Tim Stringer, Partner Success Manager

On a balmy summer’s day in a bustling city square, the sudden appearance of strangely marked vehicles stands out. It’s clear they’re military from a glance.

Heads turn as the trucks come to a halt. Their passengers disembark and heavy boots hit the ground. When they draw their weapons, cautious intrigue turns to panic. Some flee. Others take out their phone to snap the drama, or take to social media to discuss and digest the event.

Within hours, the military presence has gone viral. Routes out of the city become congested as word spreads and families seek safer surroundings. In other areas, crowds amass in protest, driven by a new-found desire to rid their city of the hostile inhabitants. Word spreads on social media, sentiment starts to form, and, for a very short while, the narrative is up for grabs.

This may be a hypothetical example. But it emphasises a crucial point, and one I certainly experienced during my time in service: online is undeniably part of the modern frontline. The web-enabled battle of narratives and prevalence of disinformation in a world more connected than ever has fundamentally altered the character of war once again.

As Elisabeth Braw writes in The Defender’s Dilemma: Identifying and Deterring Gray-Zone Aggression, ‘The ease of information access – especially the un-vetted information disseminated on social media – provides hostile states with enormous opportunities to sow discord in the targeted country, and the disinformation can seize on any subject.’

It’s not a stretch to think of social media as part of the battlefield. Each social application and its host phone, computer and tablet harbours the potential for distorted propaganda and newspeak, accessible at the touch of a button or share of a post. Amidst the fog, gaining and sustaining the support of the population is paramount. Moulding common sentiment is critical at local, regional, national and international levels. After all, those who own the narrative often win the war.

Carl von Clausewitz tells us that war is an extension of politics. We go to war to achieve political ends. What we sometimes fail to appreciate is that these political ends are a manifestation of the societal and cultural values and beliefs that are played out in the information environment every second of the day.

Understanding and influencing how people think, feel and act, both on and offline, means exploring and exploiting the cognitive dimension. But humans are complex, and – perhaps unsurprisingly – it’s difficult to determine how to cognitively affect the human brain in a measurable way. Harder still is figuring out the exact route to changing a long-held perception or belief.

Over time, this has led to a defence capability gap. Unaddressed, it threatens to undermine the West’s ability to keep pace with our authoritarian and unconventional rivals as both sides strive to win the information war.

Synthetic environments (SEs) are a means by which we can start to fill this gap. They can help us gain a better grasp of the cognitive dimension by untangling some of the tricky problems that persist around it. But for these SEs to work, they need the latest, most reliable and most relevant data and content, coupled with the expertise to tie both into accurate models and analytical streams.

Representing online behaviours: simulating the cognitive dimension

Recently, I’ve had the pleasure of working with the bright minds at Montvieux. They specialise in using advanced AI and novel methodologies to give defence organisations a competitive edge in the cognitive dimension.

Montvieux’s Autonomous Social Agent-based Influence Range (ASAIR) is a Twitter-like offline social media platform consisting of autonomous agents that reflect the attitudes, sentiment and interactions of real-world social media audiences. End-users can glean the thoughts, feelings and emotions of these same audiences in response to a variety of scenarios.

The ASAIR is trained on anonymous observation of social and news media outputs of a particular region over a period of time. All content including conversations, themes and common sentiment is then synthetically generated in the range environment. What results is a library of anonymous, highly representative and high-fidelity data.

Users can inject artificial events into the environment to observe audience actions and reactions. If a simulated bomb goes off in the same region on which the ASAIR was trained, for example, do the agents respond with panic, patriotism, or protests? In each case, how do these reactions impact wider sentiment around warfare activities? Would – or should – this alter the course of action for either an ally or an adversary?

Elevating collective training for complex future warfare

The potential that Montvieux’s ASAIR holds for enhancing immersive, realistic collective training environments is profound. The virtual and cognitive dimensions of these are rarely addressed, if at all.

A burgeoning collaboration between Montvieux and Improbable Defence sees first-person training users share the experiences of a simulated population that is able to respond intelligently and accurately to information events, newscasts and social media feeds as they happen in real time.

To return to our opening example: an adversarial military vehicle driving through a city, injected as an artificial event and autonomously reported to the ASAIR as a social media feed, will cause an uptick in online activity. The simulated civilian population takes to social platforms to discuss its location and appearance. If an influencer wades in on the discussion, there’s every chance for the event to go viral.

The synthetic Twittersphere might hum with wary chatter: what does the appearance of that vehicle mean for the safety of my city? Neighbouring entities may ask: is it our turn next? Previously disparate values and beliefs start to coalesce and strengthen in the face of unfamiliar threats, constantly rebutted by adversarial bots and disinformation. Initially spurred online, many desire to express their discontent more actively.

In the case of the above scenario, online protestations begin to physically manifest in the first-person training environment. Geographically and culturally representative AI convene in relevant parts of the city, actively seeking to confront threats in the physical dimension. Crowds might form and violence might follow – the information environment absorbs it all. This action-consequence loop adds a substantial and profound layer of realism to scenarios and significantly amplifies the complexity of a collective training experience.