PEP was based on comparing independently valid quantitatively-defined target information with similarly quantitatively-defined response information. The definition of a particular target site was contained in the yes/no answers to a set of cruestions (30) called descriptors. The only target information to be considered for analysis was completely contained in the yes/no answers. A binary number (30 bits for PEP) was constructed for the target and the response descriptor respectively. YFS. = binary "1" and NO = binary "0". The resulting two 30-bit binary numbers were than compared by a variety of mathematical techniques to form a score for the specific RV session. For a series of RV sessions, a quantitative assessment was made by comparing a given response (matched to its corresponding target site) against the scores computed by matching the response to all other targets used in the series. There were four basic problems with the PEP system: Bit descriptors were not appropriate for the training environment. PEP was not interfaced to a standard data base management system. Cross-target scoring was not sensitive to intelligence requirements. Cross-target scoring was not appropriate for a training environment. This led to the the development of the SRI Evaluation Procedure. UN6pjM~ypgJEpr Release 2000/08/08 : CIA~RDP96-00789RO02100350011-9 Approved For Release 2000/08/08 : CIA-RDP96-00789ROO2100350011-9 UNCLASSIFIED SRI Evaluation Procedure (SEP) As in the PEP, the SRI Evalu.ation Procedure quantified the target material into binary numbers corresponding to yes/no answers to a set of descriptors. Descriptor lists and coding techniques were devised for each RV response for analysis. However, it modified the PEP by using two different types of analysis -- the target-pool-dependent analysis and the target-pool-independent analysis. The target-pool-dependent analysis measured the uniqueness of the target/response match compared with the remainder of the target pool. The target-pool-independent analysis measured the accuracy and reliability of a single RV response matched only against the target material used in the session. Both methods of analysis met the need for a system scientifically measuring RV ability and accuracy, but they did not address the real world of operational intelligence collection. Intelligence_.Ratings This system rates 6 descript~orq_- ranging from major gestalt to ,articular site reie-v-a-nt information on a scale of 0 to 3, 0 .P representing little correspondence to the site and 3 .representing excellent correspondence to the site. Unknown and non-applicable factors are also considered, but not scored. Information perceived by the viewer is then rated for overall intelligence utility, ranging from no utility to very useful, or in some instances, indeterminable at the time of analysis. This procedure has not been extensively applied because there have been relatively few operational taskings since it's development. The next milestone will be the development of a data- base management system that will provide the following: Storage of operational RV session evaluations. Analytic information concerning RV effectiveness when targeted against a variety of intelligence collection tasks. Identification of reliable RV sources for each collection requirement. UNCW$'p'rbRVWFor Release 2000/08/08 : CIA-RDP96-00789ROO2100350011-9 Approved For Release,4000/08/Ga CIA-RDP96-00789ROO2100350011-9 DATABASE MANAGEMENT RV Intelligence Utility Assessment 1. (S/SK/WNINTEL) The RV Intelligence Utility Assessment will be expected to provide us with data which can then be manipulated to provide information on: a. Which categories of targets lend themselves best to remote viewing intelligence collection. b. The accuracy of each RV source. C. The accuracy, dependability, and saleability of the RV process in the intelligence collection arena. 2. (S/SK/WNINTEL) In order to properly assess the results of the RV Intelligence Utility Package, we must first establish a database which will track specific items of crucial interest. (See Fig. 1) These include, but are not limited to: a. Intelligence Utility Function and Production codes: Target categories which have been selected for examination by remote viewing. b. Specific sub category of tasking: Specific category of interest to the RV process: target location, specific details, presence/absence of person/thing at site, etc. C. Job number and job numbers of related tasking: For filing, reporting, and data management purposes. d. Source ID (number or code): Tracking the results of each source will not only indicate the personal abilities of each source, but will allow a diagnosis of that source's tendencies towards greater accuracy with differing target types. e. Date and Time of session: Recent ELF studies indicate that there is a positive correlation between sunspot HANDLE VIA SKEET CHANNELS ONLY SPECIAL ACCESS REQUIRED WARNING NOTICE: SENSITIVE CLASSIFIED BY: DIA/DT INTELLIGENCE SOURCES AND DECLASSIFY ON: OADR METHODS INVOLVED Approved For Release 2000/08/08 CIA-RDP96-00789ROO2100350011-9 Approved For Release 2 Q0 : 0,ARDP96-00789ROO2100350011-9 00W9W activity and dependability of RV collected information. Positive prediction of best RV session times will allow for the most efficient planning of work schedules. f. Session number within a tasked series: Since RV results are statistical in nature, more than one Source and more than one session is run against each task. This field will allow for more refined accuracy in analysis. g. General and specific tasking: A short explanation of tasked activity. h. List of information known at beqinninq of RV work: In order to allow an assessment of the necessary information requirements to be made of future tasking agencies. i. 'List of information generated during RV session:A list of the main information items returned to the tasking agency. e. Accuracy of results: An assessment of the accuracy of generated information, as made by the tasking agency. g. Value of results: An assessment of the intelligence value of the generated information, as made by the tasking agency. Analysis of this factor will provide the only-reii-able means of predicting the efficiency and usefulness to future tasking agencies. It will be the only accurate vehicle for generating credibility in and saleability of our product. 3. (S/SK/WNINTEL) Once the data base has qrown to sufficient proportion to have statistically analytic value, questions which are statistical in nature and data-oriented can be tasked and answered. An example of these -questions is presented in Fig. 2, below. Approved For Release 2000i0tiot ''lc' lk.,,R,--,DP96 -00789 R0021 00350011 -9 Approved For Release 20 MW 96-00789ROO2100350011-9 18F.C. rN F.T FIELD Tasking data: Function code: PGT Prod. code 2630 Specific cat.: Location Job number 86-062 Related jobs 86-078 86-112 Session data: Source ID 132 Date 860312 Time 1302 hrs Sess. number 003 Tasking/site data: General task : RV hostaqe situation Specific task: Hostage location EEI State of health fTar~-etinq inirormaticn) 4 SM) (Intornat'l terror.L (Peacch problem) (Fvcoded for security) G i v (_:~ Hostage name TInformation provided n by tasker Hostage DOB Ly tasking agency) Hostage photo Last known location Results Alive as of 860312 (Information provided from RV Well-fed to tasking aqency) V Desert town Radio station antenna to west Airport to east Evaluation session: of Accuracy 3 (Assessment made by Value 2 tasking ag ency) FIGURE 1. FXAMPLE OF DATABASE RECOPD SECRET Approved For Release 2000/08/08 : CIA-RDP96-00789ROO2100350011-9 SG1A Approved For Release 2000/08/08 : CIA-RDP96-00789ROO2100350011-9 Next 3 Page(s) In Document Exempt Approved For Release 2000/08/08 : CIA-RDP96-00789ROO2100350011-9 Approved For Release 2000/08/,08 -,CIA~-RDP96-00789ROO2100350011-9 SUN STREAK OPS/TNG OBJECTIVES JANUARY 1986 JANUARY 1988 OPERATIONS 900 SESSIONS (4 SOURCES) UTILITY ASSESSMENT 80% (720) COLLECTION 20% (1870) TRAINING 450 SESSIONS (2 TRAINEES) Si Sill 66% (300) Siv - Svi 33% (150) Approved For Release 2000 Pl~~'Pi5P96-00789ROO2100350011-9 Approved For Release 20LECE ~'X-96-00789RO02100350011-9 SUN STREAK OPERATION PLANNING JANUARY i~bb - JANUARY 198~ GIVEN: .1. FOUR FkAINED SOURCES. 2. EACH SOURCE CAPABLE OF ~'OING FOUR SESSIONS "Ek WEEK NORKING A TWO WEEK ON/ONE W'EEK OFF PATTERN. CAPABILITY: 530 SESSIONS PER YEAR, -15% FOR LEAVES, SICKNESS, AND PERSONAL APPOINTMENTS. 45U SESSIONS PER YEAR. 10 TWO YEAR PLAN: 900 SESSIONS 80% OF CAPABILITY DEVOTED TO UTILITY ASSESSMENT, 20% OF CAPABILITY FOR QRf AND RESPONSE TO COLLECTIO11 REQUIREMENTS. Q. r '.' R i T Approved For Release 2 &- P96-00789ROO2100350011-9