Q&A with Daden: the specialist business breaking new ground with synthetic social media
In this ongoing blog series, we look at how specialists across industry are using Improbable’s synthetic environment development platform, Skyral, to collaborate in new ways, and helping defence transform how it trains.
At age 12, sitting on his bedroom floor and surrounded by Airfix figures, David Burden’s military leanings stood out from the start. He’d go on to serve ten years in the British Army, tackling technical projects that would foreshadow his future work on emerging technologies.
Wargaming specialist, independent consultant and tech enthusiast, David founded Daden in 2004. He’s been helping organisations explore and exploit the social and commercial potential of using AI and virtual environments ever since. Applied to defence and national security, these technologies hold huge potential for immersive synthetic training environments.
Take chatbots. Most are designed to help, inform or educate. But some are intentionally malicious, built to spread distorted newspeak and propaganda. In an era more connected than ever, when operations are won and lost in the information domain as much as in the physical ones, it’s vital that training exists to better simulate the effects and influence of web-enabled disinformation. So we need malicious chatbots. And we need to harness the AI technology behind them to give users as real a training experience as possible.
During the early days, Daden and defence had little to do with each other. But when David received a call out of the blue from Dstl asking if Daden would like to bid for some work on conversational AI, all that changed.
Daden has since completed over two dozen projects for Dstl and MOD, a testament to the sector’s commitment to forging new and closer relationships with industry. They’ve worked across government, industry and academia to bring the latest innovations in VR and AI to wargaming, collective training and decision support. Their vision is defined by a strong belief in the power of platform-enabled ecosystems and what they can offer: increased choice, greater innovation and improved access to a wider range of larger programmes for industry specialists.
It’s a vision shared by Improbable Defence. Over the coming months, we’ll partner with Daden to explore the application of synthetic social media in a first-person training environment for added immersion and realism.
We caught up with David for more on Daden’s work, how it’ll fit into this project and what it means for the modern warfighter.
Q: Daden is unique in the defence space in that it’s a single-person business, with outsourced engineering. What challenges and opportunities has this presented for you?
A: Last year, I had the opportunity to restructure how I ran both my business and my life. Daden has always been a business that’s more focused on R&D over production. As a small company, this meant that having a permanent staff on the development side of R&D was always relatively precarious. By outsourcing Daden’s engineering, I can still work on what I enjoy around the thought process and design. But I can also collaborate with a company that takes care of the development side, which derisks things from my perspective. It also lets me leverage a larger pool of developers to work with depending on the thrust of the project. Ultimately it’s about more flexibility and reduced risk.
There are pain points, of course. Since I’ve no longer got developers in-house or on-hand, it’s that little bit harder to communicate. I need to be much stricter and clearer about the specifications for what I want developed, and very hot on checking what’s been done. Sometimes the developers tend to over deliver on some areas when I’d rather they invested time and effort elsewhere – but I can rarely blame them for sharing my excitement for a particular project, since what we work on is, in my opinion, endlessly fascinating.
Q: Daden has previously worked with Dstl. Are you able to reveal anything about this work?
A: This was 2014. Daden had been asked by Dstl to explore the application and potential of virtual humans for defence. This meant investigating how they can be used in both overt and covert ways – for example, as influence bots, or cyber-buddies to support military personnel. It was a Monday morning and I’d just sat down with my coffee. When I turned to page three of the Guardian, I saw Daden’s name across the spread with a detailed story on what we were working on, and suddenly my Monday wasn’t so mundane. Montvieux featured alongside us, too.
During this period, one of the projects we completed involved covert Turing tests. We had a theory, which turned into a paper: ‘The problem with the standard Turing test is that people know they’re taking part. This means that they ask questions of a bot that they wouldn’t ask of a person, which risks invalidating the test from the outset.’
We received funding from Dstl to test this theory in practice. Over the course of a week, consisting of three 1-hour sessions, a group of 40 students took part in what they believed to be online extracurricular activity. But they didn’t know that we’d added a chatbot to each session, masquerading as a student from another institution. Of the 40, not a single student recognised the chatbot as anything other than a real person. One even colourfully called it a ‘git’.
This spun into bigger projects around virtual humans that tied us more closely to the defence sector. MOD and Dstl tasked us with digital human capture. If a staff officer or commander was to move on or leave the force, for example, was there a way we could digitally capture them and so preserve their experience and advice for future personnel to use in training and wargaming?
‘We’ll do it,’ I said. ‘But we’ll do it badly to start with because the technology won’t let us do it well yet. Then we’ll do it better.’ Happily, both MOD and Dstl were on board with this approach and the project ran for about three years. In that time we worked with an individual to make a digital copy of them. Friends and family were able to talk to this copy and rate it to see how closely it resembled the original (and, by extension, how successful the project was).
The interesting thing about this project was that it led us into areas we hadn’t anticipated. Questions arose around digital immortality. What happens if that person passes away, for example? Who has ownership of the digital copy? Is it MOD? Or us – the creator? Perhaps their next of kin? Alternatively, can it have a life of its own, and what implications does this have for the memory of the real person? These sorts of questions can fill a book – and so they did. Together with Professor Maggi Savin-Baden, I wrote Virtual Humans: Today and Tomorrow, which not only explores the technical approaches to creating a digital human but also the ethical challenges that come with it.
Q: It sounds like these projects, alongside countless others, have naturally fed into the development of Daden’s Automated Social Media Synthetic Wrap (ASMSW). Can you speak to this journey and tell us a little bit about how the ASMSW works?
A: Over the years we’ve developed a good understanding of how bots should behave and communicate either between themselves, or over a representative social media platform. Gaining this understanding has been crucial as modern military operations increasingly happen in urbanised spaces. Even in rural areas, there’s going to be a civilian with a mobile phone, watching events unfold. Since social media is so pervasive, it’s only a matter of time before they share the event online and the message starts to spread.
So if you want to realistically model an exercise, you need to be able to replicate that information element alongside the physical domain. The traditional way of capturing this involves two or three content creators writing up posts and tweets and then feeding these into the exercise as injects. But this is typically costly and time-consuming to set up. It led the military to ask: is there a way we can use technology to do this? Can we create a system that generates social media based on what’s happening in the exercise?
We’ve done two projects for the MOD in response to these questions. The first involved capturing key environment parameters as part of a turn-based wargame system – things like are there any airstrikes or artillery fire? Is the electricity still on? Is the hospital overloaded? Based on these parameters, the ASMSW would then generate social media messages ready to be played out during the course of the following turn, giving players a sense of what’s happening in the human domain.
This had an immediate impact. It’s very easy in a military wargame to totally forget about the civilian population. But if you’ve got a screen with a Twitter-like feed, showing people complaining about the state of affairs, it brings home what’s actually going on.
We can even glean intelligence cues. If the ASMSW tells us that there are ‘tanks firing outside my street’, can we work out what street this is? Synthetic social media not only gives a broader sense of operating within the environment, but it also gives intelligence staff the opportunity to start practising their skills by taking intelligence cues and asking: can we make sense of that? Is it fake news? Can we corroborate it with something else?
This first iteration of the project took us to the end of 2021. By mid 2021 we’d already started work on a second version, driven by VBS rather than a turn-based system. We were able to use the same basic system but changed the way we triggered it, enabling the ASMSW to run in parallel with the soldier’s avatars as they move around the first-person training environment. If a tank were to drive up the road, for example, a civilian would tweet about it and the ASMSW would update in real time.
Q: Over the coming months, you’ll be collaborating with Improbable Defence and members of the Skyral Partner Network. Where does the value lie in integrating applications like the ASMSW and Montvieux’s Autonomous Social Agent-based Influence Range (ASAIR) with Skyral?
A: As I understand it, the ASAIR measures and represents the sentiment of a blueprinted audience. It aims to understand the attitudes of autonomous agents towards an ally, an adversary and the state-of-play.
What we’re able to do is provide simulated triggers to affect the sentiment of this synthetic audience. We’re able to baseline each of the agents within the ASMSW with a sentiment that’s designed to change over time depending on the sorts of activities that are happening. This sentiment will affect the words and expressions they use. Montvieux can then feed this into their ASAIR to effectively analyse and measure sentiment and how it affects autonomous agent activity within the information range.
By working together, we’re closing a feedback loop to better illuminate the cognitive domain. This gives personnel sharper insight into how shifting sentiment might affect decision making at the tactical and strategic levels. All of a sudden, there are lots of different facets of the command and decision making structure which you wouldn’t be exercising if you didn’t have that social media presence.
Q: How does this elevate the collective training experience?
A: There’s something to be said for making collective training more realistic by adding a deeper level of sheer noise and confusion through social media. To achieve this, we had to make the system dumber. A layman wouldn’t know what a T-72 tank is, for example. For the average Twitter-user, a tank’s a tank. To this end we gave each agent in the simulation a recognition skill for how good they were at recognising military vehicles. Most had little-to-none. This altered how they reported events online to better replicate real-life social media noise, and so create a more realistic collective training environment.
A lesson we learned is: most chatbots are caught out because they’re too clever. The way to simulate human behaviour is by making computers dumber. It’s about creating very human-sounding information by making it sound imprecise.
Q: What are the barriers to developing and delivering capabilities like this to defence organisations?
A: Lots of commercial barriers still remain. These are innovative requirements and solutions. All of the early development on them is going to be quite risky, so you need people with an appetite for risk to get them to a point where people say ‘there really is something in this’, leading to a willingness to invest in it in a more commercial undertaking.
In the past, Daden has typically struggled with defence contracts that don’t leave space for this. We might have completed a project for an organisation that people are excited by but never really goes anywhere for whatever reason. It’s better when one project leads onto something bigger and better, and those involved on both sides start to see the potential. Hopefully though we’re beginning to see more ‘pull-through’ from MOD. They’re starting to make earlier investments in more speculative work and better anticipate where these might lead to. This sort of approach is what led to our partnership with Improbable Defence, and no doubt will be the springboard for many similar industry partnerships and initiatives.
On a broader scale, such partnerships will bring better visibility to a whole range of innovative products and services. Defence doesn’t know what it doesn’t know. Organisations won’t build Daden’s offerings into their requirements if they don’t know what we make possible. But by partnering with Improbable Defence, our tools and services will feature on the Skyral Exchange, visible and accessible as and when needed.
Q: The project will involve integrating Daden’s tools and technologies with that of other industry partners via the Skyral Developer Platform. At this early stage, what’s got you most excited about the prospect of this collaboration?
A: We originally designed the ASMSW so that it wasn’t dependent on any one simulation system. This was key for us, and we’ve proven it by using it as part of both a turn-based and real-time system. Now we’ve got the chance to prove this again with the ASMSW working at scale via Skyral to deliver a serious quantity of synthetic social media messages throughout the system.
Besides that, we’re effectively creating value greater than the sum of the parts by integrating the ASMSW with Montvieux’s ASAIR. The collaboration will allow us to take triggers from one environment, create a social media feed out of these and then pass this onto another social media feed for analysis. Both sides are painting a smaller part of a bigger picture, but doing it extremely well according to our specialisms. This way of working has always matched Daden’s vision from the outset.
Q: What do you think are the benefits to the customer of a platform approach over a more traditional pipeline model?
A: We’ve always believed in doing things through APIs wherever possible. You’re far better off if you’ve got a particular bit of software which does one task very well and then connecting this to other parts to complete the system.
It’s all about flexibility. You don’t want a rigid system. When you don’t know what the future’s going to look like, the ability to swap elements out to reflect the changing operating environment is a key advantage.
Q: Speaking of the future, as the Fourth Industrial Revolution gathers pace, we’re seeing huge technological advancement across most sectors and industries. When it comes to the future of defence, what are you particularly excited about?
A: It’s fascinating watching technology shape and evolve over time. I’ve been working with virtual worlds since 1996. I talked about the metaverse at a 2008 conference in Amsterdam. Here we are, fourteen years later and on the cusp of seeing it pervade the common zeitgeist in a very powerful way.
Things are certainly gathering momentum around these virtual worlds and environments. Now more than ever, their potential for use across defence and national security is being explored by some of the industry’s brightest minds. Should we harness the power to train, plan and test scenarios in a risk-free virtual world, our actions will be that much more effective and impactful in the physical one – and that’s a future I’m certainly looking forward to.
The Skyral Partner Network consists of organisations like Daden that share our commitment to developing the best possible capabilities and getting them into the hands of end users fast.