The uncanny valley and the virtual briefer

We have long had a soft spot for the idea of virtual briefers – that is, interactive software agents which can deliver customized presentations to intelligence consumers. While nothing will ever likely beat the effectiveness of a real mind delivering insight regarding issues on which they have directly worked in a manner tailored the consumer’s interest, and being available for discussion and debate in real time, the virtual briefer could take the place of a large segment of the other briefing cases. Anyone who has spent time in the community knows these well – the interminably long, bullet point by bullet point monotone recital of someone else’s key judgments by a speaker who ought never have been placed in that position.

There have been multiple experiments along these lines, dating back ill-fated virtual newsreaders in the comparative dark ages of the Internet. Recently, lightweight video and audio production capabilities have seen a resurgence for delivery of prerecorded briefings and lectures (for which we are grateful, given the need to fill otherwise unproductive downtime in the friction of travel or commuting.) We are also aware of one relatively recent academic experiment at Mercyhurst’s intel studies program to create a new style of video briefing (or rather, a modernized lightweight version of the kinds of products frequently seen in the Reagan era White House, or even in visually dominant shops such as NGA) – although the results of that study have still regrettably not been made public. While these efforts all have their place, and have provided valuable lessons learned – especially when it comes to the gaps which must be overcome – the concepts are simply not sufficiently advanced to allow for effective widespread implementation.

However, the need for alternatives to the classic briefing continues to grow, especially as the range of consumers outside of the traditional spectrum served by the IC also grows. How many otherwise productive analysts have been sapped on a more or less fulltime basis by the need to support the multiple times daily briefings given in the many watch & warning oriented shops around the community? How many thinkers are burning themselves out at 3am to prepare for a consumer’s morning ritual? And how many of those briefings continue to be delivered to ever lower echelons of consumers merely interested in avoiding having to read a written product rather than in the true back and forth discussion and engagement of well executed briefing?

We thus continue to feel that there is a strong case for the further development and implementation of virtual briefers, at least in certain situated instances. We have been encouraged by advances in the key enabling technologies that would be required to make a robust implementation of this happen. Among the most critical are those that overcome the previous gaps in the cartoonish delivery offered by previous generations of avatars. However, most attempts at photorealistic avatars have thus far foundered in the uncanny valley – the response of the audience to appearance and behavior which is nearly, but not quite, entirely human.

Recent news out of the video gaming industry may however offer a potential solution to at least this part of the problem. At least of one firm has made exceptional progress – at least by evidence of the demonstration video – in bridging the uncanny valley.

Of course, conversational agents still have a long way to go. Without such agents, we would reiterate our previously expressed our concerns regarding the removal of the opportunity for human interaction between analysts and high value intelligence customers.

As a stopgap, one can however easily envision a hybrid structure evolving, in which basic briefings are delivered in a syndicated fashion by virtual talking heads, while a lightweight response interface captures any comments or questions from the consumer for routing to the appropriate analyst. It is certainly not real time, but for a wide variety of consumers such asynchronous but more tailored service may be exactly what they have been looking for when they cannot afford – or sustain – the level of analytic effort required for full time briefing support.

We have reason to believe that it will be the private sector intelligence shops that may adopt such virtual briefers first, as they have also been the early adopters of podcasts, webcasts, and other innovative delivery options. (We would not venture to lay odds on whether it will be a Stratfor, an iJet, an Economist Intelligence Unit, or some other entirely unknown shop - perhaps in the competitive intelligence arena - that takes this plunge first. And we would be even more foolish to predict which one might first succeed.) Yet one can easily see how a subscription based service could easily syndicate the dissemination of selected products – crafted by the usual suspects who lurk in the cubicle mazes or darkened basements – but delivered in a unique format tailored for the busy consumer. Likewise, one might envision a successful offering by a small competitive intelligence department located an otherwise large organization with extensive demands as yet unmet by traditional product formats. And with the new options for flash based video available from smartphones (including the increasingly ubiquitous iPhone), such a service - in whatever form it may emerge and evolve to thrive - is likely not too far off in the distant future.

Imaginary constellations meet visionary capabilities

We rarely comment on ongoing matters in the intelligence world due to the natural restrictions which accompany a professional’s responsibilities. However, the very public intercept of USA193 does touch upon an issue that we have previously discussed in these pages – the disconnect between what assets those planning for the future intelligence community thought they would have, and those that exist in an IC in which we have gone to war with what we actually have.

The shot itself has been extensively discussed by others, including In From the Cold, Arms Control Wonk, Ares, Danger Room, and many of the more mainstream blogs. We will not rehash these public details.

However in this particular instance, we find the contrast between the notional pieces on the chess board (and the detritus they left behind when the model came tumbling down) quite striking against the very effective real world capabilities offered by one of the most condemned warfighting concepts of recent history - hit to kill. These differences are also at the heart of the classic tensions between intelligence and operators. But it also demonstrates the value of innovative vision – and the need for persistent effort towards what may seem an unachievable goal. The operation also shows the adaptability of those capabilities towards missions that may never before have been conceived – particularly when the tools to execute them were first on the drafting boards.

The debates of the past few decades across all of the systems involved, and last night’s operation itself, also brings to mind the Bill Whittle’s now seminal essay on Tribes. The defense space community gave the public an unprecedented glimpse of world of the Gray at the high frontier. The press briefing (found here) is as specific a discussion as ever has occurred regarding a once entirely classified area. One has to remember that only a few short years ago, even the name of the National Reconnaissance Office was not publicly acknowledged – let alone the kind of systems of systems approach evident in this operation. But the briefing also displays strong analytic tradecraft, and an excellent use of estimative language to communicate intelligence and operational information in a transparent manner. It was indeed an impressive display, and is well worth studying for those that are in the business of facing tough talks.

Altogether, this was a unique mission, and those involved will be able to look back on the operation with pride. They have most deservedly earned the drinks being poured last night and this morning. Space control has a new face. Let us hope that this will also be the symbol that inspires future generations of capabilities on the intelligence side of the house, and the desire to avoid any other holes in the imagined sky.

Revisiting analytic rigour

The research currently being done at Ohio State University into the problems of intelligence analysis – including information overload, cognitive processes, and other aspects of the methodology – has from time to time caught our interest. Among the more interesting of these items now in circulation is an excellent lecture that we most highly recommend to our readers, recorded last year during the too often overlooked Google Talks series. The discussion focuses on the evaluation of analytic rigor, and means by which analysis may be strengthened.

We particularly favour the philosophy that Dr. Woods presents, which seeks to avoid dictating a single best methodology or process. We are more than willing to listen to the methodologists, but too often we find a dictatorial approach significantly at odds with the realities of line analysis. We think that the observed case study technique used in the Ohio State team’s research – something too infrequently done by many academics – is key to the validity of their findings. One cannot discuss analytic ideals without involving those who are actually involved in applying tradecraft to real problems. It is also not enough to conduct such research in artificial environments within student populations – real line analysis is too different.

We certainly cannot agree with the apparent off-hand condemnation of “folk” psychology of intelligence analysis – clearly aimed at taking on Heuer’s “bible”. While we think that there is a clear role for the methodologists and their research into strengthen analytic tradecraft, there is also a very real need for interdisciplinary adaptation from other areas of social science, as well as the kind of internal discussions that make up a key part of the maintenance of those oft-criticized, but entirely vital, guilds that are the backbone of the community.

We do however find several key concepts of great interest that deserve wider attention, including the concept of the Supervisor’s Dilemma – the balance of customer outcomes requirements and analytic resource opportunity costs against the relative depth of analytic rigour. We also find the study techniques themselves of interest, especially the concept of elicitation through critique – something we feel will likely have a far greater applicability in capturing the kind of intergenerational knowledge that the community is in danger of losing. We see the technique as one means of making more formal – and scalable - some of the kinds of subtle interactions that characterized the experiences of apprentice and journeyman analysts under the mentorship of a master.

We also find great merit in the good professor’s comments regarding the overconfidence of new analysts, and the satisificing biases that result. We definitely have observed a level of arrogance in too many new hires – and especially those coming out of the intelligence studies programs. The first lesson that an analyst student should learn is the fear of God – and of their own error. Too many programs of instruction are not affording the student the chance to learn that fear from the visceral experiences of their own mistakes, and to take away from the experience a humility that will cause them to productively question their future work toward its improvement. Such experiences are far better gained when the consequences are not fatal, in line with the lessons taught by Red Flag.

There is much food for thought in this lecture, as well as the contributions to the literature that the Ohio State program has generated. We will no doubt have further commentary on the subject in the near future.

New technologies for facilities characterization

We recall – and not all too fondly - the early days of attempts to create architectural CAD renderings of target facilities. This was a clumsy process, which almost always required an engineer of some description to be involved, and frankly created an end product that most consumers didn’t see as anything more than a low resolution graphic. To be sure, there was always the bragging rights of adapting a new system to a classic target, but on any given day we preferred a good graphics artist with a keen eye for perspective and proportion far better. Kind of like architects themselves, really, given their preference for covers showing artist’s concepts of buildings rather than displaying the old blueprint style plans. Newer generations of architectural software have apparently made this task easier, but have not been as enthusiastically explored.

But a new technology may change that, by bringing back a level of interaction with the system that the insulating layer of engineering specialization took from the CAD models. We are already on record as being fans of the concepts behind multitouch style screens – especially in their as on just fictional style incarnations (courtesy of the futures studies folks). There appears a lot more to be explored in the space, though, as new applications are continually popping up that offer to reconceptualize human computer interface for a variety of tasks across the intelligence profession.

The latest incarnation that has come to our attention promises a “virtual” view of factories – and in particular, aggregate views of systems data for reactions that cannot be seen directly. One can readily imagine the utility such a system would have for those engineers and analysts attempting to assemble a composite view of a competitor’s industrial processes – or an adversary’s chemical or biological weapons production facility. The key to the technology’s innovation in our view is not the S&T solution, although this could indeed be valuable, but rather the engagement with a high fidelity visualization that the multitouch screen (and its follow on evolution) could bring. We strongly believe that innovations like this are vital in opening up the more arcane collection and analytic disciplines to all source generalists – and more importantly, the intelligence end user – in ways that graphs and pictures of CAD renderings could not.

We look forward to the day – hopefully sometime very soon – when we might be able to host a briefing around such a table screen, discussing some hard target with a consumer that can literally get their hands (and heads) around the issue. We think, however, that in this our counterparts in the commercial sector might lead the way, but if nothing else the use of such visualization techniques in competitive technical intelligence might provide an excellent example to reference when building out an acquisition justification elsewhere in the community.


h / t Smart Mobs

Virtual talking heads

The circulation of the survey instrument for Mercyhurst’s new research study on video format delivery of finished intelligence has occasioned no small measure of debate among participating professionals within the community. While we are glad to see the research being conducted to advance the understanding of the field, we have some serious questions about the wider implications of the use of video for product delivery. Its historical applications have occasionally resulted in unique value – particularly for certain high level consumers, such as its famous use as part of President Reagan’s daily briefing. However, for routine production it imposes burdens on development, editing and coordination, archival processes, and technology / connectivity requirements that create issues far beyond the initial considerations of communications efficacy.

The entire discussion also reminds us of the earlier attempts to re-create intelligence briefings into virtual formats through automated processes. First popularized in response to Web 1.0 start-ups such as the virtual news readers such as Anna Nova, the concept never moved beyond research and development phases precisely for some of the same reasons. The removal of the human briefing element also created a disconnect between attention and pacing, which alienated many consumers whose areas of interest diverged from the set piece delivery – but who may have been more productively engaged by a dynamic briefer. There is also the subtle but very real impact of the uncanny valley – the noticeable aspects of the artificial in “too human” like virtual entities. (We wish to note one of our close colleagues’ wise but whimsical suggestion to replace a human virtual briefer with an object of classic children’s theatre. If nothing else, we would love to have such a product on hand to inject a bit of levity into meetings in which the participants take themselves too seriously.)

Set piece delivery tends to work best in environments where the consumer can devote only partial attention to the product – such as audio delivery of briefing or lecture materials, designed for those who are looking for cognitive engagement during other routine tasks such as driving, physical exercise, or the like. We have not yet seen an environment in which video delivery could meet this same niche – particularly for products which demand a higher degree of attention than watching CNN with the sound turned down in the typical watch center fashion. Virtual talking heads suffered the same problem.

There also remains the serious question of the longer term impacts of removing one of the few opportunities for direct interaction between consumers and intelligence professionals, where the intangibles of human relationships are nowhere as easily quantified. Perhaps there are some aspects of intelligence that simply ought not be automated, to the ultimate benefit of consumer outcomes.

Nonetheless, we will be interested to see the results of the study, and happy to see the literature of intelligence advanced in this area. We can only hope other academic studies will continue to build understanding of the nature of intelligence communication.

The future of the VTC

The Video TeleConference (VTC) is one of the little heralded advances which has so transformed the nature of command and staff decision making during this Long War, as it moved down from the most senior command ranks (as in Kosovo) to ever progressively more connected lower echelons. (For good or for ill is for the historians to decide, should there ever be a concerted effort to review this humble administrative technology amongst all of the other wiz-bang geek toys now given to us humble knuckledraggers.)

The future of the VTC may be emerging, however, from an unlikely spot. Virtual “actors working in real time from remote locations have been beamed onto a stage where they performed with live, in-the-flesh actors.”

As fascinating as this sounds from the futurists’ perspective, we cannot help but remember the old teleconferencing system originally established for nuclear command and control, for use by the National Command Authorities in the dark ages of the 1960’s. This older system was quite visually striking, involving the use of mannequins seated at a conference table, whose features were a transparent model of the individual designated to be virtually present. Into that shaped screen model was projected a video of the speaker from the remote location. All in all, for the time no doubt the very model of the modern virtual environment, and the think tank (RAND, we believe) and its government sponsors were undoubtedly and justifiably as proud of it as those actors are of their mediated creation. However, looking back, will we find such artifacts as quaint, and as disturbing, as their early warfighting counterparts?

Among the great unpublished historical images of the Long War will be any one of the number of rooms installed in numerous palaces throughout the theatre, marked by a high dollar cost plasma flat screens, with diligent staff clustered around the only faint light in the otherwise darkened room, the other circuits on overstressed electrical systems having blown out but the all important VTC still active… sic transit gloria mundi.