At various times during the workshop, especially at the end of the day and at the end of a session, many issues relating to strategic nuclear deterrence, the usefulness of various analytical tools, and the content of a possible follow-on consensus study were discussed by workshop participants. The sections below summarize collections of such comments by individual workshop participants, particularly those who attended the entire workshop and contributed significantly to the summary sessions. These comments reflect the considerable diversity of opinions expressed during the workshop on a range of issues. The last section contains illustrative terms of reference for a possible follow-on study.
INSIGHTS OF VARIOUS INDIVIDUAL WORKSHOP PARTICIPANTS
Many participants noted that strategic deterrence, as with strategic stability, means different things to different people and that strategic deterrence is not simply the nuclear deterrence of the Cold War. The nuclear dimension, however, was the key focus of this workshop. While strategic use of conventional weapons is clearly an alternative to nuclear use, the participants did not focus on other things that could be used in a campaign. To illustrate the range of views expressed by the participants, one view was that the record shows the Russians always overestimated the United States, not so much in capability but more likely in resolve. Another view was that a rich set of challenges currently exist, such as how can the United States verify what weapons China possesses, given its extensive underground tunnels.
On General Chambers' Presentation
Gen Chambers asked, "What is it we give to the President to deter and assure? We need to develop and foster critical thinking on deterrence and assurance." These notions align with one of the Air Force's important vectors. He also reminded the participants that an examination and critical evaluation of appropriate analytic tools would be of great value to the Air Force in understanding its mission of organizing, training, and equipping two legs of the strategic triad.
On General Klotz's Presentation (Nuclear Posture)
Gen Klotz believes the administration's orientation has been clear, and there has been some consensus in Congress, but it is fragile. A participant indicated that, although the orientation is "toward" global zero, there are lots of cautions about maintaining a reliable and secure force in the meantime. A major concern also raised by this participant was that the budget tightening will lead to increased disagreements. That participant also noted that Gen Klotz partially bought into the argument that the India-Pakistan proliferation might conceivably have been avoided had the world had its act together on the Treaty on the Non-Proliferation of Nuclear Weapons and the Comprehensive Nuclear-Test-Ban Treaty, for example. On that point the participant was extremely skeptical.
A participant believed this was a very interesting presentation on the relevant issues to deterrence but noted that Gen Klotz seemed to argue for status quo with no reduction in resources, which did not seem realistic. Also, he noted that Gen Klotz did not address how to do more with less. Another participant noted that Gen Klotz’s message was to not get hung up on Global Zero rhetoric; U.S. policy is not to go there unilaterally and to keep nuclear weapons safe, secure, and effective so long as others have nuclear weapons. But, that participant argued that strategy will not drive decisions made in an austere budget climate by politicians with higher priorities than nuclear deterrence. Noting that Gen Klotz believes the time is right for establishment of a new national consensus on the support and sustainment of nuclear deterrence, a participant observed that this will require two schools of thought to agree (those who say nuclear weapons are needed and those who advocate the elimination of all U.S. nuclear weapons); can both be satisfied? He concluded that this will require fact-based analysis plus the tools of such analysis.
Ms. Woolf's candid presentation during the second session elicited many favorable comments, as summarized below. She indicated there tends not to be a congressional perspective, per se. Ms. Woolf noted that nuclear weapons do not have a high profile among members and that institutional knowledge has decreased over time as important members and staff members that were present during the Cold War era have retired. As a result, she stated members tend to vote along the same lines as the more knowledgeable members, which has the advantage of meaning fewer people need to be convinced. Ms. Woolf also stated that the reasoning used by staffs, often driven by advocacy groups, reflects first-order arithmetic only, which can be misleading because these calculations do not take into account underlying strategic, conceptual, or operational issues. It is possible, she argued, to change their focus, sometimes, but analysis has to be convincing and relevant to home districts or budgets.
The participants agreed that Ms. Woolf gave a fascinating description of the congressional process and explained the difficulty involved in getting traction for deterrence issues. One participant summed it up as follows: “Congress 101” means this: youth, other priorities, the Cold War took place in “ancient” times, nuclear non-proliferation and security are today's problems; what's in it for my district? cut the deficit; follow the leader, but who will
lead as the experienced ones leave Congress? The participant added that the information age empowers non-governmental organizations and bloggers—a challenge for getting sound analysis into the decision process.
A participant voiced the need for showing how the maintenance of deterrence force structure at appropriate numbers can in fact be useful to a particular interest of someone who would otherwise advocate the sharp reduction or elimination of such weapons. Ms. Woolf’s response was to postulate outcomes of deterrence that help satisfy something else of interest to a would-be detractor.
Several participants identified many tools that may be of value to the Air Force (see Box 5-1), and many comments during the sessions related to them. A synopsis of those comments follows.
A general view by several participants was that it would be of interest to look for a suite of complementary tools and then close the aperture on bounds for possible decision making. It could be useful to do many of these, but some participants stated that they did not know how to work in some of the military environments or how it might work in a classified setting. Many participants believed they must know what information is needed and what tools could be used to get it.
Other views were as follows. One participant noted that a lot of the problem is that theory and data to support such tools is not there. Validation of such tools is most important; one would like empirical validation to be 90 percent, but there will never be an empirical way to prove some of this. Users will have to be exposed to different elements as bounding mechanisms. Another participant pointed out that there has to be some assessment of these things; how much can they be trusted? For many methods, one needs to see what works with real people.
A pessimistic view from one participant was that most of this is not ready for use now. Numbers from some of the decision tools may be worse than random. On the other hand, a more positive view came from another participant who noted that there is a huge amount of information available on new analytic techniques that is just beginning to be tapped. New concepts and methods should continue to be searched for and examined, even if some might at first be considered wild and crazy. Other participants affirmed that the real value is considering types of data that can be generated to attribute motives and perspectives to various entities (e.g., adversarial nations, terrorists). Social neuroscience research is showing “us-them” reactions and is very interesting relative to combat and ethical or moral dilemmas.
Regarding the notion of using neuroscience, a concern expressed was that one must worry about biases. A person steeped in deterrence thinking may not behave the same as a college student getting paid by the hour. Also, a lot of these studies are based on trivial tasks, which are unlikely the same as complex international tasks. One participant stated that psychological studies largely represent averages over many people, but some risk seekers and
BOX 5-1
List of Analytical Tools Considered During the Workshop
A key focus of the workshop was to identify different techniques or methods the Air Force might use in addressing strategic deterrence capabilities. At least 16 approaches were discussed:
- Qualitative analysis (international relations/strategic studies/estimative intelligence);
- Historical case studies;
- Historical statistical-empirical analysis;
- Operations research;
- Simulations and war games;
- Game theory;
- Simple deterrence analysis using synthetic cognitive models;
- Actor-specific behavioral modeling and leadership profiling;
- Agent-based computational modeling (both simple and complex cognitive decision models);
- Social network analysis/influence diagrams/data mining;
- Subject-matter-expert elicitation;
- Crowd sourcing;
- “Evil genius” and “crafty bastard” efforts;
- Insights provided by neurobiology as related to behavior;
- Heuristics; and
- Systems engineering models.
_______________________________
NOTE: In light of Dr. Todd's beliefs regarding the value of "simple heuristics," at least one expert cautions that with projected advances in computational capabilities, such as exaflop computing by 2020 or sooner, there will be a temptation to take a systematic modeling approach to address the higher-order complexities of deterrence techniques and capabilities.
others have different characteristics, which are not suitable for specific situations. A counter view was offered that, nevertheless, some data may be able to narrow the possibilities. Some blending of historical record and profiles with some of these techniques could have value, but one must be sure not to set decision makers up with biases.
On Profiling (Including the Panel Presentation)
Some believed strongly that there should be no shortfall on resources devoted to developing leadership profiles, which are crucial. They noted that profiling can identify tendencies, trends, and patterns, but it is not for predicting. Psychological operations are very important (for example, telling a population about luxurious life-styles of its leaders), and it is unimaginable to not know about a leader. They concluded that to augment Department of Defense decision making, more must be known about leaders. More intense intelligence effort is needed to get at closed societies.
Other participants were not convinced that profiling had significant value in all deterrence contexts, believing instead that more information is not necessarily better. Not all insights are useful; they must be tested along the way. Nevertheless, others thought profiles by different teams might help. There is merit to seeing how the other analytical tools discussed might help with profiling; social networking tools, such as sentiment mining, could play a big role in understanding adversaries and their populations. According to several participants, how to make these tools more robust is a big issue. Profiles appear to fit well with social networking tools.
The panel on leadership profiling approaches was held during the second workshop session. Drs. Winter, Walker, and Hermann mentioned being frustrated in their work by the insufficiency and immaturity of computer-based coding software plus the difficulty and time-sink of translating documents to be coded into English. A participant thought it would be better to have coding schema and tools that could handle documents in native language, but natural language coding is not there yet. Several participants believed these speakers did good work and that this is worth looking at further. They argued that the various leadership profiling approaches used in concert can yield an outcome greater than the sum of its parts.
Additional points of view were as follows. One participant stated that textual analysis—for example, use of verbs, power language, and other linguistic clues—has already been developed in some detail. These methods appear to be potentially useful for recognizing changes in leader (and influencer) attitudes and intentions based on their published speeches and remarks. A related point was made that inter-judge concordance is already high enough to suggest that these methods have some reliability. That participant also claimed that interpretation of historical experience suggests that leader language may help to predict aggressive versus less aggressive actions in escalating or resolving conflicts. Careful independent validation of these methods may be useful in determining whether they are ready for use in the context of deterrence. Extensions to detect shifts in the thinking of key influencers and shifts in power among factions, as well as hardening, softening, or changing positions or intents of factions, might be especially valuable.
Dr. Hermann and the other two panel members described approaches to leadership profiling. Although she gave fewer details than the other two, a participant thought Dr. Hermann’s approach could be more amenable to computer tagging. Most thought all three panelists' work could be very useful in improving deterrence. It was noted that Dr. Winter's approach requires manual labeling of concepts from a taxonomy that includes concepts such as power imagery. He provided quantitative support for his work and noted that since it requires manual labeling it is difficult to use it with social media sources, but it could be very useful with selected document sources. A participant noted that Dr. Walker presented a very similar approach to Dr. Winter's and suggested that both approaches can be automated, but there was not a chance to discuss it further.
Dr. Todd gave a presentation on an important topic. In addition to a learning tool for analysts, a participant wondered if there could be a way to help planners learn about decision
biases? Another participant noted that heuristics have been shown to provide value as an aid in decision making in a variety of enterprises—but not yet in strategic deterrence. His additional views were that the key would be to pick the right heuristics and to know when and when not to rely on them; use of the wrong heuristics could be disastrous if the wrong one is picked; and application of subject-matter expertise is essential.
Participants also noted that Dr. Todd showed how, in some instances, less information is better than more. He pointed out that simpler algorithms can outperform more complex ones and provide answers in a shorter time.1 He also described his ongoing work in cognitive bias amelioration as part of the Intelligence Advanced Research Projects Agency’s Sirius program. Another participant observed that fast, frugal heuristics have proven useful for some problems—for example, guessing which cities are biggest based on recognition. However, they have not been studied yet in the context of deterrence (or other game theory settings, such as multi-way negotiations, or formation of a consensus decision starting from factions with different preferences). The participant added that understanding fast, frugal heuristics for conflict escalation and resolution (if they are used by people in reality) could be useful.
Mr. McKenna described U.S. Strategic Command’s (STRATCOM's) generic approach to analysis of force structure issues. Strategy should come first because it drives results. A participant noted that the reason for this is that they do "requirements analysis," which assesses ability to do a well-specified job, rather than characterizing capability. The participants understood that the USSTRATCOM approach handles the kinetics pretty well, but it has not done very well on issues relating to individual decision makers, political context, and world environment. USSTRATCOM does, however, consider different futures and conduct an "attribute" based parametric analysis. Regarding Mr. McKenna's framework for thinking about deterrence, the most interesting component according to some participants was his explicit separation of the overall deterrence process into ends/ways/means. He showed how to link policy/strategy to outcomes; sound analysis from two different staffs (USSTRATCOM and A9). It was also understood by the participants that sorely lacking with this type of analytical approach is an ability to understand adversary perceptions and intentions. This lack is in great contrast to the well developed ability to understand an adversary’s capabilities.
On the Approach of the Air Force Office of Studies and Analyses, Assessments, and Lessons Learned
Maj Sorice described his organization's systematic efforts to analyze the implications of lower force levels across many possible conflicts with different strategies. The primary
___________________
1 While not stated explicitly during the workshop, some experts caution that the power of computation should not extend beyond the power of comprehension. However, one should not discount the understanding that may come from modeling and simulating highly complex problems. With the expected progress over the next decade in "reverse engineering" the human brain, one can expect rapid progress in expanding "natural bridges" between what the human brain can do best with what the computer can do best.
takeaway according to many participants was a table showing that the ability to support various classic targeting options changes (or disappears) at lower force levels. The analyses were not discussed in detail; instead, Maj Sorice presented a rather detailed assessment of force postures by numerous metrics across the uncertainty space (because of concerns about classification, there were no numbers given). Participants believed his presentation offered a framework for thinking about deterrence. They noted that there were no analytic inputs (i.e., no weights associated with items), but it appeared to be a pretty complete framework for analyzing deterrence approaches. A participant observed that Maj Sorice showed how to link policy/strategy to outcomes; again, sound analysis was provided by two different staffs (USSTRATCOM and A9). The A9 organization is expecting to drive forward in fleshing out and applying the analytic framework it presented.
On Scenario-Based Analytic Tools
Gen Elder described complex modeling that is state of the art, but at least one participant questioned how it can be validated, asking, “What are the criteria for selecting subject-matter experts?” Gen Elder described an analytic framework that he developed, but at least one participant had difficulty understanding the details of its use and could not assess the value of the approach. He noted that models and integrated ensembles of models for generating insights are already available, such as Pythia, Construct, and the framework developed by the Concepts and Analysis of Nuclear Strategy study. A participant's view was that these models generally have uncertain validity and stop short of supporting decisions, except by providing possibly useful (but possibly misleading) insights into connections among variables. More expressive models and better validation are probably essential for closing the gap between insight and well-supported decisions.
Ms. Russell described interesting analytic tools, including Narrative Pattern Analyzer (NPA) and Influence Net Modeling (iNET/SIAM™), which show the power of new methods. A workshop participant believed, however, that it was unclear how they would be readily adapted to nuclear deterrence. Another workshop participant believed that the more useful one of the two for the purposes of this workshop appeared to be SIAM ™. It provides a Bayesian framework for improving estimates with incoming information and appears capable of being employed for deterrence work with modest effort. Another participant suggested that software such as iNET and Palantir make it practical to track patterns in space and time. NPA and similar software may provide valuable clues about emerging patterns and potential threats, including shifting attitudes toward use of nuclear weapons.
On Threat Anticipation and Intelligence Analysis
Although one participant did not see any takeaways from the panel on threat anticipation and intelligence analysis, a few observations were offered. Dr. Wagner sees strategic forces as a training base for the United States to maintain knowledge and skills until—perhaps decades from now—there is again a need for "real" nuclear forces. Dr. Wagner believes we are in a strategic pause and need to be ready with appropriate analysis and analytic
tools when we come out of this period. Another participant noted that Dr. Wagner did present an additional framework, the most interesting part of which was identification of a matrix that breaks up deterrence situations into four elements—negotiated monitoring/not versus treaty verification/threat.
Mr. Hamon’s experience could be very valuable in developing a tool taxonomy in the future. Another participant believed there are limitations of the estimative process (manage expectations—it gets back to understanding intentions). That participant also noted that a Defense Science Board study of nuclear monitoring is expected to be available soon, and many nuclear-related studies sponsored by the DTRA’s Advanced Systems and Concepts Office were done over a decade or so.
INSIGHTS FOR A FOLLOW-ON STUDY
During both workshop sessions, but especially the second, workshop participants offered many insights regarding the content of a possible follow-on study. The dialog focused on an illustrative TOR that could form a framework for such a study. Several versions of this TOR were discussed and modified during the workshop, taking into account a wide range of individual views of the participants. The notional TOR in Box 5-2 reflects comments from various participants and could serve as a starting point for decisions by the Air Force and National Academies regarding a follow-on study. During discussion of the TOR, two other suggestions offered by workshop participants were (1) for the longer-term study, why could it not look at simulation scenarios and games to see what tools might work, and (2) the study could begin with a presentation of the security environment by using a geographical schema to present conditions in applicable areas of interest; deterrence matters should be considered region by region as well as in a strategic sense.
BOX 5-2
Notional Terms of Reference for a Follow-on Study
As identified during the workshop, possible items in the terms of reference for a follow-on study by an ad hoc committee were as follows:
1. Identify the broad issues and factors that must be considered in seeking nuclear deterrence in the 21st century. Describe a program of analysis to address those issues and support planning, resourcing and managing U.S. nuclear deterrence in the 21st century.
2. Identify the major components of the analysis and the relationships among them to serve as a basis for the identification, development and use of necessary tools and methods.
3. Evaluate and recommend tools, methods, including behavioral science-based methods, and approaches for improving the understanding of how nuclear deterrence works in the 21st century, how it might fail, and how failure might be averted by the proper choice of capabilities, postures, and concepts of operation of American nuclear forces.
4. Recommend a way ahead for evolving and adapting methods and approaches in a coherent, systematic approach. This will include identifying what questions need to be addressed, and assessing what questions each tool, method, or approach is most and least valuable for this purpose.
5. Recommend how these methods and approaches can be drawn upon as a package, or used to inform each other. It is likely that any tool, method, or approach will have strengths and weaknesses.
6. Recommend criteria and a framework for validating the tools, methods, and approaches and for identifying which classes of tools, methods, and approaches are the most promising.
7. Recommend a balance of resourcing across the classes in today’s austere financial climate and that can be reserved for future resourcing when and if it becomes available.
________________________________
NOTE: While the workshop committee did engage in much discussion between the two workshop sessions on what could constitute the basis for the TOR of a follow-on study, the TOR reflects a much broader discussion that occurred at both workshop sessions among the many participants.
