We’ve published another study. I thought I’d share it here. This new study investigates how symptoms of PTSD, cannabis use, and negative emotions vary on a day-to-day basis among US military veterans. We used ecological momentary assessment (EMA) to gather real-time data from participants multiple times a day, providing a nuanced understanding of how psychological states and behaviours shift in close proximity to each other.

What makes this work particularly exciting is the use of MAVERICK (built on the Ration framework), a custom-built app developed to collect high-frequency, in-the-moment data from users. MAVERICK has been designed to integrate with digital devices and wearables, offering a scalable way to track mental health and behavioural patterns over time. This study showcases how MAVERICK can support mental health research through reliable, user-driven data collection outside of traditional clinical environments.

The findings suggest that on days when veterans experienced more severe PTSD symptoms, they were more likely to use cannabis and report greater negative affect. Importantly, the study moves beyond cross-sectional assumptions to highlight the dynamic interplay between mental health symptoms and substance use in everyday life.

This work adds to a growing body of research using digital tools to better understand mental health in real-world contexts. It also demonstrates how apps like Maverick can support longitudinal behavioural health monitoring, particularly in populations with complex needs.

Read the full article here: Daily associations between post-traumatic stress disorder, cannabis use, and negative affect among veterans

How do we prepare minds for conflict? And how do those preparations ripple outwards, shaping the society that supports – or questions – military service? On Thursday 17 July 2025 in London, we will bring historians, health researchers, serving personnel, veterans and policy specialists together to explore these questions from both ancient and contemporary perspectives.

Hosted by the King’s Centre for Military Health Research (KCMHR) and University of Bristol, this workshop forms part of the interdisciplinary impact project "Preparing the Body and Mind for War in the Ancient and Modern Armed Forces".

Why attend?

• Fresh perspectives: hear how classical sources on warfare foster modern debates about recruitment, training and societal attitudes to those who serve.
• Cross‑disciplinary dialogue: mix with scholars of history, psychology, public health and defence studies, as well as practitioners and veterans.
• Policy relevance: discuss how historical insights can inform today’s approaches to supporting military personnel before, during and after service.

Programme outline

• 10:30 – Arrival & coffee (Bush House, King’s College London)
• 11:00 – Welcome and opening reflections
• 11:15 – The ancient roots of military mindset – lightning talks (5–7 mins) and round‑table discussion
• 12:30 – Sandwich‑platter lunch & networking
• 13:00 – Modern contexts and societal influence – discussion with Q&A
• 15:15 – Synthesis: bridging past and present
• 16:00 – Prompt close

Who should join?

Anyone with an interest in the interface between military service and wider society can attend this free event. This includes:

• Academics in history, classics, psychology, public health and defence studies.
• Serving personnel and veterans.
• Health professionals and charity representatives.
• Policymakers shaping Armed Forces welfare and public engagement.

Registration

Places are free but limited. To reserve a seat (in person) please email Dr Daniel Leightley (Department of Population Sciences) at daniel.leightley@kcl.ac.uk AND Dr Hannah‑Marie Chidwick (University of Bristol) at hc6198@bristol.ac.uk by Monday 30 June 2025.

Something a little different from me today. If you’re into collecting Pokémon cards (like me), you’ve probably heard of WhatNot, a live auction and shopping platform that’s part eBay, part Twitch stream, and 100% chaotic (and fun). I dove in recently to explore the world of live breaks, graded cards, and last-second bidding wars. Here’s what I learned; plus some top tips if you’re thinking of jumping in. I was keen to expand my Pokémon TCG collection with some vintage cards.

What Is WhatNot?

WhatNot is a livestream auction platform where sellers (often collectors themselves) showcase Pokémon cards and products in real-time. You can:

• Buy single cards, sealed products, or graded slabs
• Join live ”breaks” (more on this below)
• Chat with sellers and other collectors
• Bid in fast-paced auctions, some as short as 5 seconds!

It’s fast, exciting, and occasionally overwhelming. But once you get the hang of it, it can be a great way to find deals, or just join a community of collectors.

Key lingo to know

If you’re new, the chat and seller talk can feel like a different language. Here’s a crash course:

• Break: A seller opens sealed Pokémon products (e.g. booster boxes) live. You buy a “spot” and get the cards pulled from the packs assigned to you. Popular formats include rip-and-ship (you get everything opened on your behalf) and pack battles (winner takes all).
• GG: Short for Gold Grading or Graded Gem, depending on context. More casually, it’s used to say “good game” when something great is pulled or won.
• Razz: A raffle-style listing. You pay a small amount for a slot in a draw, and the winner gets the item.
• Slab: A graded card encased in a plastic case (from PSA, CGC, BGS, etc.).
• BIN: “Buy It Now” - a fixed-price item rather than an auction.
• Sniping: Jumping in at the last second to win an auction. It’s part of the game and can be thrilling or annoying depending on which side you’re on.

Top tips for buying on WhatNot

1. Do Quick Research - Fast - Auctions often last just 15 to 30 seconds, so you need to be able to instantly judge value. Have a second tab open with eBay sold listings or TCGPlayer to check prices. It’s very easy to overpay if you get caught in the hype of the swipe.

2. Understand the Format of Breaks - Before buying into a break, read the listing carefully (often in the show notes):

• Are you getting just the hits? (i.e. no bulk commons/uncommons)
• Do you get reverse holos?
• Is shipping included?
• Is the break randomised (e.g. random energy) or do you pick your pack?

Always ask in the chat if you’re unsure, most sellers are happy to explain.

3. Be Cautious About Slabs - Graded cards can be a bargain on WhatNot, but not all slabs are created equal. PSA, CGC, and BGS are generally the most respected brands. You’ll also see a lot of lesser-known companies like GetGraded or other small-scale UK-based graders. These often don’t hold the same market value or buyer confidence, and some sellers appear to mass-sell these slabs cheaply to shift stock. Always check resale value on eBay before bidding on a slab, and if in doubt, pass.

4. Set a Budget and Stick to It - It’s easy to get carried away; especially during high-energy breaks or when you’ve just missed a great pull. Have a spending limit in mind and pace yourself. It’s a marathon, not a sprint.

5. Follow Trusted Sellers - Look for sellers with:

• Good ratings and reviews
• Transparent listings (no vague “mystery packs” without value range)
• Professional setup (clear camera, good lighting, no “lost” pulls)

You can follow your favourites and get notified when they go live.

6. Enter Giveaways - Smartly - Giveaways are common and can be a great way to snag a free item, or more importantly, free shipping. My personal strategy: enter a giveaway, and if I win, that unlocks free shipping for the entire stream. I’ll then make my purchases after that. It’s a simple tactic that can save you money. But be mindful, some sellers do then go on to charge a small shipping fee if you do make a large number of purchases.

7. Be Aware of Shipping Costs - Shipping is usually calculated per seller, not per stream. If you’re buying several items from one seller, this can work in your favour. If you’re buying one card from five different sellers, less so.

8. Don’t Skip the Small Streams - Some of the best deals and genuine communities are in smaller streams with fewer viewers. Less competition = lower prices and more chances to interact directly with the seller.

Final thoughts

WhatNot brings a real social and adrenaline-fuelled aspect to Pokémon collecting that’s hard to replicate. I’ve had both wins and near misses, but each stream teaches you something. Whether you’re in it for nostalgia, investment, or fun, knowing the basics (and the lingo) makes a huge difference.

Taking on a marathon was, without exaggeration, one of the toughest challenges I’ve faced. I picked it precisely because it was beyond my comfort zone and I wanted to see how I did. I thought I’d share some of my thoughts of the run and tips leading up to it.

• Start early and stick to a plan to hit 30km before race day. I’d convinced myself that doing a half-marathons meant I was ready. Wrong. At 28km I went into the wall: legs cramping, brain fog for what next and self-doubt. The physical and mental challenge was hard. Meeting that moment in training would have helped.
• Run in all kinds of weather. During training I was a fair weather runner. Come race day we got hail, sheets of rain, then blazing sunshine (I did burn). My soaked shoes filled and I’d never practised the shuffle strides to get water out. Practise this, and take sun cream.
• Layer up like the military. I shivered on the start line, then baked mid-course, then the rain slapped me and I couldn’t even see. The hail brutally assaulted me. I wish I had packed a rain coat, or at least had a runner pack to store this in.
• Test the food beforehand. On a training I used salt tablets and gels; they were magic on the day. Important to test this beforehand to see how the tummy handles it.
• Bring tech with real power (or the ability to power up). My Apple Watch died at 30km, which meant my music went and I lost my pace maker. This made it hard to pace and I really went overboard. Check your technology in advance.
• Respect the runners’ code. A few people zig-zagged without looking and two went down hard. This helps my strategy early on, go slow to avoid these dangers and then speed up.
• Sip, don’t gulp. Volunteers handed us 250ml bottles of water. First station I gulped half of the bottle of water and paid with stabbing stitches. The fix: two mouthfuls, bottle over the head, move on. Be smart with this.
• Use the loo. Always use the loo. Comfort beats pain while running. And the queues weren’t that bad. The staff did a great job.
• Prepare for vomit (well for me). I crossed the line feeling relief, then a St John Ambulance volunteer offered a sick hat. I filled it immediately, thanked her and nooted off. I felt fine after, I just didn’t think I needed it, but they clearly felt I did. Happens I guess when pushing yourself so hard.
• Cramps are impossible to manage. My Calf seized at around 32km; I hobbled and stretched. I carried ibuprofen to mask the pain. Not a smart move. But it was my only solution.

Those 42.2 kilometres were brutal, but I would recommend running in the Edinburgh Marathon! The staff, the route and the support were excellent. Really made all the difference to moral. 

We’ve just launched a new survey — and we need your help.

Digital technology is a huge part of everyday life. From booking a GP appointment to managing benefits, keeping in touch with family, or watching the news; it all happens online. But for many veterans in the UK, getting online isn’t always easy. Some don’t have the right equipment, others struggle with internet access, confidence, or simply knowing where to start.

That’s where the Veterans’ Digital Needs UK Survey comes in.

We want to understand how veterans are using (or struggling to use) digital tools and services. Our aim is to make sure no veteran is left behind in our increasingly digital world.

Who can take part?

• If you’re a veteran of the UK Armed Forces (including National Service), aged 18 or over and living in the UK; we’d love to hear from you.
• If you support a veteran — maybe you're a family member, friend, or carer; you can complete the survey on their behalf too.

It doesn’t matter whether you’re online all day or barely touch a computer — your experience is valuable, and your voice counts.

What’s involved?

• It’s a short, anonymous survey (just 10–15 minutes).
• You’ll be asked about your service history, health, and how you use digital technology.
• There’s a chance to win one of ten £50 Love2Shop vouchers as a thank you.

Why it matters

Your feedback will help shape better support, improve digital services for veterans, and guide national policy.

Take the survey

If you have any questions, feel free to reach out to us at ukmrg@kcl.ac.uk or reach out to Dr Daniel Leightley.

Thank you for helping us build a more connected, inclusive future for veterans.

The goal for this trip was twofold: cycle around Guernsey and carve out time for a writing retreat. The cycling part was a clear success; the writing, however, didn’t go quite to plan. I probably set the bar a little too high. This trip forms part of my broader ambition to cycle all cyclable parts of the UK and the Channel Islands islands (simply anything I can cycle in a loop 20km or more, but as an island).

Travel to Guernsey 

I arrived at Poole Harbour for a Condor Ferry, where the staff were incredibly friendly. The process was straightforward: check-in, police check, then queue for the ferry. Boarding the Condor Catamaran was quick and easy. I was impressed by the drive-on/drive-off loop system, a smart bit of design that keeps traffic flowing.

Being the smaller Catamaran service, only basic hot food (courtesy of a TurboChef oven) was available — not the full menu you get on the larger ferry. The crossing itself was quite choppy. I sat in the middle of the vessel and focused on work to fend off seasickness. Whether it worked or not is still up for debate. The staff were brilliant, particularly with younger passengers, offering advice and tips to ease any seasickness worries.

The Cycle

This was the big ride — 41.91km around the island. Despite the rain and some traffic, it was a manageable and enjoyable day out on the Brompton. I began at Le Pommier Hotel & Restaurant and cycled up to Cobo Bay, then followed a clockwise route around the island, sticking as closely to the coast as possible without venturing off-road. (note: Bromptons are not made for gravel adventures I have found out)

The ride was fairly straightforward until I hit the incline near Torteval — fine for a road bike, a bit tougher on a folding one. I’d originally planned to swing out to Jerbourg, but the heavy rain persuaded me to press on directly to St Peter Port. The descent into the town is sharp and requires firm braking; worth noting if you’re doing the same route.

I stopped at Coco for a well-earned coffee — a lovely spot, busy but welcoming to laptop users. From there, I continued to Mont Crevelt, Fort Doyle, Rousse Tower, and followed the coast back to the hotel.

This was one of the shortest island loops I’ve done, but it’s a great ride. Roads are tight and compact, yet very cycle-friendly. You can see my full Strava route here.

The Return 

The ferry departure was moved an hour earlier due to “adverse” weather; always adds a touch of stress when plans shift suddenly. Thankfully, the crossing wasn’t too rough and the weather cleared as we went. Again, the Condor Ferries team were fantastic; cheerful, helpful, and clearly enjoying their work, which made the experience much better.

Tips

• Island speed limits range from 15–35mph.
• Fuel is more expensive than on the mainland — arrive with a full tank if driving, though you likely won’t use much during your stay.
• You can comfortably see Guernsey in three days, but I spent five — partly to explore the World War Two sites and partly to relax and catch up on work.

Food recommendations:

• The Rockmount
• Le Pommier
• Imperial Hotel

WW2 highlights:

• German Signals HQ
• The Occupation Museum

There’s plenty more to see as you cycle the island — these are just the standout ones for me. 

Moral injury, the inner turmoil arising from betraying one's own ethical code, is gaining traction in both military and civilian contexts, and research as well as clinical settings. However, unlike PTSD or depression, many argue that moral injury is not a medical disorder to be diagnosed or treated with medication or given the person a label. Instead, it is considered a natural human response to extreme moral conflict, a "wound of the soul" or "violation of one's beliefs”, that calls for understanding, meaning, and ethical growth, rather than clinical medicalisation.

Understanding Moral Injury: A wound to the conscience

Moral injury occurs when an individual does something (or fails to do something) that deeply violates their moral beliefs, or when they witness others (or leaders they "trusted") committing moral wrongs. It's the gut-wrenching guilt or shame experienced when one's inner self screams "I did wrong" or "I let this happen". For example, a soldier might experience moral injury after following a lawful order that resulted in civilian casualties, or a nurse might feel it after being forced to prioritise one patient’s life over another due to limited resources, such as during the Covid-19 pandemic.

It is important to note that moral injury is not currently classified as a mental disorder. It does not appear in the DSM, and one cannot be "diagnosed" with moral injury. PTSD, on the other hand, is an established mental disorder characterised by fear-based symptoms such as flashbacks and hyperarousal resulting from real and measurable life-threatening trauma. Moral injury has a different psychological framing; it is less about fear and more about guilt, shame, violation of one’s ethics or beliefs, and a crisis of trust in oneself or others. It has been described that PTSD is a racing heart, and moral injury is a broken soul.

For veterans, moral injury is considered a "dimensional problem"; a spectrum of inner conflict with no clear threshold. Moral injury is not pathological – rather, it is a human response to a moral breach. Unlike PTSD, the nervous system is not on high alert; instead, the hallmark is a collapse of trust and questioning, either in oneself, in others, or in the wider world.

In short, moral injury isn’t about a hardwired dysfunction; it’s about one’s conscience reacting to something that went against their individual values, standards, and moral code. Rather than medical symptoms, individuals experience moral emotions: remorse, shame, anger, and disgust at their own actions or those of others. These feelings can be intense and painful, but many argue they are also fundamentally normal. In essence, these feelings are a normal part of everyday life; indeed, experiencing them indicates that a person's moral compass is still functioning and that they have a conscience.

Impossible choices: Moral Injury in war and healthcare

War often places soldiers into impossible moral dilemmas. One scenario used during officer selection describes a person in charge of two outbound flights in a war zone, one carrying badly needed blood for transfusions, the other carrying reinforcements and ammunition for troops under fire. Only one plane can take off in time due to weather. Whichever is chosen, at least one life will be lost. How does one decide? How does one live with that decision? Many personnel carry the weight of such choices long after the decision has been made.

For some personnel, these experiences leave a “soul wound”; a deep sense of having violated “what’s right”, even if they followed orders. Shay in 2014 described a form of moral injury as “a betrayal of what’s right, by someone who holds authority, in a high-stakes situation”. In war, that betrayal might even be by oneself; individuals do what they have to do, but may still feel they betrayed their own values in the process.

Real-world examples are vast. But focusing on the war setting, the first accounts and reflections come from the US war in Vietnam. US soldiers participated in several massacres and atrocities during the war, some without any apparent remorse or guilt. However, others who attempted to intervene and stop these events reported feeling guilt for years, believing they couldn’t stop the killings sooner.

One example is that of Hugh Thompson, who intervened to stop a massacre and suffered for many years afterward, feeling that he bore a moral burden. This was despite the fact that he did the right thing by trying to stop the massacre. Witnessing, feeling of duty, and responsibility to prevent it from happening left him feeling morally injured. This illustrates that moral injury isn’t about legal guilt; it’s about ethical guilt. Individuals often hold themselves to high standards of right and wrong, sometimes unfairly.

Moving away from a military context, some of the clearest examples of moral injury have emerged in health, notably during and after the Covid-19 pandemic. Medics found themselves in difficult situations: too few ventilators or beds for too many patients, guidelines requiring them to delay or deny care to people who might have lived if resources were available, or watching patients die alone without friends or family due to quarantine rules. These aren’t just “stressful work conditions”, they are moral challenges.

Frontline health workers reported feelings similar to those of morally injured soldiers; guilt, shame, and a sense of betrayal. For example, an ICU nurse might think, “I let that patient die because we had to ration care; that goes against everything I believe as a nurse”. Or a doctor needing to make a triage decision might be haunted by thoughts such as, “did I do the right thing? Did my choice cost someone their life?” One survey of UK healthcare workers during the pandemic found a spike in experiences of betrayal: nurses feeling let down by their seniors, doctors feeling society abandoned them by ignoring public health advice.

These examples highlight the universality of moral injury across professions involving high-stakes and often quick decision-making. Whether in war or healthcare, individuals who hold frontline public roles face ethical dilemmas that can leave lasting emotional and hidden wounds. However, moral injury is not necessarily a sign of moral failure; it can also indicate deeply held values and ethical integrity.

Moral Injury: A human struggle, not a disorder

Moral injury is real, but it is not a disorder. It is not a medical condition requiring diagnosis and treatment, but rather a moral and ethical struggle best addressed through community, reflection, and meaning-making. A psychiatrist who worked with Vietnam veterans saw how some wounds went beyond PTSD. He described moral injury as a “violation of what’s right”, separate from psychiatric diagnoses. PTSD is about fear; moral injury is about losing trust and meaning. Recovery, he argued, comes from storytelling, communal reflection, and ethical reckoning; not medication. 

Further research has cautioned against medicalising moral injury. Standard PTSD treatments such as exposure therapy or medication do not address guilt and shame. Instead, approaches like “adaptive disclosure” focus on ethical reflection and making amends. This position appears to be endorsed by the U.S. Department of VA, which acknowledges moral injury but states explicitly that it is “not a mental disorder”. Many veterans reject medicalisation, fearing it suggests that guilt is pathological rather than a natural response. The VA has explored alternative support, including peer groups, chaplain services, and ethical discussions.

One philosopher views moral injury as “part of the human condition”. Unlike PTSD, which is often viewed as a private struggle, moral injury is relational; about navigating guilt, responsibility, and justice. Healing is not about erasing guilt but learning to live with it and use it as a moral guide. Moral injury is painful, but it is not an illness; it is a sign of a working conscience. Those affected do not need a diagnosis but recognition, space for reflection, and pathways to make amends. By understanding moral injury as a human experience rather than a disorder, society affirms our collective ability to wrestle with right and wrong and grow from it.

Addressing moral injury requires more than just individual resilience; it demands systemic support, open dialogue, and acknowledgment from institutions and society. By fostering environments where these struggles can be discussed and validated, those affected are able to process their experiences in ways that promote healing and personal growth, rather than isolation or guilt.

Finding relief: Recovering from Moral Injury

If moral injury isn’t a disorder, how do people recover or find relief? Research has highlighted key elements that help soldiers, doctors, and others struggling with a damaged conscience. Acknowledgment and open discussion is important. Many suffer in silence, believing they are alone in their guilt and shame. Simply naming the pain can be a relief. Some military units now incorporate reflections into after-action debriefs, while peer support groups in health create spaces to share moral dilemmas, reducing isolation and self-blame.

Self-compassion and forgiveness help ease moral distress. Those with moral injury often hold themselves to impossible standards, focusing on perceived failures; “I should have done more”. Research shows self-forgiveness can foster growth. Therapists often ask veterans to consider what they would tell a fellow soldier in the same situation; usually, they offer understanding, not judgment. Also, making amends and finding meaning can be transformative. This may involve direct restitution, such as an apology, or symbolic action, like advocacy or humanitarian work. Philosophical and faith-based perspectives can help reframe guilt as a commitment to do better, shifting the focus from self-blame to responsibility.

Community support is an important aspect. Moral injuries often fester in secrecy, but healing happens in connection with others. Support groups for veterans, health workers, and first responders provide non-judgmental spaces for sharing experiences, reinforcing that moral distress reflects integrity, not failure. This also includes finding solace in religion. Research has linked unaddressed moral injury to increased suicide risk, making peer support vital. Systemic and leadership actions also play a role. Moral injury is often created by institutions, not just individuals. When leaders acknowledge that personnel were placed in impossible ethical situations, whether in combat or understaffed hospitals, it validates their suffering and helps prevent future crises.

Conclusion

In conclusion, moral injury is a hard truth of life and is inevitable, especially in professions such as the military or health, where life-and-death decisions occur and must occur. It’s real, it’s painful, but it’s not a mental illness in the typical sense. It’s the by-product of having a moral conscience in an imperfect world. While suffering shouldn’t be romanticised, it also doesn’t need to be pathologised away. As long as there is war, as long as there are emergencies, as long as tough choices must be made, there will be moral injuries; moments when people feel crushed by the weight of doing (or not doing) something under difficult conditions.

The silver lining is that where there is moral injury, there is also moral courage and integrity. The very existence of moral injury demonstrates the goodness in people, the aspiration to be just, loyal, and compassionate with clear values and standards. Supporting moral injury taps into that same goodness: through empathy, honesty, and striving to make things right. 

It reminds us that being human is a moral endeavour, and even when we falter, we can find our way back to what’s right with the help of others. In the end, moral injury is not a sentence to lifelong guilt; it can be a journey to deeper wisdom and strength. By sharing and listening to these stories, whether from the battlefield or hospital ward, we all become a bit more human, and a bit more equipped to do what’s right when it matters most.

Repost from UK Military Research Group.

After the Christmas rush, I was keen to find somewhere reasonably warm, affordable, and easy to reach. Gibraltar, often referred to simply as “Gib,” checked every box. Tucked away on the southern tip of the Iberian Peninsula, this British Overseas Territory combines a dash of the UK (think red post boxes, full English breakfast and paying in Pounds) with plenty of Spanish influences. The flight from London Gatwick via EasyJet only took about two hours and thirty minutes. Here’s a rundown of my visit, plus some handy tips for anyone looking to plan a short getaway.

Why Gibraltar?

Gibraltar’s tiny size (a jog around The Rock is about 20km) makes it perfect for a long weekend or a quick city break. English is widely spoken, though you’ll also hear Spanish. Also another handy element. Pounds are accepted everywhere and no card processing fees.

How long to stay?

You really only need three to four days. That’s enough time to see the main attractions, wander through the city centre, and even pop over the border into Spain. Despite Gibraltar’s compactness, there’s plenty to keep you busy, from the heights of the Nature Reserve to the waterfront promenade.

If you stay longer, you might want to look into additional activities, like trying out local spas or swimming pools. However, walking options can be limited because Gib is so compact: your best longer walk will be The Rock itself.

Getting around

I chose to walk most of the time because Gibraltar isn’t very big. Taxis are available but can be expensive considering the distance (although traffic was very light when I visited). Another option is renting a scooter through a company called Hoppy. It’s £1 to unlock, then you pay per mile (£0.25). A fun, zippy way to explore, but watch the costs if you plan on riding a lot. Many of the locals have their own scooters. But mindful, these can only be ridden on the road, and a helmet is required. You do get a helmet with Hoppy. 

Quick trip over to Spain

If you fancy a cross-border jaunt, it’s very straightforward to walk into La Línea de la Concepción, the Spanish town next door. The guards asked why I was visiting, so I showed them a walking route I’d planned. That was enough to get two stamps (one in, one out) in my passport. Some locals have reported issues trying to cross the border. It does seem some post-Brexit tension still remains. Therefore, the advice I had was to ensure you had a detailed plan of what you were intending to do in Spain. Visitors and locals are getting turned away daily.

The Rock: Gibraltar Nature Reserve

No trip to Gib is complete without tackling The Rock. I bought the £25 Nature Reserve pass, which granted access to:

• Thrill Seekers Trail and Mediterranean Steps – Amazing coastal views, but be prepared for steep climbs and it isn’t for the faint hearted. It is a challenging hike.
• St. Michael’s Cave – Impressive limestone formations and a light show inside. It is a shame that they’ve placed light effects everywhere, it means you only see the light effects and not the natural effect,
• O’Hara’s Battery, WWII Tunnels, Great Siege Tunnels, Moorish Castle – This was a great adventure of Gib’s role in the War. It had a significant role and the extensive tunnel network is impressive. You can only walk a section of it; but gives you a sense of what it must have been like during the War. A fun little fact. There are more tunnels than roads in Gib.
• Skywalk and Windsor Suspension Bridge – Heart-in-mouth moments with unforgettable panoramic views. When I went it was a bit windy; but hold you nerve and you can get some great snaps on the bridge

I walked about 20km in one day here, so sturdy shoes are a must. Also, the resident Barbary macaques are cute but can get pushy if they suspect you’ve got snacks.

City Centre vibes

Gibraltar’s central district is set within old fortress walls, so it’s brimming with history. You’ll see familiar British chains like Costa and Pizza Express alongside smaller local restaurants. Plenty of pubs selling UK beers at very affordable prices. Whether you’re after a quick coffee or a hearty meal, you’ll find plenty of choices. It’s a lovely place to wander around, shop duty-free, and soak up the mix of British and Mediterranean cultures.

Running in Gibraltar

I’m an avid runner, so I hit the pavements (and occasionally the tunnels) for a few different routes. A word of caution: some roads just end abruptly, so a bit of planning goes a long way. Locals run through Devil’s Tower Road Tunnels, though it’s not officially advised: if you decide to do it, make sure you’re visible and prepared to hug the wall when cars pass. Also, keep a few coins handy if you need water breaks along the way.

Final thoughts

For such a small place, Gibraltar delivers an outsized travel experience. You can walk across an airport runway, explore war tunnels, admire incredible sea views, and even hop over to Spain if you feel like it: all within a few days. I’d recommend three to four days to take in everything without feeling rushed.

We are pleased to announce the release of the Consensus Roadmap for the Use of AI in the Armed Forces Charity Sector. This roadmap, developed through a collaborative event held in October 2024, offers insights and guidance on how charities can approach the use of AI in a responsible, transparent, and effective manner.

The roadmap comes at an important time, as Armed Forces charities face growing demands and financial pressures. This consensus statement aims to support charities in understanding the opportunities and challenges associated with AI adoption and offers a pathway for responsible use.

Insights

The roadmap draws on insights shared at the Exploring the Role of AI in the Armed Forces event sponsored by the Forces in Mind Trust, Combat Stress and King's College London. Leaders from academia, government, industry, funders, and the charity sector came together to discuss how AI could impact and support Armed Forces charities. Through expert talks, roundtable discussions, and attendee feedback, key considerations for AI adoption were identified and synthesised into this roadmap.

Speakers included Michelle Alston (CEO, Forces in Mind Trust), Dominic Murphy (Head of Research, Combat Stress and King’s College London), Zoe Amar (Founder, Zoe Amar Digital), Stella Harrison (Senior Analyst, RAND EU), Daniel Leightley (lecturer, King’s College London), Stuart Middleton (associate professor, University of Southampton), Samantha Ahern (Senior Digital Research Trainer, UCL), and Nicholas Cummins (lecturer, King's College London). Their insights underscored the importance of ensuring that AI adoption aligns with the needs, values, and ethical obligations of Armed Forces charities.

Key takeaways

1. Current use of AI by charities

• 61% of Armed Forces charities currently use AI, primarily for operational tasks like automating routine processes.
• There is a significant gap between operational AI use and strategic AI adoption, highlighting the need for increased understanding of AI’s capabilities and limitations.

2. Barriers to AI adoption

• Limited AI and data literacy across the charity sector.
• Concerns around data privacy, GDPR compliance, regulation and the responsible use of AI.
• A lack of clear guidance from government bodies, the Information Commissioner’s Office (ICO), and the Charity Commission.

3. Opportunities for AI

• AI can streamline beneficiary support, triage services, and provide quicker, more tailored responses to those in need.
• Automation of administrative tasks like meeting minutes and document processing allows staff to focus on higher-value strategic activities.
• By analysing donor behaviour, AI can tailor engagement strategies, identify potential supporters, and optimise fundraising campaigns.

4. Responsible AI considerations

• Transparency, fairness, and accountability are essential for AI systems, especially when handling sensitive personal data.
• The use of AI in mental health support must be complemented by human oversight, particularly in high-stress and crisis situations.

A need for greater clarity and support

The roadmap highlights the need for greater support from government, regulators, and collaboration across the industry to ensure responsible AI adoption in the Armed Forces charity sector. Our key recommendations include:

• Upskilling charity leaders and staff so they understand AI’s risks, responsibilities, and potential.
• A focus on practical accountability to ensure models are transparent, fair, and aligned with charity sector values.
• Promoting partnerships between Defence, charities, and industry to support inclusive AI adoption and share best practices.
• Encouraging the use of AI that aligns with sustainability goals and contributes positively to environmental and social objectives.

Looking ahead

The updated Charity Digital Code of Practice (2025) will offer additional guidance on digital transformation and AI use. While not a set of mandatory rules, the Code will support charities in navigating complex issues like leadership, data security, and digital skills development.

The Consensus Roadmap for the Use of AI in the Armed Forces Charity Sector is a valuable resource for charities at all stages of their AI journey and to create discussion. By engaging with the roadmap, charities can better understand the potential of AI, address key barriers, and ensure that AI adoption aligns with the needs of the communities they serve.

Join the discussion

We invite all stakeholders — charity leaders, funders, regulators, and beneficiaries — to review this roadmap and share their feedback. This is an opportunity to contribute to a shared understanding of how AI can best support the Armed Forces charity sector.

To learn more about the roadmap or to get involved, please contact Daniel Leightley (daniel.leightley@kcl.ac.uk) or Dominic Murphy (dominic.murphy@combatstress.org.uk).

Earlier this week, academics came together with charity leaders, industry members and civil servants for an event at King’s College London that explored opportunities for the use of artificial intelligence (AI) in the Armed Forces sector.

Organised by the Department of Population Health Sciences at King’s in partnership with Combat Stress and funded by the Forces in Mind Trust, the event aimed to build a consensus within the Armed Forces charitable sector on the use of AI. 

There is currently little guidance on the use of AI within the charitable sector. According to a survey conducted by the Department of Population Health Sciences ahead of the event, there is fear around its use within the Armed Forces charitable sector. This means there could be missed opportunities to use to AI to improve charity functions such as resource allocation, service delivery and engagement with clients.

During the event, the 40 attendees discussed how charities within the Armed Forces sector could make use of AI, what barriers exist for the use of AI, and what policies should be in place for its use within the sector. 

“The event focused on the opportunities of AI for the Armed Forces sector, but also on some of the risks. It’s about getting a balance between the risks and opportunities and this event aimed to help us develop this consensus,” said Dr Daniel Leightley, Lecturer in Digital Health Sciences at King’s.

Dr Leightley’s research is focused on military health, including on how technology can be used to improve the health and wellbeing of serving and ex-serving personnel.

“The attendees were a mix of academics and charity, industry and government members, which made for really lively discussions with broad perspectives,” he added. 

The event was split into three sessions. After a welcome by Professor Dominic Murphy, Co-Director of King's Centre for Military Health Research and Head of Research at Combat Stress, the first session focused on what AI is and how it is currently being used in the sector.

Dr Leightley, spoke about ‘Demystifying AI: Separating Hype from Reality.’ His talk highlighted the hype around AI within the media, the real capabilities of AI, and its present limitations while calling for further improvements on guidelines, technology access and collaboration across the sector.

The session also included a talk from Zoe Amar FCIM, founder of Zoe Amar Digital, who discussed how AI is being adopted within charities. Her presentation highlighted the growing use of AI, but pointed out that while AI offers opportunities, many charities still face barriers, such as a lack of skills and concerns over data privacy, GDPR, and ethical considerations. Also, Dr Nick Cummins, Lecturer in AI for Speech Analysis at King’s, provided a practical example of how AI can be used to analyse speech as a digital biomarker. His work emphasised the potential of AI in clinical decision-making, while also acknowledging the challenges, including variability in speech analysis and the need for reliable datasets.

The second session explored the opportunities and risks associated with using AI. Dr Stuart Middleton, Associate Professor at the University of Southampton, presented on the capabilities of Large Language Models (LLMs), focusing on their potential applications in areas such as mental health and defence, while acknowledging the challenges posed by model biases and the need for responsible AI integration​.

Samantha Ahern from the Centre of Advanced Research Computing at UCL, discussed the complexities of designing and implementing AI tools, addressing issues such as the risks of foundation models and the importance of explainable AI in healthcare​. Finally, Dr Stella Harrison, Senior Analyst at RAND UK, explored AI regulation, highlighting challenges such as the ‘pacing problem’ and cross-border consensus, alongside the implications of the EU AI Act and UK regulatory frameworks​.

 “Understanding the opportunities AI can offer and the risks that need to be managed is key for those working with and supporting the Armed Forces community. It was great to support an event starting the conversation between senior leaders to explore these issues and discuss how we can most effectively utilise and manage AI to support our community,” said Michelle Alston, Chief Executive of Forces in Mind Trust.

The final session of the day was a roundtable discussion, where attendees could delve deeper into the topics of the day. A live scribe was present at the event, who created an illustration of the conversations. 

Some of the important points raised included: the need for an ethical framework and the right policies to put AI into practice within the sector; how can safety be ensured when using AI; the need for appropriate training so that people can make use of AI safely and confidently; and the need for more collaboration within the sector to understand the opportunities AI can have to improve services for clients. 

The findings of the roundtable discussions will be aggregated to develop a consensus statement and shared with the community later in the year.

Reposted from This week at King's. Original here.

The Ministry of Defence recently released a call for evidence for the upcoming Strategic Defence Review. You can read more here. The following is my position statement in response to this call. This is a collective submission with Colleagues from King's College London.

The following contributed to this response: 

• Dr Daniel Leightley, Lecturer in Digital Health Sciences at King’s College London
• Professor Vasa Curcin, Professor of Health Informatics, King’s College London

This response has the following core themes:

• Theme 1 - Operational readiness and personnel sustainability: Prioritising the physical and mental health of personnel is essential for maintaining a capable Defence. By reducing the incidence of injury, long-term sickness (P0), and medical discharges, we can promote a healthier environment. Detecting condition onset, and delivering early intervention is key to this ambition.
• Theme 2 - Inclusive Force development: Promoting gender and health equality is essential for building a diverse, adaptable, and resilient Defence. This inclusivity is not only crucial for operational effectiveness but also strategically important given the global diversity and the nature of current conflict regions. By embracing diversity, Defence is better equipped to understand and operate effectively in varied cultural contexts and enhancing strategic success.
• Theme 3 - Force modernisation and skill enhancement: Investing in advanced training and education programs is critical to upskilling Defence personnel, which maintains Defence's long-term capability. This approach reduces dependence on external entities such as industry contractors or allied states, thereby strengthening Defence's autonomy and operational readiness. By focusing on internal skill development, Defence can maintain a highly capable and self-sufficient force, better prepared to meet future challenges independently or jointly with NATO.
• Theme 4 - Personnel retention and resilience: The most effective retention strategy is to make personnel feel valued and that the organisation values them. Optimising workload management, providing clear career progression, and offering recognition through improved pay and conditions are essential. By fostering a supportive environment where personnel feel appreciated and rewarded, they will naturally become Defence's strongest advocates, enhancing both retention and recruitment.
• Theme 5 - Optimised resource allocation: Maximising the use of Reserve Forces' civilian expertise for targeted mobilisation enhances operational flexibility and efficiency while significantly reducing overall costs. By effectively integrating civilian skillsets, Defence can achieve greater versatility in operations, minimise reliance on external contractors, and ensure a more cost-effective approach to meeting strategic objectives.
• Theme 6 - Research and development: While significant strides have been made in researching personnel health and wellbeing, these advancements are largely due to individual efforts rather than a coordinated organisational approach. Investing in dedicated, systemic research will align the UK with international best practices, ensuring that Defence remains competitive on the global stage. By institutionalising this focus, Defence can build on existing successes and foster a more sustainable and impactful research environment. An approach to this would be to create an organisational research strategy with key priority areas.

Describe the strategic, threat, and operational context for UK Defence 2024- 2050. 500 words.

The strategic environment facing the UK in the coming decades will be marked by a rapidly evolving global landscape, where conventional and unconventional threats will challenge UK Defence capabilities. A failure to recruit and retain personnel poses a significant threat, undermining the UK's ability to operate effectively on the world stage. This deficiency will weaken our standing in international alliances, particularly NATO, and encourage our adversaries who might exploit perceived weaknesses.

While technology and a professional standing fighting force is critical, the importance of maintaining sufficient numbers cannot be overstated. Conflicts in recent years has demonstrated that strength in numbers remains vital. The UK’s current Defence structure is insufficiently agile to match current demands, necessitating a more robust recruitment strategy, a focus on health & wellbeing and an expanded use of the Reserve Forces.

Our support of NATO is essential, but the UK must play an active and credible role within the alliance. A diminished capacity to contribute due to recruitment and retention challenges would not only reduce our operational effectiveness but also our political influence within the alliance and on the global stage.

A major threat is the health and wellbeing of the force. Operational readiness is directly linked to the physical and mental health of personnel and should be treated as such. High injury rates, long-term sickness, and medical discharges not only reduce the effectiveness of Defence but also increase costs and undermine morale. Addressing these issues is vital to sustaining a capable and resilient Defence.

To address personnel shortfalls, the role of the Reserve Forces should be expanded. However, this must be matched with legislative changes to protect civilian employment for reservists, including enshrining the right to protected training in law. This would ensure that the Reserve Forces are better integrated into Defence strategy, providing a flexible and cost-effective means of enhancing the UK’s military capacity. Reserves have made the commitment to Serve, they should be allowed the opportunity.

Public attitudes towards Defence also present a strategic challenge. Negative perceptions of those who Serve, often characterised by the "mad, bad, and sad" narrative, can be exploited by external actors to weaken national resolve and support for Defence. This narrative must be countered to strengthen public support and ensure that UK’s Defence is seen as a valued and integral part of national security.

In conclusion, the UK's Defence strategy for 2024-2050 must prioritise personnel sustainability, inclusive personnel development, and modernisation to ensure operational readiness. A proactive approach to recruitment, health, research and public perception is essential to maintaining a capable and resilient Defence that can effectively navigate the complex threats of the future.

This response is focused on these aspects outlined, and not the wider threats faced. Other highly qualified and eminent responses should be considered in this regard.

Propose, in order of priority, the roles UK Defence must be capable of fulfilling 2024-2040. 

This response is focused exclusively on health and wellbeing, recognising it as a critical threat to Defence capability. The roles UK Defence must prioritise to address this challenge are (in order):

1. Defence must prioritise the physical and mental health of its personnel to ensure sustained operational readiness and resilience. Improved health outcomes are essential for the effectiveness and longevity of the force.
2. Defence should invest in advanced health technologies to support better monitoring, tracking, and management of personnel health. This includes the development and integration of systems that allow for proactive health interventions and efficient resource management. Defence should own these technology innovations. This includes digital interventions, use of secondary Defence data for research, and expanded academic collaborations.
3. Defence must minimise the dependency on external industry contractors by ensuring that systems are interoperable, and that data can be transferred seamlessly between providers where necessary. Where third-party services are required, the ability to transition between providers without disruption must be ensured. Where third party services are required, they must be required to train Defence personnel, and be compelled to sign the Armed Forces Covenant.
4. Defence should focus on building internal capabilities in digital by upscaling its personnel and strategically use Reserve Forces. This approach will reduce long-term external costs and enhance Defence’s self-reliance in providing critical services.
5. Defence should leverage (and capture) civilian expertise of Reserve Forces to bolster Defence’s health capabilities. This will create a robust and flexible support structure that can adapt to evolving challenges and ensure that Defence is well-prepared for future operational demands.

By prioritising these roles, UK Defence will be better equipped to maintain a capable and resilient standing fighting force, ensuring that health and wellbeing are integral to its strategic success from 2024 to 2040.

Propose in order of priority what defence capability (as a function, not defined as equipment or organisation) UK Defence requires to deliver the roles required of it 2024-2050.  This should distinguish between:  - The capability required for enduring, standing commitments and tasks, or to be held at high readiness.  - The capability required on mobilisation in times of crisis.

No comment.

Describe how existing UK Defence capability and the current 10-year programme match the demands of the roles and capabilities identified for 2024-2040, highlighting and prioritising critical gaps. Identify areas that require urgent attention to ameliorate immediate strategic level risks or vulnerabilities, including in logistic support and medical capability.

The current state of UK Defence capability and the 10-year programme reveals significant gaps that must be urgently addressed to meet the demands for 2024-2040. Defence has consistently failed to meet recruitment targets, creating a shortage of personnel that weakens operational readiness. Therefore, the highest priority should be improving recruitment and retention, particularly during Phase 1 and Phase 2 training where personnel are often discharged due to injury or leave due to the misleading promotional material suggesting Defence life is ‘easy’. Retaining long-term Service personnel, who are considering transitioning out of Defence, is equally crucial as their expertise is invaluable and difficult to replace.

Also, there are critical gaps in logistic and medical capabilities that require immediate attention. The current infrastructure is not robust to support future demands, particularly in high-tempo operational environments. To mitigate these vulnerabilities, Defence must implement strategic retention programmes, focus on reducing dropout rates in early training, and bolster medical and logistic support systems. Addressing these issues is essential to maintaining a capable and resilient Defence in the face of evolving global challenges.

Propose how UK Defence's contribution to NATO should be enhanced as a driving requirement of capability and activity, including through prioritising defence export potential.

No comment.

Describe how the current support to Ukraine is integrated into UK Defence programmes and activities and propose measures in priority order to sustain and enhance this support to at least 2027.

The current support to Ukraine has been integrated into UK Defence programmes, but this has come at the expense of training our own personnel. While supporting Ukraine is vital, it should not take precedence over the needs of UK Defence. The integration of this support should be balanced, ensuring that UK training systems are not compromised. To sustain and enhance support to Ukraine until at least 2027, it is crucial to allocate resources effectively, ensuring that both UK Defence needs and international commitments are met without undermining the training and preparedness of UK forces. 

Simply put, priority should be given to maintaining robust UK training programmes while continuing to provide necessary support to Ukraine, ensuring that both objectives are achieved in parallel.

Describe what advancing UK Defence from a multi-domain to an integrated force concept means and identify the key characteristics of each, including the relationship with non-kinetic means and resources.

No comment.

Describe and recommend how the UK Defence homebase is made secure and resilient in confrontation, crisis and conflict, including in enabling full mobilisation and campaign endurance when needed.

To ensure the UK Defence homebase is secure and resilient during confrontation, crisis, and conflict, the recruitment and retention of Reserve Forces must be a priority. Developing depth of skill within the Reserve Forces is key, as their current training opportunities are limited. For example, Combat Medical Technicians in the Reserve Forces are restricted to CMT 3 and CMT 2 levels, without access to CMT 1 training, which is available only to regular forces. This is the same across many different speciality trades. This limitation hampers progression and reduces overall capability. Expanding specialist training for Reserve Forces, while potentially increasing liability, will enhance their effectiveness and ensure that the UK Defence homebase is prepared for mobilisation and sustained campaign endurance when needed.

Propose how UK Defence rapidly becomes a global pacesetter in how Digital Age technology evolves equipment, organisation and method across all aspects of capability, identifying the first major steps to be taken in each domain, how quickly this can be achieved, and recommend what legacy capability currently in service and in the extant future plans should be modified or dropped.

To rapidly become a global pacesetter in the evolution of Digital Age technology, UK Defence must undergo a significant cultural shift, particularly in recognising and valuing digital skills, especially within the Army. Current Defence culture prioritises training only for specific trade-relevant skills, which limits the identification and use of digital talent. The first major step is to overhaul training programmes to include broader digital skill development and to actively identify and harness talent, such as through events like the Defence Hackathon. This does not mean giving each trainee an iPad, but developing in-depth digital skills such as software development, cyber security, systems management etc. 

Additionally, legacy systems that do not support this digital transformation should be critically reviewed and either modified or phased out. Only with these changes can Defence effectively integrate advanced digital capabilities across all domains. Further Defence should seek to embrace digital interventions for supporting, managing an intervening in health; leverage data for predictive analytics and seek to embrace high-value based healthcare.

Propose how UK Defence builds and sustains a relationship with technology providers and industry (major primes and SME) that enables the technology choices that will drive globally pace-setting modernisation and transformation.

To build and sustain a transformative relationship with technology providers and industry, UK Defence must focus on several key strategies. First, interoperability and transferability of software (including systems) should be prioritised in contracts to avoid vendor lock-in and ensure seamless integration across platforms. Defence should commit to fair market pricing, challenging unjustified price hikes, especially for software that hasn’t seen significant development. Leveraging its collective bargaining power, Defence can hold suppliers accountable, ensuring they deliver value and adhere to ethical standards.

Contracts should require suppliers to participate in the Armed Forces Covenant at the Bronze level (Defence Recognition Scheme), reinforcing a commitment to supporting Defence personnel. Additionally, Defence should collaborate with technology providers to establish clear training and development pathways for its personnel, ensuring they remain at the cutting edge of technological innovation. Finally, fostering long-term partnerships with both major primes and SMEs will allow Defence to integrate innovation from smaller, agile companies alongside the stability offered by larger firms. This comprehensive approach will enable UK Defence to drive globally pace-setting modernisation and transformation.

Propose how UK Defence acquisition and support (for all classes of equipment) continues to modernise to: maintain pace with key technology, enable the transition to an integrated force, advance interoperability in NATO, and deliver agile, cost-effective acquisition and support at optimum value for money.

No comment.

Propose measures in priority order that establish how UK Defence transitions to the optimum balance of regular, reserve, civil servant, and industry people across all standing tasks and commitments and on mobilisation. Recommend the recruitment, education, training and retention measures required to achieve this.

To transition to the optimal balance of regular, reserve, civil servant, and industry personnel across all standing tasks and mobilisation, UK Defence should prioritise the following measures:

1. Focus on targeted recruitment strategies for regular and Reserve Forces, addressing current shortfalls outline in this response. Enhance retention by offering career progression, improving work-life balance, and recognising experience.
2. Invest in comprehensive education and training, especially for Reserve Forces, enabling access to advanced and specialist courses to build depth of skill. Incorporate more flexible, digital-based training solutions for civil servants and military personnel.
3. Leverage the civilian expertise of Reserve Forces, ensuring they are integrated strategically into Defence operations. Update legislation to protect their civilian employment and provide guaranteed access to training.
4. Expand collaboration with the private sector to share skills and knowledge, with flexible arrangements to integrate industry personnel into Defence tasks when required.
5. Foster specialised roles for civil servants to support logistics, technology, and administration, reducing reliance on external contractors. Provide ongoing professional development opportunities to sustain a high level of operational support.

By implementing these recruitment, education, and retention measures, UK Defence can create a sustainable, flexible, and balanced workforce that meets its standing tasks and mobilises efficiently when necessary.

Propose the digital enablement required by an Integrated Force across all domains and to recommend and prioritise the steps to accelerate this transformation to place UK Defence as a global pacesetter in this area.

To digitally enable an Integrated Force focused on health and wellbeing, UK Defence should prioritise key measures. First, a centralised health data system will track personnel health and wellbeing, enabling efficient care management and early intervention. Wearable technologies and AI-supported analytics can enhance proactive health monitoring, improving the early detection of physical and mental health issues. Integrated digital health records will ensure continuity of care across postings. Expanding telemedicine and virtual healthcare will provide remote treatment options, particularly for deployed personnel. 

Digital tools for mental health support, such as resilience apps, should be developed to improve overall mental health care. Examples already exist, such as the DrinksRation, ACTVet and MeT4VeT apps developed by King’s College London and used across the Armed Forces Community. Additionally, robust data security and interoperability standards will ensure safe, seamless sharing of health information across all services.

By implementing these digital solutions, Defence will not only improve the health and wellbeing of its personnel but also significantly enhance operational effectiveness. Healthier, better-supported personnel are more resilient, reducing downtime due to illness or injury, and improving overall readiness. This ensures that Defence is fully capable of sustaining operations and responding swiftly to emerging threats, making it a global leader in digital healthcare for military personnel.

Propose how individual and collective education and training across all elements of Defence delivers the character and skills required by an Integrated Force for the challenges and requirements of 2024-2040, recommending the optimum division of responsibility between all elements of UK Defence and allies/partners for delivering this and identifying how the rapid and continuous application of new technology enables this.

To ensure UK Defence develops the culture and skills required for an Integrated Force between 2024 and 2040, there must be a focused effort on expanding educational opportunities, enabling career growth, skill development, and personal advancement. Collaboration with educational institutions, such as universities and colleges, should be significantly increased to offer diverse learning opportunities that foster advanced skills, critical thinking, and cross-disciplinary expertise.

One key area for improvement is the Educational and Training Services (ETS), which currently restricts eligibility to those with teaching qualifications at the primary or secondary education levels. This limits the ambition of Defence education, as highly skilled professionals, such as university lecturers with Doctrates, are unable to contribute. Broadening the eligibility for ETS to include individuals with higher academic qualifications would open up Defence to a wealth of expertise and elevate the standard of education offered. If Defence only relies on individuals with high school teaching experience, the ambition and depth of the educational content will remain limited to that level.

Further, offering Defence personnel opportunities for career changes through advanced educational programmes would enhance adaptability within the force. By facilitating skill development in new areas, personnel can transition more smoothly between different operational roles, which is vital for a modern, flexible Integrated Force.

Digital technologies should be incorporated into education and training, enabling continuous skill development through online learning platforms, simulations, and virtual environments. By using these modern educational tools, Defence can ensure that personnel are constantly upgrading their capabilities to match the rapidly evolving technological landscape.

Additionally, partnerships with private industry and educational institutions would further enhance the breadth of opportunities available, allowing Defence personnel to access specialised knowledge in fields such as cybersecurity, AI, and data analytics. These partnerships would provide new pathways for skill development, ensuring the force remains cutting-edge and globally competitive.

In conclusion, to achieve the level of educational ambition required for 2024-2040, UK Defence must broaden its approach to education by collaborating with higher education institutions, opening ETS eligibility, embracing technology, and focusing on skill development that aligns with future operational challenges. This approach will ensure the force remains adaptive, forward-thinking, and prepared to meet the strategic demands of the future.

Propose how the maritime (including littoral) domain is developed to contribute to an Integrated Force, recommending measures that accelerate modernisation and transformation and identifying existing capabilities and programmes that do not support this.

No comment.

Propose how the land domain is developed to contribute to an Integrated Force, recommending measures that accelerate modernisation and transformation and identifying existing capabilities and programmes that do not support this.

No comment.

Propose how the air and space domains are developed to contribute to an Integrated Force, recommending measures that accelerate modernisation and transformation and identifying existing capabilities and programmes that do not support this.

No comment.

Propose how UK Special Forces are developed as part of an Integrated Force and in support of UK National Security and resilience.

No comment.

Propose how Defence Medical Services ensures the levels of primary, secondary and tertiary medical support are in place for standing tasks and for mobilisation for the demands of operational context of 2024- 2040. Identify the optimum mix of Regular, Reserve, civil servant, NHS and industry capability required and recommend urgent and essential measures to ameliorate major extant risks in priority order.

To ensure the Defence Medical Services (DMS) provides adequate primary, secondary, and tertiary medical support for standing tasks and mobilisation from 2024 to 2040, several measures must be implemented. There is a critical need to address the issue of basic medical standards, such as routine dental inspections and health checks, which are not being met consistently. For Reserve Forces, allowing access to civilian healthcare providers for routine checks, with costs reimbursed, would help reduce strain on military medical services and ensure readiness for deployment.

Additionally, significant waiting lists for medical treatment pose a threat to operational effectiveness, limiting personnel's ability to deploy. This issue must be prioritised by streamlining services and increasing capacity within the DMS. Regular service auditing and quality assessments should be conducted to monitor the standards of medical care provided. Partnering with academic institutions to research existing health data would enable data-driven improvements to healthcare delivery.

Further exploitation of resources like the Academic Department of Military Mental Health and other military-funded research units is essential. Their research can help inform better mental health services and overall medical readiness within the armed forces. To meet the demands of 2024-2040, DMS should adopt an optimal mix of Regular, Reserve, civil servant, NHS, and industry capabilities. Priority must be given to addressing current medical capacity shortfalls, improving service delivery, and ensuring that the health and readiness of personnel are maintained at the highest standards.

In conclusion, DMS must focus on improving routine medical checks, reducing waiting lists, enhancing collaboration with academic institutions, and optimising the use of Reserve and civilian medical capacities to meet future operational demands.

Describe the current utilisation and condition of all aspects of Defence Estate and recommend how to accelerate restoring all Married and Single Living accommodation to standards recognised as acceptable by the Services. Make recommendations in priority order for estate rationalisation and disposal at best value to Defence and identify how to improve and modernise support to Defence estate.

The current condition of the Defence Estate, especially Married and Single Living accommodation, falls below acceptable standards. To restore these accommodations, UK Defence must prioritise meeting industry standards across the rental sector, not just military standards. By doing so, quality of life for personnel and their families will significantly improve, enhancing retention and morale.

To accelerate restoration, Defence should conduct a full audit of all housing to identify priority areas for refurbishment. Rationalisation and disposal of underutilised or non-strategic assets should be pursued to fund these improvements. Additionally, modernising the support infrastructure for the Defence Estate through partnerships with industry and more efficient property management systems will ensure long-term sustainability and better living conditions for personnel.

Propose how to improve financial management and administration across Defence, meeting HMG standards of accountability and identifying how more rapid adoption of digital technology will modernise culture, organisation and ways of working.

No comment.

Propose how UK Defence can build relationships with allies, partners and alliances as a strategic strength for the UK

No comment.

The Royal Navy, the Army and the Royal Air Force each invited to review their culture, organisation, training and institutional habits, describe how they reinforce modern war fighting imperatives, and recommend where changes are required as part of the modernisation of defence to be captured in this Review.

This has been covered in other elements of the response.

Digital health solutions are rapidly becoming a mainstay in research, transforming the way health data (e.g. wearables, questionnaires, environment) is collected, analysed, and used. These technologies, ranging from mobile health apps to digital therapeutics, offer huge potential to enhance the efficiency of research and improve health outcomes. However, researchers often face challenges of balancing the drive for cutting-edge research with the need for software that is usable, intuitive, and accessible to end-users (otherwise known as research participants).

When it comes to developing digital health software, researchers must juggle complex research-driven goals, such as ensuring scientific validity, rigour and data accuracy, with creating an interface that participants can easily navigate.

This task is further complicated by factors such as the limitations imposed by research funding, the expectations and varying approaches of principal investigators and funders, and the absence of a standardised approach to software development. This discussion will also draw upon the concept of a Research Viable Product (RVP), which I coined in a previous post.

Understanding research needs in digital health

Academic research has a distinct set of requirements that must be carefully integrated into digital health solutions. A great example would be the development of the DrinksRation app. At the fore are concerns about data integrity, participant privacy, and the overall scientific validity of the study. Rigorous data collection procedures must be in place to ensure that the information gathered is both reliable and relevant to the research objectives (keep in mind data minimisation here). Ethics also plays a crucial role, with strict protocols governing informed consent and participant confidentiality.

In digital health research, studies often focus on assessing the efficacy of an intervention, which requires precise tracking of participant outcomes (often through questionnaires). This focus on high-quality data often demands sophisticated software with features that support granular data collection, while also ensuring participant engagement.

For example, if an app is too cumbersome or difficult to use, data quality could suffer as participants may engage less frequently or inaccurately. These challenges highlight the importance of integrating and considering research-focused features into digital health platforms without losing sight of the need for usability. As discussed in my post on the Research Viable Product (RVP), achieving this balance is key to success in academic digital health projects.

The concept of Research Viable Product (RVP)

The concept of a Research Viable Product (RVP) is designed to meet the specific needs of research, offering a structured approach to software development that prioritises research goals. RVP takes inspiration from the Minimal Viable Product (MVP), a term commonly used in the tech industry to describe a version of a product that includes only its most essential features.

However, RVP differs significantly by focusing on the unique demands of research-driven projects, particularly in fields like digital health. Unlike the MVP, which aims for rapid market entry with minimal features, RVP ensures that all critical research elements, such as rigorous data collection, ethical compliance, and scientific validity, are embedded in the product from the very beginning.

One of the core differences between RVP and MVP is the emphasis on maintaining the integrity of research throughout the development process. MVPs may prioritise speed and functionality over comprehensive features, often sacrificing elements that are not immediately necessary for market testing. In contrast, an RVP must ensure that the software meets high academic standards, addressing requirements such as data robustness, participant privacy, and regulatory compliance. This is particularly important in digital health, where the software must support ethical and accurate data collection to evaluate the efficacy of interventions, as well as ensuring participant safety and privacy.

In reality, RVP provides a framework that bridges the gap between traditional product development and academic research priorities. It allows teams to develop software that not only functions well but is also suitable for academic validation and ethical scrutiny.

By integrating the essential elements of research from the start, an RVP approach helps academic teams avoid the need for major reworking later on, thus streamlining the path from concept to publication, and eventually, to real-world implementation in clinical settings. This balance between usability and research validity makes the RVP model particularly suited to the development of digital health tools.

Usability and participant engagement

In the world of digital health research, usability isn’t just an added bonus; it’s essential to deliver the effect. The success of a digital health tool often hinges on how intuitive and user-friendly it is for participants, who may have varying degrees of technological literacy. A well-designed, easy-to-use interface not only helps ensure that participants continue engaging with the software but also improves the quality of the data being collected.

Several best practices can enhance usability without compromising data accuracy. First, involving Public and Patient Involvement and Engagement (PPIE) from the start can offer valuable insights into how real users interact with the software. Agile development methodologies enable iterative improvements based on user feedback. Beyond functionality, attention must be paid to the naming of features and design choices, as these can significantly impact how users perceive and engage with the software. Additionally, user interactions should be assessed through suitable questionnaires and continuously monitored via logging and analytics to ensure data accuracy and user satisfaction.

Balancing both worlds: Practical strategies

Successfully navigating the divide between research needs and usability requires a careful, iterative approach. Academic research teams can greatly benefit from regular feedback from both researchers and participants. By incorporating this feedback into each phase of development, teams can refine both the research protocols embedded in the software and the overall user experience. Iterative development, grounded in real-world usage data, allows academic teams to stay agile, making necessary adjustments without derailing the research. This is critical to success of research software, and does not need to come at a cost.

Co-production and co-code production are key strategies for integrating the needs of both researchers and end-users. By involving a wide range of stakeholders, from developers to participants, throughout the process, teams can create software that is robust from a research perspective while also meeting the usability needs of participants. This collaborative approach ensures that academic digital health solutions can fulfil their dual mission: supporting rigorous research while remaining accessible and engaging to users.

Security and privacy considerations

Security and privacy are paramount in the development of digital health tools, especially within research. Adhering to privacy laws like GDPR in the UK/Europe or HIPAA in the United States is not optional; it is a legal requirement that, if neglected, could result in significant breaches and fines. The UK’s Information Commissioner’s Office (ICO) has reported multiple data breaches in the digital health space, underscoring the critical need for secure systems. This also includes proactively tackling cyber security. Also, it is also important to seek to develop Data Protection Impact Assessments on data that is being collected to ensure compliance.

Methods for ensuring robust data protection include implementing encryption, secure access controls, and regular security audits. At the same time, these security measures should not overly hinder the usability of the software. By integrating data protection measures seamlessly into the software’s architecture, academic teams can build systems that protect participant privacy without sacrificing functionality or user experience.

Conclusion

Balancing research needs with software usability is a challenging but essential task in the development of digital health tools within academia. As we have seen, the Research Viable Product framework provides a useful lens for achieving this balance, ensuring that academic software is not only scientifically rigorous but also user-friendly. By adopting an iterative development approach, involving stakeholders in the process, and adhering to strict security and privacy standards, researchers can create digital health tools that are both effective and engaging. 

These are just some thoughts and rambling from my 10 years of developing software for research.

I recently completed the 15-day Combat Infantry Course (CIC; Reserves; 2023) at Vimy Barracks in Catterick. Continuing my series on Army Reserve training, I thought it would be helpful to share some insights, hints, and tips for anyone considering/preparing for CIC. It's important to note that this course is continually evolving, so my experience may differ from yours.

Preparation

Prior to attending CIC, it is crucial to emphasise that fitness is important. Completing a full RFT (S) with your unit beforehand is strongly recommended to ensure you’re fully prepared. If you fail the RFT (S), you will be Returned to Unit the same day. It is the only element of the course with a hard pass/fail. It isn't difficult to pass, so long as you practise. My approach was to run 5km with a 10kg weight in the months leading up to it.

Course Structure

Day 0 – Friday: Arrive at ITC Catterick Vimy Barracks no earlier than 1800. Spend the evening preparing kit for the RFT (S) to save time and keep your mind occupied. Be sure to meet those in your section/room. They will be your rock throughout the course.

Day 1 – Saturday: The morning is spent completing the Weapons Handling Test. The afternoon and evening is focused on preparing and weighing kit for the RFT (S). The shop is closed on a Saturday.

Day 2 – Sunday: The RFT (S) is conducted in the morning on hilly terrain at the training area out back (Land of Nod). This includes a 4km tab (55 minute pace) and a 2km run (within 15 minutes). After this, we tabbed to the astroturf for the remaining RFT (S) serials. The afternoon involved briefings on harbour occupation, kit preparation for the Intro Ex, and a kit inspection. It’s a long day, so moral boosts throughout the day are a must (sweets!).

Day 3 – Monday: We deployed to the field around 0900, with a 3km tab to the harbour location (+snap ambush location). We prepared the harbour for occupation and had lessons on target indication and six section battle drills.

Day 4 – Tuesday: The day was filled with lessons, including pairs fire and manoeuvres. We also had a night exercise with HMNVS and night familiarisation.

Day 5 – Wednesday: The harbour location was attacked at reveille, leading to a bug out and a casualty drag to the Bayonet lanes. After Bayonet, we tabbed back to ITC Catterick and spent the remainder of the day cleaning weapons. Note: Clean your weapons quickly to maximise downtime.

Day 6 – Thursday: We spent the day on the ranges practising the ACMT. Due to the small course size, practice was completed by 1500, and we stood down at 1600. It’s advisable to perform a battle clean of your rifle when not on point.

Day 7 – Friday: This followed the previous day's pattern, but we completed the official ACMT. Everyone passed by 1500, and we stood down by 1700.

Day 8 – Saturday: The morning included lessons on fighting in forests and built-up areas, section and platoon attacks. We then prepared our kit for the final exercise and had the rest of the day to ourselves.

Day 9 – Sunday: Spent at the urban training complex, learning how to fight in urban areas and conduct rolling section attacks. The best part of the course.

Day 10 – Monday: Continued building upon the skills and drills for urban warfare.

Day 11 – Tuesday: Deployed on the final exercise, which included a 5km tab to the harbour location in patrol formation. This was a full tactical deployment reinforcing all prior learnings.

Day 12 – Wednesday: Continued the final exercise with a night reconnaissance mission and section attacks.

Day 13 – Thursday: Further continuation of the final exercise, including a night ambush and fighting in forests.

Day 14 – Friday: The final day of the exercise culminated in a full platoon attack over 2km of undulating terrain (with a 6km tab to the location), followed by a casualty drag. We then tabbed back to Catterick for weapons cleaning and block jobs.

Day 15 – Saturday: Completed block jobs and stood down at 1000. Course complete.

The course is physically demanding, so having a solid training programme is essential. Also, pack plenty of sweets to provide a much-needed morale boost.

This relates to 2023.

I have recently just published as an open-source pre-print my thoughts and views on the need for a new term in the development lifecycle called 'Research Viable Product'.  You can read the full article here (as a pre-print) or a copy is below.

In the research software engineering community, we frequently encounter industry jargon to describe the development phase. Terms such as the minimum viable product (MVP), minimum marketable product (MMP), proof of concept (PoC) and prototype are commonly used to describe different stages in the development and testing of new software and products [1], [2]. Each term signifies a specific phase in the product lifecycle. In each phase, developers, owners, stakeholders, and teams are designed to manage risk, use resources efficiently, execute the development process to a specific standard and create software that meets user needs and market demands. 

However, the nature of research-oriented software development sometimes stretches beyond the traditional scopes of MVPs, MMPs, PoCs, and Prototypes. This raises the following question: is there a more suitable term for describing software developed in a research context? Any term should accurately reflect the objectives and challenges unique to the field of research software engineering: the research viable product (RVP) is one possible term. This term has received little use in the research software engineering community to date but could be used to describe the type of software or product more accurately developed. It is important to note that, as part of this Short Position Statement, the term ‘product’ is used in the broadest sense, encompassing digital health apps, content management systems, and algorithms implemented on Cloud architecture, among other possible products [3].

Decoding the jargon: MVP, MMP, PoC and Prototype

Before delving into the concept of RVP, let us briefly revisit the definitions of generic software development terms:

• Proof of Concept (PoC): A realisation of a specific method or idea to demonstrate its feasibility. It also has the potential for further development, often prior to the creation of a full product. It often lacks significant functionality or all outcomes or is preplanned (e.g., mocked-up).
• Prototype: An early sample, model, or release of a product built to test a concept or process, widely used across industries to refine and validate the functionality of a design.
• Minimum viable product (MVP): A product with the minimum essential features to attract early adopters and validate a product idea early in the development cycle.
• Minimum Marketable Product/Final Product: A version of a product or service that includes the minimum set of features necessary to be released to the market and satisfy early adopters or customers.

 These stages are critical in the lifecycle of software development, each serving distinct purposes, such as feasibility (PoC), refining user needs (prototype), or initiating market interaction (MVP/MMP). However, when developing software for research purposes, these terms often fall short of capturing the nuances of creating a product for testing and use within a research setting.

Introducing the Research Viable Product (RVP) 

To bridge this gap, the term Research Viable Product (RVP) refers to projects at the crossroads of software innovation and research. An RVP is defined by its readiness and intended use in a research context, specifically designed to test hypotheses, collect data, and evaluate efficacy in research scenarios while complying with research ethics and applying academic rigour. Unlike an MVP/MMP, which prioritises early market feedback through minimal features, an RVP focuses on the core elements necessary for research integrity, including data governance, protection, and security standards.

For instance, software innovations such as DrinksRation [4], [5] and RADAR-Base [6] represent the RVP approach by prioritising research objectives over broader marketability or user interface design. Considering DrinksRation as a case study, this RVP was developed to investigate user patterns in alcohol consumption and its health impacts. Unlike a typical MVP, the DrinksRation was designed from the ground up to ensure robust data collection, scientific rigour for a randomized controlled trial, adherence to stringent privacy regulations, and facilitation of statistical analysis, all of which are key aspects of a research tool. This is not the case when we consider other existing terminology.

Why is the distinction important?

Distinguishing MVP, MMP, PoC, and Prototype from RVP is not purely semantic; it addresses the core challenges and objectives specific to research-centric software development. By focusing on hypothesis testing, data integrity, and adaptability to research settings, the RVP acknowledges the unique ecosystem within which research software is developed.

Moreover, by involving participant input from the earliest stages and adhering to the highest data governance standards, RVPs offer clarity in the development process in research contexts. This approach not only ensures the relevance and reliability of the research conducted but also enhances the potential impact of the findings.

Future steps in research software development

The introduction of the RVP signifies a shift in how we should approach software development within research settings. It is challenging to focus on the unique demands of research integrity, participant involvement, data security and professional assurance above traditional development milestones.

As the term RVP develops, fellow researchers, developers, and stakeholders are encouraged to engage in the dialog. How can we better adapt our development processes to meet the intricate demands of research? What challenges and opportunities do you foresee in adopting the RVP terminology?

 References

[1]       A. M. Davis, E. H. Bersoff, and E. R. Comer, “A strategy for comparing alternative software development life cycle models,” IEEE Transactions on Software Engineering, vol. 14, no. 10, pp. 1453–1461, Oct. 1988, doi: 10.1109/32.6190.

[2]       R. Jain and U. Suman, “A Systematic Literature Review on Global Software Development Life Cycle,” ACM SIGSOFT Software Engineering Notes, vol. 40, no. 2, pp. 1–14, Apr. 2015, doi: 10.1145/2735399.2735408.

[3]       I. Cosden, “What is a Research Software Engineer?,” Ian Cosden Blog. Accessed: May 06, 2024. [Online]. Available: https://cosden.github.io/what-is-an-rse

[4]       D. Leightley et al., “Evaluating the Efficacy of the Drinks:Ration Mobile App to Reduce Alcohol Consumption in a Help-Seeking Military Veteran Population: Randomized Controlled Trial,” JMIR Mhealth Uhealth, vol. 10, no. 6, p. e38991, Jun. 2022, doi: 10.2196/38991.

[5]       D. Leightley et al., “Drinks:Ration - The role of a smartphone application in reducing alcohol consumption in a veteran population seeking formal mental health support,” London, UK, 2022. [Online]. Available: https://kcmhr.org/pdf/2022_ration_report.pdf

[6]       Y. Ranjan et al., “RADAR-Base: Open Source Mobile Health Platform for Collecting, Monitoring, and Analyzing Data Using Sensors, Wearables, and Mobile Devices.,” JMIR Mhealth Uhealth, vol. 7, no. 8, p. e11734, 2019, doi: 10.2196/11734.

Citation:

Leightley, D. (2024, June 25). Navigating the software development landscape: Introducing the term Research Viable Product (RVP). Retrieved from osf.io/5n37j

The ICO recently released a call for evidence on the topic of generative AI, and its accuracy of training data and model output. You can read more here. The following is my position statement in response to this call, and the emerging position of the ICO. This is a collective submission with Colleagues from King's College London

Response to Call for Evidence 

We strongly support the focus on the accuracy of training data and the necessity for robust reporting and validation of model outputs. However, as highlighted in our discussions, there is a critical need for explicit provenance of the training data and model outputs to ensure transparency and accountability. Adherence to the ICO's fundamental data principles (e.g. data minimisation) should remain paramount, while fostering innovation. To achieve this balance, we propose a simplified regulatory framework that is easy to implement (e.g. a toolkit). Our response to the questions outlined in the Call for Evidence also focus on the importance of public education on the strengths and limitations of these models, ensuring all stakeholders are well-informed.

Consultation question: Do you agree with the analysis presented in this document?

Response: We agree somewhat with the ICO analysis statement. Broadly, the analysis highlights ethical, legal, and moral obligations that all stakeholders need to consider. At present, this isn’t necessarily happening. The ICO's focus on accuracy and statistical accuracy aligns with the wider academic community concern that high quality training datasets, which have a clear data provenance are required (but not available). But also, there is a clear urgent need for generative AI competencies across organisational structures [1], [2]. An important area not directly addressed by the consultation thus far is the need to train developers and stakeholders in understanding and interpreting generative AI outputs correctly. Not just developing the statistical framework, but also how to interpret the results (generative AI competencies). 

It is important to recognise the wider community concern that data in the public domain shouldn't automatically be repurposed for generative AI, particularly when it involves manipulating or aggregating data to create similar outputs. A strong emphasis should be placed on a clear consenting pathway and a straightforward explanation on data use are key for end-user comprehension. This is important when most generative AI approaches use end-user input to refine and retrain models ‘on the fly’ without manual review on the quality of these inputs. This includes organisations who use LLM (for example), trained on user inputs, often opaquely returning outputs of likely different models, with undocumented training data. This aspect needs to be acknowledged, understood, and considered in any future ICO position statement. 

We agree that it is important for developers to document, record decision and processes. The ongoing Post Office Inquiry has laid bare the disconnect between technical teams and senior management, and this needs to be considered in the context of generative AI accountability. The translatability of technical details to both the public and senior organisational stakeholders is vital to develop trust in generative AI. Furthermore, it stresses the central role of data privacy and the risks posed by insufficient understanding of data management and processes which feed into organisational risk appetite. It must be acknowledged that there is a tension between generative AI and the principle of data minimisation. This could be addressed with thorough consent practices and the careful selection of trusted, albeit potentially biased, data sources. 

Consultation question: How can organisations deploying generative AI models effectively communicate to their employees, customers, or other people the extent to which their outputs are statistically accurate? 

Response: Organisations (including developers, academic researchers, directors etc) should be required to articulate the extent to which their outputs are statistically accurate within a timely fashion. This includes, using straightforward language, such as through a transparency statement. Statements such as these should be available on websites for customers and translated in other languages. Where internal, it should be integrated into regular training sessions and staff briefings. This approach ensures that both internal and external stakeholders are consistently informed about how AI models function and their level of accuracy. It is also important to consider the models themselves, and provide information on the model version itself. It is also possible to establish routine testing of their AI models to monitor accuracy. The results of these tests should be communicated using clear, unambiguous language to those who 1) use the system, 2) those who have provided data for training (depending on content options). This practice will help in maintaining trust and transparency, providing stakeholders with a clear understanding of the model's performance over time. 

The ICO could promote the adoption of a standardised testing framework, supported by a Toolkit, to enhance consistency in reporting AI results. This simplifies understanding AI outputs and aids in industry benchmarking. Alternatively, the industry could be encouraged to develop agreements that include a voluntary framework with peer review and validation elements. Such collaborative efforts also foster the adoption of best practices and continual improvement of AI applications. While ensuring accuracy and transparency is crucial, it is equally important that regulatory measures do not stifle innovation. A light regulatory touch will encourage the ongoing development and ethical use of generative AI, while maintaining compliance.

Consultation question: What technical measures can organisations developing generative AI models use to classify or mark audio, image, or videos outputs as being AI-generated, rather than human-generated?

Response: Generative AI has the potential drive innovation, empower individuals, and increase productivity within the UK. However, it is increasingly difficult to differentiate between human-generated content, and synthetic content generated by AI. It is getting increasingly difficult to identify ‘deep fake’ content versus real content. A recent briefing by the European Parliament identified Watermarking as one viable solution for tagging AI content [3]. We believe this approach has merit, and its adoption by Google [4], Microsoft [5] and Meta [6] indicate strong industry support. This also enables the ability to track provenance of the data. 

The intent from these organisations is to develop systems and frameworks which 1) mark the generated content, 2) identify the watermarked signature within the content to identify and trace its source. Watermarking isn’t without issue, and this will be discussed later. Other technical measures have been proposed, such as using an open repository (with data provenance assured), or enabling organisations to query generative AI models to see if content has been generated by its algorithms, but these give rise to further challenges related to privacy and are not viable in a large scale.  

Consultation question: What are the benefits and limitations of these methods?

Response: Developing multiple frameworks for validation will hamper any large-scale rollout of a classification framework for content arising from generative AI. Therefore, we believe that the focus should be towards watermarking as the most promising technology. This already has a significant benefit as some of the largest players in the field are working on this technology implementation. There main limitations of watermarking are presented focused on lack of standardisation across the sector, accuracy of watermarking and the algorithmic potential for false-positives and modification where watermarks can be removed or altered. 

It is important to recognise that watermarking is only viable for image and video content generated by AI. It is very difficult, if not impossible, to accurately predict, trace and authenticate text written by generative AI. This limitation needs to be acknowledged in any arising guidelines, policy, regulation, or legalisation. 

Consultation question: What technical and organisational measures can organisations use to test the extent to which the output of a generative AI model is statistically accurate?

Response: To assess the statistical accuracy of generative AI model outputs, model owners can implement a combination of technical and internal controls. First, employing data validation techniques ensures the training data's quality and relevance. It is important that the source of the data is known, and consent for its processing of generative AI demonstrable. Regular testing and auditing should also be undertaken. There is no viable reason not to undertake these activities. As part of the testing process, the models should be tested against established benchmarks. These benchmarks would preferable be publicly available industry standardised datasets. Statistical methods like cross-validation can further validate the generalisability of the AI model's predictions. Transparent reporting of these processes enhances accountability, while collaborative benchmarking with industry standards promotes continuous improvement and consistency across similar AI applications. Models which are publicly available, the corresponding statistical outputs should also be provided.

Consultation question: What technical and organisational measures can organisations use to improve the statistical accuracy of generative AI models? 

To improve the statistical accuracy of generative AI models, model owners can adopt several technical and internal controls. Technically, enhancing data quality through rigorous pre-processing and validation ensures robust training datasets. Incorporating diverse and comprehensive datasets also helps reduce bias. Regularly updating and refining models based on new data and feedback loops can further boost accuracy. At an organisational level, establishing dedicated teams for continuous AI monitoring and implementing strict governance frameworks ensures the integrity and accuracy of AI outputs. Additionally, collaborating with external experts for audits and adherence to industry standards can provide further validation and improvement.

Consultation question: As a developer / supplier of generative AI models, how will the proposed regulatory approach affect your ability to offer services to the UK market?

As a developer or supplier of generative AI models, the proposed regulatory approach in the UK may have significant implications on your service offerings. This regulatory framework will likely require adherence to specific standards and practices, potentially increasing the overhead for compliance and operations. However, it also presents an opportunity to enhance the educational aspect of AI technologies. By integrating educational programmes and developing knowledge bases early in the development process, you can not only comply with regulatory expectations but also add value to your offerings, making your AI solutions more robust and trustworthy. This proactive approach in education and compliance can help maintain a competitive edge in the UK market.

References

 [1]           F. Fui-Hoon Nah, R. Zheng, J. Cai, K. Siau, and L. Chen, “Generative AI and ChatGPT: Applications, challenges, and AI-human collaboration,” Journal of Information Technology Case and Application Research, vol. 25, no. 3, pp. 277–304, Jul. 2023, doi: 10.1080/15228053.2023.2233814.

[2]           E. Brynjolfsson, D. Li, and L. Raymond, “Generative AI at Work,” Cambridge, MA, Apr. 2023. doi: 10.3386/w31161.

[3]           T. Madiega, “Generative AI and watermarking,” 2023. [Online]. Available: https://www.europarl.europa.eu/RegData/etudes/BRIE/2023/757583/EPRS_BRI(2023)757583_EN.pdf

[4]           S. Gowal and P. Kohli, “Identifying AI-generated images with SynthID - Google Deep Mind.” [Online]. Available: https://deepmind.google/discover/blog/identifying-ai-generated-images-with-synthid/

[5]           K. Wiggers, “Microsoft pledges to watermark AI-generated images and videos.” [Online]. Available: https://techcrunch.com/2023/05/23/microsoft-pledges-to-watermark-ai-generated-images-and-videos/

[6]           “Meta to introduce watermarking feature for some AI products.” [Online]. Available: https://www.reuters.com/technology/meta-introduce-watermarking-feature-some-ai-products-2023-12-06/