Upload
jeremy-levin
View
237
Download
1
Embed Size (px)
Citation preview
8/3/2019 The Role of Bias in Intelligence Analysis
1/16
The Role of Organizational
and Intelligence DisciplineBias in Intelligence
Analysis, and Structured
Analytic Methods to
Overcome ThisJeremy Levin; Student ID #3049427
INTL 506: Analytics II
7/30/2011
8/3/2019 The Role of Bias in Intelligence Analysis
2/16
In the wake of the 11 September 2001 terrorist attacks against the World Trade Center
and Pentagon--commonly referred to as 9/11--and the faulty intelligence that sent the United
States (US) to war in Iraq, intelligence failure has become one of the most discussed national
security topics in both academia and national security circles. From online blogs devoted to
exposing intelligence dysfunction to the 9/11 Commission Report, every aspect of the
intelligence cycle has been picked apart, and failure attributed to every aspect. A common
thread among much of this criticism is the intelligence stovepiping, or maintenance of
intelligence within an organization rather than disseminating it to the intelligence community
(IC). Most pundits assert this stovepiping was a result of either security concerns--that those
outside the originating organization would leak or spill the intelligence--or born of the desire to
use the withheld reporting make their organizations' analysis more valuable to national decision
makers. Whatever the reason, intelligence organizations were not sharing information, resulting
in several failures to piece together complete intelligence pictures for national leaders.
In 2005, leadership of the intelligence community was taken away from the Director of
Central Intelligence--the position that headed the community since 1946, predating the Central
Intelligence Agency's (CIA) creation--and given to the newly-created Director of National
Intelligence (DNI), in large part in an attempt to eradicate the stovepiping that had plagued the
US' intelligence efforts for decades.
Despite this mandate and the subsequent creation of numerous joint intelligence groups to
facilitate intelligence sharing, stovepiping continues to plague the intelligence community,
suggesting the root of the problem lies deeper than habit or neglect. This paper proposes the
stovepiping blamed for much of the intelligence community's dysfunction and several
intelligence failures results from a more fundamental problem--bias. This bias primarily takes
8/3/2019 The Role of Bias in Intelligence Analysis
3/16
one of two forms: bias within an organization in favor of that organization's reporting, or against
the reporting of other organizations, and bias for or against a particular type of intelligence,
usually due to its method of collection.
Assuming the bulk of authors and experts correctly state stovepiping was largely
responsible for previous flawed intelligence that left the US vulnerable to terrorist attack, and
that stovepiping is an ongoing issue, it logically follows that the US continues to be vulnerable to
attack due, in part, to ongoing, underlying bias that contributes to stovepiping within the
intelligence community. This paper will explore the manifestations of this bias toward reporting
from different intelligence collection organizations, and toward reporting from different types of
intelligence collectors. It will then explore whether and which structured analytic techniques
could effectively mitigate the identified biases among intelligence analysts, managers, and
consumers.
Literature Review:
While not specifically detailing intelligence stovepiping, in his book, "Bureaucracy,"
James Q. Wilson described organizational turf protectionism as a key factor in bureaucratic
operations. Wilson asserts agencies and organizations attach high priority to their decision
making autonomy, or turf, and strongly resist and resent attempts to infringe on this turf. Wilson
further states "struggles over autonomy are especially visible when the organizations involved
have similar tasks." While Wilson was referring specifically to the armed forces, this sentiment
is applicable to intelligence stakeholders as well. (Wilson 1989)
In his book, "Getting Agencies to Work Together," Eugene Bardach asserts agencies
often resist collaboration for fear it will blur agency missions and political accountabilities.
Bardach further states agencies may fear obsolescence, potentially resulting in a "ecosystem
8/3/2019 The Role of Bias in Intelligence Analysis
4/16
effect" in which policy aims to protect individual species--in this case, individual agencies and
organizations--rather than the ecosystem writ large--in this case, the IC. Bardach identifies the
same issues regarding autonomy Wilson put forth, stating agencies may reject collaboration to
preserve their decision making autonomy and minimize potential threats from necessary
relationships. (Bardach 1998)
In August 2007, former Director of National Intelligence Mike McConnell wrote his
recommendations for intelligence reform in his piece, "Overhauling Intelligence," published in
Foreign Affairs. Among his recommendations for intelligence reform, he identified several
problem areas within the intelligence community; stovepiping intelligence was one of these
areas. McConnell asserts that in addition to a divide between foreign intelligence and law
enforcement that remained in place as intelligence sharing otherwise expanded post-9/11, the IC
component organizations' unique mandates and narrowly focused missions inhibited IC unity.
McConnell specified that few analysts knew their counterparts in other agencies, and there were
few mechanisms in place to facilitate, let alone ensure collaboration and analytic exchange.
(McConnell 2007)
A 2008 Office of the Director of National Intelligence (ODNI) study into IC management
challenges, conducted by the Office of the Inspector General, identified IC information sharing
as a key challenge. The study found that most analysts relied on personal relationships with
counterparts for information sharing, as agencies responsible for intelligence collection
continued to limit access to data and products to the wider community. Part of the problem,
according to the study, was that information technology systems were largely disconnected and
incompatible, and interoperability between networks was lacking. However, the study claimed
turf battles and agency protectionism continued to be a problem, despite efforts to improve and
8/3/2019 The Role of Bias in Intelligence Analysis
5/16
increase collaboration and sharing, because there were few, if any consequences for failure to
collaborate. (Office of the Director of National Intelligence, Office of the Inspector General
2008)
Kenneth Lieberthal's "The U.S. Intelligence Community and Foreign Policy," written for
the Brookings Institution, was written to identify problem areas within the IC and recommend
changes to make the IC more relevant and responsive to policy and decision makers' needs. The
study identifies numerous problem areas, from overemphasis on a single intelligence product line
to the recruitment of poor quality analysts; for our purposes, Lieberthal's section on
overemphasizing classified reporting and de-emphasizing open source reporting that provides
greater context to intelligence products. Lieberthal also assets a "culture of insularity and
secrecy" works to the detriment of both the IC and its products by limiting dissemination of
intelligence traffic due to variations in security screenings, and lamented that participants in
National Intelligence Estimates (NIEs) represented their agencies rather than themselves, and
therefore perpetuated inter-agency rivalries and partisan manipulation. (Lieberthal 2009)
Mark Lowenthal details IC interrelationships in his book "Intelligence: From Secrets to
Policy." Lowenthal contends many of the developments within the IC since 2001 have actually
increased rather than eradicated interagency rivalries, most prominently between foreign and
domestic intelligence and between civilian, military, and law enforcement intelligence--
especially between the CIA, DoD, and FBI. Lowenthal also identified the potential for "footnote
wars," (as he describes them), in which producers of finished intelligence products attempt to
maintain separate points of view on intelligence topics. Further, Lowenthal describes
stovepiping as a result agencies attempting to highlight their individual relevance to ensure
8/3/2019 The Role of Bias in Intelligence Analysis
6/16
continued funding, and asserts one way intelligence analysts attempt to make their products stand
out is to emphasize the unique nature of their sources. (Lowenthal 2009)
Dr. Rob Johnston's "Analytic Culture in the US Intelligence Community," written for the
CIA's Center for the Study of Intelligence, is a comprehensive look at how the IC conducts
analysis, and many of the problems entailed in this analysis. Johnston identifies several factors
that contribute to intelligence stovepiping and bias, including the conflict between secrecy and
efficacy, agency reluctance to reverse or alter corporate judgments, and confirmation bias in
agency selection and weighing of data according to classification. Johnston cites interviews with
senior analysts and intelligence managers, in which common themes include bias in favor of
classified reporting over open-source intelligence, trust in technical collection over other
intelligence collection, and agency preference for intelligence collected by that agency. Johnston
further contends agency managers believe personnel, training, and readiness should be tailored
specifically for their own organizations, which inhibits interaction with other agencies. Johnston
also asserts that even when placed in a joint environment, analysts will revert to previous
isolationist patterns if not given specific joint processes, principles, and operational structures.
(Johnston 2005)
Steven W. Peterson's "US Intelligence Support to Decision Making," written for the
Weatherhead Center for International Affairs at Harvard University, details the primary
intelligence failures leading to both 9/11 and the 2003 invasion of Iraq. Peterson reiterated the
9/11 Commission's determination that intelligence stovepiping and agency parochialism as a
matter of policy and practice were key contributors to these failures. Peterson also assigns fault
to decision makers' expectation that intelligence can provide certainty, and asserts this
8/3/2019 The Role of Bias in Intelligence Analysis
7/16
expectation causes bias in favor of technical intelligence collection and reporting against
adversary actions, and against collection and reporting against adversary intent. (Peterson 2009)
Gregory Treverton's essay "Intelligence Analysis: Between 'Politicization' and
Irrelevance" addresses both decision maker and agency bias in intelligence analysis. Treverton
details several facets of politicization, including senior policy officials' direct pressure and
agency "house line" assessment. Treverton describes the same desire for certainty Peterson
detailed, but asserts the less technically fact-driven analysis on intent and adversary mindset is
still sought after and of value in shaping policy makers' strategic goals. Treverton implies
decision makers discount non-technical intelligence on intent and mindset when addressing
tactical situations. (George and Bruce 2008)
Frank Cilluffo, Ronald Marks, and George Salmoiraghi's study titled "The Use and
Limits of U.S. Intelligence" published in the Washington Quarterly's Winter 2002 edition
attributed many of the US' HUMINT collection difficulties to a long-running political shift away
from HUMINT collection in the US intelligence community. The authors argue US political
structures degraded HUMINT collection due to the perception HUMINT was more "dirty" than
technological collection methods, such as SIGINT, and attempts to make HUMINT "cleaner"--
such as directives to ensure HUMINT assets were "boy scouts," as the authors quote a
clandestine service officer--further ensured HUMINT collection would not adequately address
US intelligence needs. (Cilluffo, Marks and Salmoiraghi 2002)
The existing literature on the subject clearly indicates stovepiping is an ongoing issue
within the intelligence community, and suggests this stovepiping results at least in part from bias
on the part of policy and decision makers, intelligence managers, and intelligence analysts alike.
8/3/2019 The Role of Bias in Intelligence Analysis
8/16
Discussion:
In discussing bias that results in intelligence stovepiping, we must differentiate between
unconscious bias--that which affects the perception of intelligence reporting or analysis writ
large, without a specific goal or aim--and deliberate bias--selecting reporting for use in analysis
and assessment to support specific goals, aims, or agenda. We must further specify how
different types of bias affect participants in the intelligence production process.
Unconscious Bias:
Desire for Certainty
Johnston, Treverton, and Peterson all describe a preference for technical collection over
non-technical collection. Johnston and Treverton directly attribute this to the desire for certainty,
and the belief that technical collection and analysis based on technical collection can provide
more certain assessments. This appears to be driven by policy makers and intelligence managers
more than analysts, and often results in increased funding for organizations and agencies
supporting technical intelligence collection at the expense of those supporting non-technical
collection. This leadership expectation does impact analysis, however, in that analysts will use a
disproportionate amount of technically collected intelligence in finished intelligence products to
ensure these product appeal to, and are read by their customers.
Preference for Classified Reporting
Lieberthal and Johnston identify the preference to use classified reporting rather than
open source information in intelligence analysis, and anecdotal evidence from more than fifteen
years in the IC suggests this bias is prevalent among intelligence analysts and managers. In de-
emphasizing openly available information--especially adversary or target-nations leaders' public
statements and concerns--intelligence analysts lose sight of the context and social drivers
8/3/2019 The Role of Bias in Intelligence Analysis
9/16
impacting their analysis and assessments. Additionally, classified intelligence collection
depends on collection assets being available and placed where they are able to collect
information, and being focused on targets that ensure all relevant information is collected.
Classified collection platforms that are not perfectly placed and targeted will result in incomplete
or incorrect intelligence collection, and over-reliance on this classified but incomplete reporting
will invariably lead to intelligence failure.
Preference for One's Own Agency's Reporting
Lowenthal and Johnston both identify analyst preference to use unique sources of
information or reporting originating from their own agencies. Once again, anecdotal evidence
suggests this bias is not only prevalent, but pandemic among intelligence analysts and managers.
National decision makers require full-spectrum assessments to identify what adversaries intend,
of what they are capable, the state of their preparation, etc, and no one intelligence discipline--
such as HUMINT, SIGINT, MASINT, IMINT, and so forth--is capable of providing such full-
spectrum assessment. Ostensibly "all-source" analysis that emphasizes one discipline or
collection agency over others is disingenuous, misleading to national decision makers, and
contributes to intelligence failure.
Perception of "Clean" and "Dirty" Intelligence
Cilluffo et al identify the perception among decision makers that technical collection is
"cleaner" or more gentlemanly than clandestine HUMINT collection via spies. While this bias
appears primarily with policy and decision makers, it can prompt these leaders to discount
assessment based on HUMINT or affect budget and resource allocation which subsequently
impact both collection and analysis.
8/3/2019 The Role of Bias in Intelligence Analysis
10/16
Deliberate Bias:
Ensuring Relevance and Resources
Bardach, Lieberthal, and Lowenthal identify intelligence managers' selection and use of
intelligence reporting for analysis to ensure decision makers perceive their organizations as
relevant and necessary for national security. This results in incomplete and skewed intelligence
pictures provided to senior decision makers, which increases the likelihood of poor decisions that
leave us vulnerable to intelligence failure and attack. As with the unconscious bias in favor of
one's own agency's reporting, selecting reporting for ostensibly "all-source" national security
analysis and assessment to support an organizational agenda is disingenuous and damaging.
Protecting Autonomy
Wilson and Bardach identify the overarching desire of any and every bureaucratic
organization to seek and protect its decision making autonomy. One way to do this is to ensure
one's customers--in this case, national level decision and policy makers--rely on the products
your organization provides, in the manner your organization provides them, with the
understanding that pressure from or forced collaboration with external stakeholders will degrade
that product. Additionally, the ODNI's determination that the systems and networks amongst the
IC are unable to communicate and lack interoperability likely stems from the protectionism each
agency performs: maintaining separate and distinct means of communication translates to less
external influence.
Structured Analysis:
Advances in message handling systems--such as the Multi-Media Messaging (M3) used
by many Department of Defense components--have greatly improved cross-organizational
reporting to facilitate its availability to all analysts. However, the problem of bias in the
8/3/2019 The Role of Bias in Intelligence Analysis
11/16
selection and use of reporting in finished intelligence products continues. The top-down
approach to eliminating stovepiping and bias in intelligence analysis and assessment undertaken
in the wake of 9/11 and the Iraq conflict has so far failed; not surprising, as policy and decision
makers are unlikely to acknowledge their biases as such, and intelligence managers are unlikely
to cease their deliberately applied bias.
This leaves it up to the intelligence analyst to remove bias. Several structured analytic
techniques have been developed with the goal of removing bias from intelligence analysis. In
the initial phase of research, gathering information, and winnowing available reporting for use in
an intelligence product, analysts should evaluate the available sources of information to
determine reliability, credibility, the quality of information, and its relevance to the issue at hand.
Beyond this, using a structured technique to objectively compare reporting to hypotheses further
minimized the effect of bias in analysis and assessment.
In the selection and evaluation stage, the Defense Intelligence Agency's "A Tradecraft
Primer: Basic Structured Analytic Techniques" would prove of immense value. Detailed in this
primer are methods to perform relevance checks, source checks, and quality of information
checks--all necessary steps when producing finished intelligence. (Defense Intelligence Agency
2009)
Relevance Check
This check enables analysts to winnow available reporting to that which directly
addresses the issue at hand, mitigates bias when reviewing information and reporting, and
increases analyst confidence that all facets of the issue have been thoroughly analyzed. While
time consuming if there is a large body of reporting on the subject, relevance checks ensure each
piece of data assessed relates to the central issues or alternative possibilities being analyzed. Of
8/3/2019 The Role of Bias in Intelligence Analysis
12/16
note, the relevance check should not be used to gather information and data that supports a
particular hypothesis or opinion--it should determine reporting related to the central topic, which
will then help form hypotheses.
Source Check
This is conducted as part of the first full review of the relevant reporting, and is designed
to mitigate bias for or against a given source. To perform a source check, a series of questions
must be asked when reading each report; for example, when evaluating a HUMINT source, the
analyst should analyze the source information and context, then ask whether the source's stated
placement and access would grant him/her this information, and whether the source appears
biased or is approaching the information from a particular point of view. For ELINT or
MASINT, the analyst should ask about the capabilities and limitations of the collection platform
compared to the data collected, the frequency and duration of collection, the platform's coverage,
etc. The answers to the questions asked while performing a source check will allow the analyst
to both objectively weigh reporting from multiple sources, and to accurately convey confidence
in these sources to intelligence managers and decision makers.
Quality of Information Check
This check enables analysts to evaluate the completeness and validity of available
information. Analysts should question the reports' actual information for completeness, signs of
bias, signs of deception or intention to influence, whether the information is corroborated by
separate sources or intelligence platforms, and whether the information is consistent with
previous information--and if not, whether the information is anomalous, and therefore likely
incorrect, or signifies a paradigm shift that alters or cancels previous assessments.
Structured Analysis
8/3/2019 The Role of Bias in Intelligence Analysis
13/16
Possibly one of the most well known and best regarded analytic techniques is Richards
Heuer's analysis of competing hypotheses (ACH), intended to minimize bias and objectively
weigh reporting against mutually exclusive hypotheses. The eight-step procedure helps analysts
ensure and display thoroughness, and can be used to demonstrate the objective relevance of
reporting collected from varying platforms to intelligence managers and decision makers. (Heuer
1999)
Once the relevance, source, and quality of information checks have been completed,
analysts can use any of several techniques to determine mutually exclusive hypotheses on a
particular intelligence issue or question. ACH compares reporting against these hypotheses in a
matrix designed to determine consistency or inconsistency with each hypothesis. Such a matrix
could look like this:
8/3/2019 The Role of Bias in Intelligence Analysis
14/16
Question: Will Iran retaliate against Israel for Israel's perceived involvement in the 23 July
killing of a nuclear scientist in Tehran?
Hypotheses:H1: Iran will not retaliateH2: Iran will retaliate only with deniable covert/clandestine/terrorist operations
H3: Iran will retaliate with overt military strikes against Israel
- = Inconsistent; + = Consistent; N/A = Neither consistent nor inconsistent
H1 H2 H3
E1: CIA report of
Iranian intent to
conduct retaliatorycovert/clandestine
attacks against Israel
- + N/A
E2: Israeli liaisonreport of Iranian
intent to conductballistic missile
launches against Israel
- N/A +
E3: Iranian public
statements vowingretaliation for the
killing
- + +
E4: NSA report of
Iranian militaryleaders traveling to
Lebanon for ameeting with
Hizballah
+ + +
E5: CIA report stating
Iranian military
leaders believe theyare unprepared to
repel an Israeli air
strike
+ + -
E6: DIA report statingIranian Islamic
Revolutionary GuardCorps personnel were
photographing the
Israeli embassy inTurkey
N/A + N/A
8/3/2019 The Role of Bias in Intelligence Analysis
15/16
Once completed, the analyst can use the matrix results to support a most likely
hypothesis, as well as assign relative likelihoods to alternate hypotheses. For our purposes,
utilizing all available, relevant reporting in such a matrix effectively removes unconscious bias in
source selection and utilization, and can be used to demonstrate each intelligence collection
platform and organization's relevance to intelligence managers and decision makers.
Conclusion:
Structured analytic techniques can assist intelligence analysts avoid unconscious bias
associated with stovepiping in the creation of finished intelligence products, and assist analysts
display unbiased relevance of all organizations' reporting to intelligence consumers. While these
techniques will not directly remove or decrease deliberate unconscious bias in decision makers
and intelligence managers, consistent, transparently advertised use of these techniques can make
such biases more difficult to justify and maintain.
8/3/2019 The Role of Bias in Intelligence Analysis
16/16
Works Cited
Bardach, Eugene. Getting Agencies to Work Together. Washington D.C.: Brookings InstitutionPress, 1998.
Cilluffo, Frank J., Ronald A. Marks, and George C. Salmoiraghi. "The Use and Limits of U.S.
Intelligence." The Washington Quarterly, 2002: 61-74.
Defense Intelligence Agency.A Tradecraft Primer: Basic Structured Analytic Techniques.
Defense Intelligence Agency, 2009.
George, Roger Z, and James B Bruce.Analyzing Intelligence. Washington D.C.: Georgetown
University Press, 2008.
Heuer, Richards J. Psychology of Intelligence Analysis. Washington D.C.: Center for the Study
of Intelligence, 1999.
Johnston, Rob.Analytic Culture in the US Intelligence Community. Academic Study,
Washington D.C.: Center for the Study of Intelligence, 2005.
Lieberthal, Kenneth. The U.S. Intelligence Community and Foreign policy. Academic Study,
Washington D.C.: The Brookings Institution, 2009.
Lowenthal, Mark M.Intelligence: From Secrets to Policy. Washington D.C.: CQ Press, 2009.
McConnell, Mike. "Overhauling Intelligence." Foreign Affairs 86, no. 4 (2007): 49-58.
Office of the Director of National Intelligence, Office of the Inspector General. CriticalIntelligence Community Management Challenges. Washington D.C.: ODNI, 2008.
Peterson, Steven W. US Intelligence Support to Decision Making. Research Paper, Cambridge:
Weatherhead Center for International Affairs, 2009.
Wilson, James Q.Bureaucracy. Basic Books, Inc, 1989.