- This paper discusses different biases which have been identified in Intelligence Analysis and how TIACRITIS, a knowledge-based cognitive assistant for evidence-based hypotheses analysis, can help recognize and partially counter them. After reviewing the architecture of TIACRITIS, the paper shows how it helps recognize and counter many of the analysts' biases in the evaluation of evidence, in the perception of cause and effect, in the estimation of probabilities, and in the retrospective evaluation of intelligence reports. Then the paper introduces three other types of bias that are rarely discussed, biases of the sources of testimonial evidence, biases in the chain of custody of evidence, and biases of the consumers of intelligence, which can also be recognized and countered with TIACRITIS.
Bias, cognitive assistant, intelligence analysis, evidence-based reasoning, argumentation, symbolic probabilities.
I.
Intelligence analysts face the difficult task of drawing
defensible and persuasive conclusions from masses of
evidence, requiring the development of often stunningly
complex arguments that establish and defend the three major
credentials of evidence: relevance, believability, and inferential
force [
In this paper we propose an approach to the identification
and countering of the biases in intelligence analysis. The
approach is based on the observation that the best protection
against biases comes from the collaborative effort of teams of
analysts, who become skilled in the evidential and
argumentational elements of their tasks, and who are willing to
share their insights with colleagues, who are also willing to
listen. As we discuss in this paper, this could be achieved by
employing an intelligent analytic tool like TIACRITIS [
In the next section we introduce the architecture of the TIACRITIS cognitive assistant which is based on semantic technologies for knowledge representation, reasoning, and This research was partially supported by the Department of Defense and by George Mason University. The views and conclusions contained in this document are those of the authors and should not be interpreted as necessarily representing the official policies or endorsements, either expressed or implied, of the Department of Defense or the U.S. Government. learning. Then, in Section III, we address the analysts’ biases discussed by Heuer [2, pp.111-171]: biases in the evaluation of evidence, in the perception of cause and effect, in the estimation of probabilities, and in the retrospective evaluation of intelligence reports. After that we address three other origins of bias that are rarely discussed, even though they may be at least as important on occasion as any analysts’ biases.
II.
TIACRITIS is a knowledge-based system that supports an
intelligence analyst in performing evidence-based hypothesis
analysis in the framework of the scientific method. It guides the
analyst to view intelligence analysis as ceaseless discovery of
evidence, hypotheses, and arguments in a non-stationary world,
involving collaborative processes of evidence in search of
hypotheses, hypotheses in search of evidence, and evidentiary
testing of hypotheses [
G
M very likely
H likely
G likely
F very likely E*
H’ G’ no support
F’ likely ••• ••• •••
H likely N En*
likely min almost Q certain min very likely max •••
S very likely Evidence in search of hypotheses
Hypotheses in search of evidence
Evidentiary tests of hypotheses
Fig. 1. Scientific method framework of TIACRITIS.
Let us assume that we have shown that F and G are more likely than their corresponding competing hypotheses. Next we have to assess H, H’, … . To assess H we need additional evidence which is obtained by successively decomposing H into simpler and simpler hypotheses, as shown by the blue tree in the right part of Fig.1. H would be true if G and M would be true. Then M would be true if N, Q, and S would be true. But if N would be true, then we would need to observe evidence En*. So we look for En* and we may or may not find it. This is the process of hypotheses in search of evidence that guides the evidence collection task. Now some of the newly discovered items of evidence (e.g. En*) may trigger new hypotheses, or the refinement of the current hypotheses. Therefore, as indicated at the bottom part of Fig.1, the processes of evidence in search of hypotheses and hypotheses in search of evidence take place at the same time, and in response to one another.
Then we use all the collected evidence to assess the
hypothesis H. This assessment is probabilistic in nature
because the evidence is always incomplete, usually
inconclusive, frequently ambiguous, commonly dissonant, and
has various degrees of believability [
TIACRITIS was developed by first customizing the
Disciple learning agent shell (a general agent building tool [
Knowledge Base Transactional Access
Tutoring &
Testing e c a Iftren ESlicceitnaatrioion r seU Problem la Solving c i rahpG REevBaidasesoenndcineg
Repository Management
Ontology Development Multistrategy
Learning MixedInitiative
Interaction Asynchronous Message-Based Interaction Knowledge Management Domain KB
Domain KB Scenario KB
Scenario KB
Each of these knowledge bases is structured into an
ontology of concepts and a set of general problem solving rules
expressed with these concepts. The rules are learned from
specific examples of reasoning steps, by using the ontology as
a generalization hierarchy [
Learned general rules from the IA KB include those for directly assessing a hypothesis based on evidence. These rules automatically reduce the assessment of a leaf hypothesis, such as Q in Fig.1, to assessments based on favoring and disfavoring evidence and, further down, to the assessment of the relevance and the believability of each item of evidence with respect to Q. Once these assessments are made, they are combined, from bottom-up, to obtain the inferential force of all the items of evidence on Q, which results in the likeliness of Q.
An example of a learned rule is shown in Fig. 4. It is an
ifthen problem reduction rule that expresses how and under what
conditions a generic hypothesis can be reduced to simpler
generic hypotheses. The conditions are represented as
firstorder logical expressions [
The ontology and the rules from the knowledge repository of TIACRITIS allow it to support the analyst in formulating hypotheses, developing arguments that reduce complex hypotheses to simpler and simpler ones (as discussed above), collecting evidence relevant to the simplest hypotheses, and finally assessing the relevance, the believability, and the inferential force of evidence, and the likeliness of the hypotheses. Additionally, TIACRITIS continuously learns from the performed analyses.
As discussed in the rest of this paper, TIACRITIS has one additional important capability. It supports the analysts in recognizing and countering many of their biases. Because Heuer has made a detailed and very well-known analysis of biases in intelligence analysis [2, pp.111-171], we follow his classification and identified characteristic of biases to show how TIACRITIS helps recognizing and countering many of them.
III.
BIASES OF THE ANALYST A. Biases in the Evaluation of Evidence
Heuer first mentions vividness of evidence as a necessary criterion for establishing its force. Analysts, like other persons, have preferences for certain kinds of evidence and these preferences can induce biases. In particular, analysts can have a distinct preference for vivid or concrete evidence when less vivid or concrete evidence may be more inferentially valuable. In addition, their personal observations may be over-valued.
First, as discussed in the previous section, the hypothesis in search of evidence phase of the analysis helps identify a wide range of evidentiary needs. For example, the argumentation in Fig. 1 shows that we need evidence relevant to N, evidence relevant to Q, evidence relevant to S, etc. It is unlikely that we would have vivid evidence for each basic hypothesis. So we would be forced to use less vivid evidence as well.
Second, as illustrated by the abstract analysis example in Fig. 5 and discussed in the following, TIACRITIS guides us to assess a simple hypothesis Q by performing a uniform, detailed, and systematic evaluation of each item of evidence, regardless of its “vividness”, helping us be more objective in the evaluation of the force of evidence.
Let us first consider how to assess the probability of Q based only on one item of favoring evidence Ek* (see the bottom of Fig. 5). First notice that we call this likeliness of Q, and not likelihood, because in classic probability theory likelihood is P(Ek*|Q), while here we are interested in P(Q|Ek*), the posterior probability of Q given Ek*. With TIACRITIS, to assess Q based only on Ek*, we have three judgments to make by answering three questions:
The relevance question is: How likely is Q, based only on Ek* and assuming that Ek* is true? If Ek* favors Q, then our answer should be one of the values from “likely” to “certain.” If Ek* is not relevant to Q then our answer should be “no support” because Ek* provides no support for the truthfulness of Q. If, however, Ek* disfavors Q, then it favors the negation c (or complement) of Q, and it should be moved under Q .
The believability question is: How likely is it that Ek* is true? Here the answer should be one of the values from “no support” to “certain.” “Certain” means that we are sure that the event Ek reported in Ek* did indeed happen. “No support” means that Ek* provides us no reason to believe that the event Ek reported in Ek* did happen. For example, we believe that the source of Ek* has lied to us.
The inferential force question is: How likely is Q based only on Ek*? TIACRITIS automatically computes this answer as the minimum of the relevance and believability answers. Indeed, to believe that Q is true based only on Ek*, Ek* should be both relevant to Q and believable.
Q
When we assess a hypothesis Q we may have several items of evidence, some favoring it and some disfavoring it. The favoring evidence is used to assess the likeliness of Q and the c disfavoring evidence to assess the likeliness of Q . Because disfavoring evidence for Q is favoring evidence for Qc, the assessment process for Qc is similar to the assessment for Q.
When we have several items of favoring evidence, we evaluate Q based on each of them (as was explained above), and then we compose the obtained results. This is illustrated in Fig.5 where the assessment of Q based only on Ei* (almost certain) is composed with the assessment of Q based only on Ek* (likely), through the maximum function, to obtain the assessment of Q based only on favoring evidence (almost certain). In this case the use of the maximum function is justified because it is enough to have one item of evidence that is both very relevant and very believable to make us believe that the hypothesis is true.
Let us now assume that Qc based only on disfavoring evidence is “likely.” How should we combine this with the assessment of Q based only on favoring evidence? As shown at the top of Fig.5, TIACRITIS uses an on balance judgment: Because Q is “almost certain” and Qc is “likely,” it concludes that, based on all available evidence, Q is “very likely.”
Heuer also mentions the absence of evidence as another origin of bias. The bias here concerns a failure to consider the degree of completeness of available evidence. Consider again the argumentation from Fig. 1 which decomposes complex hypotheses into simpler sub-hypotheses that are assessed based on evidence. This argumentation structure makes very clear that S is not supported by any evidence. Thus the analyst should lower her confidence in the final conclusion, countering the absence of evidence bias.
The next source of bias mentioned by Heuer is a related
one: oversensitivity to evidence consistency, and not enough
concern about the amount of evidence we have. This kind of
bias can easily manifest when using an analytic tool like
Heuer’s ACH [
According to Heuer [2, pp. 121-122]: “When working with a small but consistent body of evidence, analysts need to consider how representative that evidence is of the total body of potentially available information.” The argumentation from Fig. 1 makes very clear that the available evidence is not representative of all the potentially available information. We have no evidence relevant to S. If we would later find such evidence which would indicate “no support” for S, then the considered argumentation would provide “no support” for the top-level hypothesis H. When faced with sub-hypotheses for which there is no evidence (e.g., S in Fig. 1), TIACRITIS allows the analyst to consider various what-if scenarios, making alternative assumptions with respect to the likeliness of S, and determining their influence on the likeliness of H. This should inform the analyst on how to adjust her confidence in the analytic conclusion, to counter the oversensitivity to evidence consistency bias.
Finally, Heuer lists the persistence of impressions based on
discredited evidence as an origin of bias. If Heuer had written
his book in 2003, he might have used the case of Curveball as a
very good example [
TIACRITIS helps countering this bias by incorporating in the argumentation an explicit analysis of the believability of evidence, especially for key evidence that has a direct influence on the analytic conclusion. When such an evidence item is discredited, specific elements of its analysis are updated, and this leads to the automatic updating of the likeliness of each hypothesis to which it is relevant. For example, as shown in the left hand side of Fig. 6, the believability of the observations performed by a source (such as Curveball) depends on source’s competence and credibility. Moreover, competence depends on access and understandability. Credibility depends on veracity, objectivity, and observational sensitivity under the conditions of observation. Thus, the bias that would result from the persistence of impressions based on discredited evidence is countered in TIACRITIS with a rigorous, detailed and explicit believability analysis.
But there are additional biases in the evaluation of evidence that Heuer does not mention, particularly with respect to establishing the credentials of evidence: relevance, believability, and inferential force or weight. An analyst may Competence of S
Credibility of S min min
min Access of S
Veracity of S confuse the competence of a HUMINT source with his/her credibility. Or, the analyst may focus on the veracity of the source and ignore source’s objectivity and observational sensitivity. Analysts may fail to recognize possible synergisms in convergent evidence, as happened in the 9/11/2001 disaster. Analysts may even overlook evidence having significant inferential force.
As noted by Heuer, analysts seek explanations for the occurrence of events and phenomena. These explanations involve assessments of causes and effects. But biases arise when analysts assign causal relations to those that are actually accidental or random in nature. One related consequence is that analysts often overestimate their ability to predict future events from past events, because there is no causal association between them. One major reason for these biases is that analysts may not have the requisite level of understanding of the kinds and amount of information necessary to infer a dependable causal relationship.
According to Heuer, when feasible, the “increased use of scientific procedures in political, economic, and strategic research is much to be encouraged”, to counter these biases [2, p.128]. Because TIACRITIS makes all the judgments explicit, they can be examined by other analysts to determine whether they contain any mistakes or are incomplete. Because different people have different biases, comparing and debating analyses of the same hypothesis made by different analysts can also help identify individual biases. Finally, as a learning system, TIACRITIS can acquire correct reasoning patterns from expert analysts which can then be used to analyze similar hypotheses.
Now, here is something that can occur in any analysis concerning chains of reasoning. It is always possible that an analyst’s judgment will be termed biased or fallacious, on structural grounds if it is observed that this analyst frequently leaves out important links in his/her chains of reasoning. This is actually a common occurrence since, in fact, there is no such thing as a uniquely correct or perfect argument. Someone can always find alternative arguments to the same hypothesis; what this says is that there may be entirely different inferential routes to the same hypothesis. Another possibility is that someone may find arguments based on the same evidence that lead to different hypotheses. This is precisely why there are trials at law; the prosecution and defense will find different arguments, and tell different stories, from the same body of evidence.
There are different views among probabilists on how to
assess the force of evidence [
Since Heuer only considers numerical probabilities conforming to the Kolmogorov axioms, any biases associated with them (e.g., using the availability rule, the anchoring strategy, expressions of uncertainty, assessing the probability of a scenario) are either irrelevant or not directly applicable to a type of analysis that is based on different probability systems, such as the one performed with TIACRITIS, which is based on the Baconian and Fuzzy probability systems. Indeed, analysts using TIACRITIS never assess any numerical probabilities.
Heuer [2, p.122] mentions coping with evidence of uncertain accuracy as an origin of bias: “The human mind has difficulty coping with complicated probabilistic relationships, so people tend to employ simple rules of thumb that reduce the burden of processing such information. In processing information of uncertain accuracy or reliability, analysts tend to make a simple yes or no decision. If they reject the evidence, they tend to reject it fully, so it plays no further role in their mental calculations. If they accept the evidence, they tend to accept it wholly, ignoring the probabilistic nature of the accuracy or reliability judgment.” He then further notes [2, p.123]: “Analysts must consider many items of evidence with different degrees of accuracy and reliability that are related in complex ways with varying degrees of probability to several potential outcomes. Clearly, one cannot make neat mathematical calculations that take all of these probabilistic relationships into account. In making intuitive judgments, we unconsciously seek shortcuts for sorting through this maze, and these shortcuts involve some degree of ignoring the uncertainty inherent in less-than-perfectly-reliable information. There seems to be little an analyst can do about this, short of breaking the analytical problem down in a way that permits assigning probabilities to individual items of information, and then using a mathematical formula to integrate these separate probability judgments.”
First, as discussed in the previous section, concerning the believability of evidence, there is more than just its accuracy to consider. Second, as discussed above, Heuer only considers the conventional view of probability which, indeed, involves complex probability computations. With TIACRITIS, the analyst does precisely what Heuer imagined that could be done for countering this bias. It breaks a hypothesis into simpler hypotheses (see Fig.1), and assesses the simpler hypotheses based on evidence (see Fig.5). Also, TIACRITIS allows the analyst to express probabilities in words rather than numbers, and to employ simple min/max strategies for assessing the probability of interim and final hypotheses that do not involve any full-scale and precise Bayesian or other methods that would require very large numbers of probability assessments.
There are many places to begin a defense of verbal or fuzzy
probability statements. The most obvious one is law. All of the
forensic standards of proof are given verbally: “beyond
reasonable doubt”; “clear and convincing evidence”, “balance
of probabilities”; “sufficient evidence”, and “probable cause’.
Over the centuries attempts have been made to supply
numerical probability values and ranges for each of these
standards, but none of them have been successful. The reason,
of course, is that every case is unique and rests upon many
subjective and imprecise judgments. Wigmore [
We conclude this discussion by recalling what the
wellknown probabilist Professor Glenn Shafer said years ago [
D. Hindsight Biases in Evaluating Intelligence Reporting
As Heuer notes, analysts often overestimate the accuracy of their past judgments; customers often underestimate how much they have learned from an intelligence report; and persons who conduct post-mortem analysis of an intelligence failure will judge that events were more readily foreseeable than was in fact the case. “The analyst, consumer, and overseer evaluating analytical performance all have one thing in common. They are exercising hindsight. They take their current state of knowledge and compare it with what they or others did or could or should have known before the current knowledge was received. This is in sharp contrast with intelligence estimation, which is an exercise in foresight, and it is the difference between these two modes of thought—hindsight and foresight—that seems to be a source of bias. … After a view has been restructured to assimilate the new information, there is virtually no way to accurately reconstruct the pre-existing mental set.” [2, p.162]
Apparently Heuer did not envision the use of a system like TIACRITIS that keeps track of the performed analysis, what evidence we had, what assumptions we made and what were their justifications, and what was the actual logic of our analytic conclusion. We can now add additional evidence and use our hindsight knowledge to restructure the argumentation and re-evaluate our hypotheses, and we can compare the hindsight analysis with the foresight one. But we will not confuse them. As indicated by Heuer [2, pp.166-167]: “A fundamental question posed in any postmortem investigation of intelligence failure is this: Given the information that was available at the time, should analysts have been able to foresee what was going to happen? Unbiased evaluation of intelligence performance depends upon the ability to provide an unbiased answer to this question.” We suggest that this may be accomplished with a system like TIACRTIS.
IV.
SOME FREQUENTLY OVERLOOKED ORIGINS OF BIAS So much of the discussion of bias in intelligence analysis is directed at intelligence analysts themselves. But we have identified three other origins of bias that are rarely discussed, even though they may be at least as important on occasion as any analysts’ alleged biases. The three other origins of bias we will consider are: (1) persons who provide testimonial evidence about events of interest (i.e. HUMINT sources); (2) other intelligence professionals having varying capabilities who serve as links in what we term “chains of custody” linking the evidence itself, as well as it’s sources, with the users of evidence (i.e. the analysts); and (3) the “consumers” of intelligence analyses (government and military officials who make policy and decisions regarding national security). A. HUMINT Sources
Our concern here is with persons who supply us with
testimonial evidence consisting of reports of events about
matters of interest to us. Heuer [2, p.122] does mention the
“bias on the part of the ultimate source,” but he does not
analyze it. In our work on evidence in a variety of contexts, we
have always been concerned about establishing the
believability of its sources, particularly when they are human
witnesses, sources, or informants [
As discussed above, assessing the credibility of a human source S involves assessing S’s veracity, objectivity, and observational sensitivity. We have to consider that source S can be biased concerning any of these attributes. On veracity, S might prefer to tell us that event E occurred, whether S believed E occurred or not. As an example, an analyst evaluating S’s evidence E* might have evidence about S suggesting that S would tell us that E occurred because S wishes to be the bearer of what S believes we will regard as good news that event E occurred. On objectivity, S might choose to believe that E occurred because it would somehow be in S’s best interests if E did occur. On observational sensitivity, there are various ways that S’s senses could be biased in favor of recording event E; clever forms of deception supply examples.
These three species of bias possible for HUMINT sources must be considered by analysts attempting to assess the credibility of source S and how much weight or force S’s evidence E* should have in the analyst’s inference about whether or not event E did happen. The existence of any of these three biases would have an effect on an analyst’s assessment of the weight or force of S’s report E*. As we know, all assessments of the credibility of evidence rest upon available evidence about its sources. In the case of HUMINT we need ancillary evidence about the veracity, objectivity, and observational sensitivity of its sources. In the process, we have to see whether any such evidence reveals any of the three biases just considered. TIACRITIS supports the analyst in this determination by guiding her to answer specific questions based on ancillary evidence. For instance, the veracity questions considered are shown in Table 1. 3. Exploitation potential? Is this source subject to any significant exploitation by other persons or organizations to provide us this information? 4. Any contradictory or divergent evidence? Is there any evidence that contradicts or conflicts with what the source has reported to us? 5. Any corroborative or confirming evidence? Is there any other evidence that corroborates or confirms this source's report? 6. Veracity concerning collateral details? Are there any contradictions or conflicts in the collateral details provided by this source that reflect the possibility of this source's dishonesty? 7. Source's character? What evidence do we have about this source's character and honesty that bears upon this source's veracity? 8. Reporting record? What does the record show about the truthfulness of this source's previous reports to us? 9. Source expectations about us? Is there any evidence that this source may be reporting events he/she believes we will wish to hear or see? 10. Interview behavior? If this source reported these events to us, what was this source's demeanor and bearing while giving us this report?
Unfortunately, there are other persons, apart from HUMINT sources, whose possible biases need to be carefully considered. We know that analysts make use of an enormous variety of evidence that is not testimonial or HUMINT, but is tangible in nature. Examples include objects, images, sensor records of various sorts, documents, maps, diagrams, charts, and tabled information of various kinds.
But the intelligence analysts only rarely have immediate and first access to HUMINT assets or informants. They may only rarely be the first ones to encounter an item of tangible evidence. What happens is that there are several persons who have access to evidence between the times the evidence is first acquired and when the analysts first receive it. These persons may do a variety of different things to the initial evidence during the time they have access to it. In law, these persons constitute what is termed a “chain of custody” for evidence.
Heuer [2, p.122] mentions the “distortion in the reporting chain from subsource through source, case officer, reports officer, to analyst” but he does not analyze it. In criminal cases in law, there are persons identified as “evidence custodians”, who keep careful track of who discovered an item of evidence, who then had access to it and for how long, and what if anything they did to the evidence when they had access to it.
These chains of custody add three major additional sources
of uncertainty for intelligence analysts to consider, that are
associated with the persons in chains of custody whose
competence and credibility need to be considered. The first and
most important question involves authenticity: Is the evidence
received by an analyst exactly what the initial evidence said
and is it complete? The other questions involve assessing the
reliability and accuracy of the processes used to produce the
evidence if it is tangible in nature (see the right side of Fig. 6),
or also used to take various actions on the evidence in a chain
of custody, whether the evidence is tangible or testimonial. As
an illustration, consider an item of testimonial HUMINT
coming from a foreign national whose code name is
“Wallflower”, who does not speak English [
Now, here is where forms of bias can enter that can be associated with the persons involved in these chains of custody. The case officer Bob might have intentionally overlooked details in his recording of Wallflower’s report. The translator Husam may have intentionally altered or deleted parts of this report. The report’s officer Marsha might have altered or deleted parts of the translated report of Wallflower’s testimony in her editing of it. The result of these actions is that the analyst Clyde receiving this evidence almost certainly did not receive an authentic and complete account of it, nor did he receive a good account of its reliability and accuracy. What he received was the transmitted, edited, translated, recorded testimony of Wallflower. Fig. 7 shows how TIACRITIS may determine the believability of the evidence received by the analyst. Although the information to make such an analysis may not be available, the analyst should adjust the confidence in his conclusion, in recognition of these biases.
Believability of transmitted, edited, translated, recorded testimony of Wallflower
min
Believability of edited translation
Believability of translated recording
Believability of recorded testimony Believability of Wallflower min min min
Believability of editing by Marsha Believability of translation by Husam Believability of recording by Bob
Believability of transmission by Marsha
Fig. 7. Chain of custody of Wallflower’s testimony.
The policy-making consumers or customers of intelligence
analysts are also subject to a variety of inferential and
decisional biases that may influence the reported analytic
conclusions. As is well known, the relationships between
intelligence analysts and governmental policy makers are much
discussed and involve considerable controversy [
V.
CONCLUSIONS
A wide variety of biases affect the correctness of intelligence analyses. In this paper we have shown how the use of TIACRITIS, a knowledge-based cognitive assistant, helps analysts recognize and counter many of them. TIACRITIS integrates several semantic technologies (knowledge representation through ontologies and rules, evidence-based reasoning, machine learning and knowledge acquisition). It can run in a browser as a web-based system, or it can be installed locally, and has been used in many civilian, military, and intelligence organizations.
There are two complementary ways by which TIACRITIS helps mitigate biases. First, as a cognitive assistant, it helps automate many parts of the analysis process, making this task much easier for the analyst. Thus it alleviates one of the main causes of biases, which is the employment of simplified information processing strategies on the part of the analyst. Second, TIACRITIS performs a rigorous evidence-based hypothesis analysis that makes explicit all the reasoning steps, evidence, probabilistic assessments, and assumptions, so that they can be critically analyzed and debated. Indeed, the best protection against biases comes from the collaborative effort of teams of analysts, who become skilled in solving their analytic tasks through the development of sound evidence-based arguments, and who are willing to share their insights with colleagues, who are also willing to listen. TIACRITIS makes all this possible.
Finally, this paper adds a strong argument in favor of using structured analytic methods, in the debate on how significantly improve intelligence analysis [26].