Panellists:
- Alison Heppenstall, Professor of Geocomputation, University of Glasgow
- Raffaele Esposito, City Design Manager, Glasgow City Council
- Caroline Parkinson, Creative Industries Lead, Edinburgh Futures Institute
- Tom Wilkinson, Chief Data Officer, Scottish Government
- Jeremy Doherty, Edinburgh Office Leader, Arup
Moderator:
- Stephen Jardine, Scottish Journalist, Broadcaster
Part 1: Understanding AI
Introduction
In October 2025, the AI for Collective Intelligence (AI4CI) Hub, Architecture and Design Scotland (A&DS) and Edinburgh Futures Institute (EFI) hosted ‘AI in the City’, a public forum inviting the local community to explore a technology that is reshaping how we live and is reimagining the places we call home. Over 90 minutes, Scottish broadcaster and journalist Stephen Jardine guided a panel of experts through questions about what AI is, how it’s transforming Scotland’s built environment and how we can navigate concerns around data ethics, energy consumption and governance to harness AI’s potential responsibly.
What is AI?
To kick things off Alison Heppenstall, Professor of Geocomputation, University of Glasgow, explained that AI at its most basic level is a machine doing something that only humans could do before, such as understanding language, recognising images, processing and manipulating data or complex calculations.
In 1950, Alan Turing, English mathematician and Computer Scientist, proposed the Turing test to measure a machine’s capability to exhibit human-like intelligence through conversation. Essentially, if you interact with a machine via text and you can’t determine if the responses are coming from a machine or a human, then that machine can be considered to demonstrate intelligent behaviour.
Alison went on to explain that although AI has been around since the 1950s, the technology and data weren’t advanced enough to deliver on its initial hype. We went into long periods of time known as “AI winters” where the promise was there but the ability to deliver it wasn’t. So, what’s changed? The combination of increased computational power, vast amounts of data and more sophisticated algorithms sped up the advancement of emerging AI technologies which are now being put to work in Scotland’s infrastructure and urban spaces.
Part 2: AI in Action – Current Applications
Infrastructure and Maintenance: The Queensferry Crossing
Jeremy Doherty, Edinburgh Office Leader for Arup, used the Queensferry Crossing as an example of how Arup have been using AI for the construction and maintenance of major infrastructure projects.
The bridge was built over a period of 5-years and was completed in 2017. During this project they used rudimentary AI to optimise construction phasing, a critical consideration for a multi-billion-pound public project subject to intense scrutiny. AI models helped assess how to mobilise teams efficiently and ensure the bridge was constructed on time and within budget.
At the point of construction Arup made the decision to pepper the entire bridge with Internet of Things (IoT) sensors to collect data relating to climatic conditions, load barriers etc. This forward-looking action recognised the significance of data and has since allowed them to learn and deliver measurable improvements. The IoT sensors enable predictive maintenance, meaning engineers can assess how the bridge is behaving based on the weight of the load on top of it and the climatic and environmental conditions. With the use of AI, they can assess if there are specific components that are taking more load and wearing down at a faster rate than expected.
“Prevention is better and cheaper than cure, so we are able to predict when there may be a requirement for us to intervene and provide that maintenance upfront and lower the cost of asset maintenance and expand the lifecycle of that asset.” – Jeremy Doherty, Arup
One exciting opportunity that Arup identified from the data being collated involved information from the MET office. Using this data they were able to predict weather conditions and assess where and when ice would be present, causing significant risk to life from traffic collisions. The data available means that they can now predict those conditions, and allow the human in the loop to decide how best to intervene.
The information that has been collected on the bridge can also be used more broadly to influence the planning and construction of future assets, optimising public use and minimising costly errors.
Jeremy explained that, ethically, there is a need for human intervention in using AI in major infrastructure projects. Where there is potential danger to life, the risks are too high to exclude humans. We are not yet at the point, in engineering or construction, to allow a machine to make a decision on our behalf.
There is also a legal obligation. There is professional indemnity where the owners, operators and maintenance engineers have a responsibility to ensure that the asset is doing the job that it was designed to do. If it fails, there are consequences, so for now we need a human to be able to make the final decision. There must not be any ambiguity about who is ultimately responsible.
The Queensferry crossing is a great example of AI’s potential in large-scale infrastructure, but the technology is also being deployed at street level to reshape the design of our cities.
Urban Planning and Traffic Management: Glasgow’s Approach
Raffaele Esposito, City Design Manager for Glasgow City Council, shared some of the ways Glasgow is integrating AI into its future planning. Viva City is a pilot project that has partnered with Glasgow City Council to improve traffic flow and road safety for pedestrians and cyclists. AI sensors have been installed at key intersections to improve cycle and vehicle detection, enabling traffic signals to respond more effectively. In addition to collecting multimodal data to feed the traffic signals, these sensors also gather accurate, detailed and anonymous data on other travel modes. This provides a clearer understanding of how the intersections are used to ensure that the new junction layouts meet the needs of all road users.
It is Raffaele’s belief that AI has the potential to optimise and improve every aspect of city life starting with the relationship between decision–makers and end-users. AI technologies could be used to facilitate citizen engagement in the design process, making it more visible and therefore more accountable. If citizens are encouraged to participate more, it will naturally improve the relationship between people and places.
Glasgow City Council is also using digital twin technology as a stepping stone to AI to try and better understand conditions that were previously outside of its control. City officials and planners are looking at examples of the wrong building appearing in the wrong place or being designed in the wrong way. This careful analysis work will improve city planners’ ability to make informed decisions about these types of challenges.
“We are at the very beginning of a journey, and we are starting to see those first shoots, but they are nothing compared to what AI could bring”. – Raffaele Esposito, Glasgow City Council
While Glasgow is laying the groundwork for AI-driven urban planning, other Scottish institutions are exploring how emerging technologies can transform the relationship between planners and the communities they serve.
Community Engagement Through Creative Technologies
Caroline Parkinson, Creative Industries Lead, Edinburgh Futures Institute, added that digital twin technology can be used alongside augmented reality (AR) tools for planning and construction projects.
Caroline explained that these types of digital tools are expanding beyond the creative industry, where they were first created, into architectural planning and construction spaces. Architects are utilising 3D fly throughs and modelling tools that animators would traditionally use for gaming and entertainment projects, to look at planning scenarios. These technologies allow city planners to create virtual walkthroughs of proposed buildings, allowing architects, clients and citizens to examine the design and functionality in a realistic and immersive setting before construction begins. It is important to recognise that AR and digital twin technologies are not AI but rather separate, complementary technologies. The combination of a digital twin, AR and AI allow for a more immersive experience. AI can analyse data and improve the interactivity of AR applications.
In terms of urban planning, Caroline suggested that there are greater opportunities to codesign with the community. The use of 3D fly through can aid people in picturing how a project could look when completed. Caroline is working with a project called Convergent Screen Technologies and performance in Realtime (CoSTAR) which has created two virtual production hubs in Dundee and Leith. The hubs specialise in virtual production – a new cinematography technique which uses computer-generated imagery (CGI), augmented reality and motion capture to create virtual film and performance sets.
If city officials, architects and planners were to use this type of virtual production space, they could create a more immersive realisation of a design for consultation with the community. These technologies would not only allow the community to see and better understand a proposed plan but also enable planners to make real-time changes. For example, they could add or remove buildings, park benches, zebra crossings, green spaces or cycle lanes based on the feedback they receive.
“Not everyone can visualise the idea that they have in their own head or describe it, but maybe through using some new technology in mobile phone apps, they can feedback to planners and communicate in discussions and consultations about what they visualise. – Caroline Parkinson, Edinburgh Futures Institute
These immersive technologies offer exciting possibilities for public engagement, but their potential impact extends beyond consultation to the planning system itself. AI and creative technologies could be used to improve the planning process, but it requires disrupting the established system, which is already struggling with significant delays, capacity shortages and lack of strategic oversight.
Streamlining the Planning Process
Raffaele highlighted that although architects and designers shape our city landscapes, the process of developing built spaces is complicated and multifaceted. If AI can be integrated into the planning system, it could allow architects and designers to focus less on the mundane operations and more on creating meaningful experiences for the public.
Jeremy agreed that the current planning process is tricky and laborious, adding additional pressures to the UK’s housing crisis. As part of the UK Innovation fund, Arup is working with a client to assess the applicability of two types of AI. They are using Generative AI to support and expedite the generation of application data to a high standard, therefore making them more likely to get a positive outcome, and they are developing Agentic AI models to check against the criteria. These practical AI applications are happening now and it’s worth recognising that this wouldn’t have been possible until very recently. If this pilot project is successful, it could be deployed across all local authorities, helping to resolve one of the major challenges in the UK housing crisis.
As these examples show, AI holds remarkable promise for transforming Scotland’s built environment. But realising this potential requires confronting serious questions about how we develop and deploy these technologies responsibly.
Part 3: Addressing the Challenges
Data Ethics, Bias and unequitable distribution
‘AI doesn’t happen without some sort of precursors like data and, obviously I’d say this as Chief Data Officer, but a critical barrier to making the most of AI in a lot of situations is having good and well understood data underneath – Tom Wilkinson, Scottish Government
Many opportunities for AI, whether in the built environment or elsewhere, require enabling work on data before AI can be successful in delivering real value. The importance of having access to good data has been repeatedly highlighted by our panellists, but it is information about us, it is our data. What happens if our personal information ends up in the wrong hands? That’s the question concerning many people.
Alison highlighted that the University of Glasgow, alongside other academic institutions, have strict ethical guidelines to mitigate against the misuse of data. Within her research Alison uses mobility data to try and understand how people experience city spaces. This data is captured from a range of sensors like environmental sensors that are designed to pick-up air pollution and foot traffic and mobile phone sensors that track GPS, Wi-Fi triangulation, and accelerometer apps, which can track where you are and how you got there. These sensors are tracking our individual behaviours but when combined with others, reveal larger patterns.
When you pool these different data sets it offers a broader view of what’s happening in that specific area. Researchers can use this data to train their models and make projections; ‘what if’ scenarios on what the future might look like. Alison explained that she can develop models to explore scenarios like closing the M8 motorway, examining who would be impacted and what alternative uses the space could serve. These models allow researchers to generate scenarios and take them directly to the local community for feedback and citizen engagement.
“For my area new forms of data, new insights, new models and algorithms are fantastic but this additional layer of public engagement and getting that feedback is important. We are not interested in trying to identify people. From my perspective we need data to make these models work, to develop methods and tools that can help decision-makers to make informed decisions”. – Alison Heppenstall, University of Glasgow
A present concern with those developing AI tools is not only getting access to data but also having bias within the data. It is a huge concern for researchers like Alison that the data they have access to is not inclusive and therefore the decision-making is also biased.
There are many factors that can contribute to data bias within AI systems and technologies. A simple example would be if the data you’ve collected has come exclusively from smartphone technology. You might assume that this device covers a broad range of characteristics like gender, race, disability, sexual orientation etc however it doesn’t properly consider the digital divide. Individuals from lower-income households or rural areas may not have reliable access to these types of technologies and therefore their perspectives and experiences are excluded from the data. There are also digital literacy considerations, with younger or more educated individuals having an advantage over those from an older generation or with less education. Researchers and those working within AI technologies need to be vigilant about bias and take rigorous steps to minimise its affects or they could potentially exacerbate inequality.
Outside the ethical and bias concerns for AI there is also worry that investment in the technology will pool within major cities like Edinburgh and Glasgow to the detriment of smaller cities, towns and rural communities. It may seem like good economic sense to invest in areas of higher populations as these are often the areas that have big issues like congestion, air pollution, waste, construction challenges, healthcare pressures etc that AI could tackle. However, this strategy creates inequity in other areas which could have consequences not only for those living in less populated areas, but also those living within cities. One example would be that if there isn’t adequate investment in towns and rural communities there won’t be an incentive for younger populations to stay which threatens the sustainability of these communities and puts more pressure on city resources.
Beyond questions of privacy, bias and unequitable distribution, AI’s rapid growth raises another pressing concern: its environmental footprint.
Environmental Impact and Sustainability
There are high hopes that AI can help tackle some of the world’s biggest environmental emergencies but there are also concerns about the significant resources and energy it uses.
When it comes to the environment, there is a negative side to the sudden explosion of AI technologies and its associated infrastructure, according to a growing body of research. The proliferating data centres that house AI servers produce electronic waste. They are large consumers of water, which is becoming scarce in many places. They rely on critical minerals and rare elements, which are often mined unsustainably. And they use massive amounts of electricity, spurring the emission of planet-warming greenhouse gases.
In 2021, the Scottish Government launched its Green Datacentres and Digital Connectivity Vision and Action Plan, which works alongside UK initiatives to ensure that the energy demands of data centres, including those supporting AI, are carefully considered.
AI systems, particularly large foundation models, are highly computationally intensive and require significant energy and specialised hardware. This creates challenges for aligning AI development with renewable energy sources, which can be intermittent. For example, pausing high-cost computing infrastructure until wind or solar power is available is not economically viable.
Looking ahead, there is optimism that improvements in battery technology and energy storage will help mitigate these challenges. There is also speculation that current LLM technology is stabilising in terms of its base performance and that the greatest hope for reducing the energy implications of AI will come from some of the design shifts that are happening with the technology.
What that might mean is shifting strategies to focus more on training better versions of those foundation models and putting more creative energy into the design of more efficient implementations.
In early 2025, Chinese startup DeepSeek released DeepSeek R1, claiming performance comparable to technology developed by ChatGPT-maker OpenAI, but at a fraction of the cost and energy consumption. DeepSeek’s approach uses much cheaper computer chips and much lower amounts of energy to match the performance of other foundation models by using techniques such as model distillation and a mixture-of-experts architecture. This design activates specialised components only when relevant, reducing computational overhead. If we continue to see improvements in energy efficiency, it could make the use of green energy a more viable option.
“I’m optimistic that those types of changes will start to see the current tensions between AI and energy reduce over the coming years”. – Tom Wilkinson, Scottish Government
Balancing AI’s environmental costs with its benefits requires not just technical innovation but clear governance frameworks.
Regulation and Governance
The European Union introduced the EU AI Act in 2024, creating the first comprehensive legal framework for AI across member states. The UK is not subject to this framework and the UK Government, which holds responsibility for AI regulation as a reserved matter, has not introduced a dedicated AI law.
The EU’s approach to AI governance is grounded in a strong precautionary and ethics-driven philosophy, while the UK’s pro-innovation approach is more light touch when it comes to regulation. Tom explained that the Scottish Government has always tried to bridge the gap between the UK and Europe for historical reasons and so has built much of its thinking around trying to satisfy both regulatory landscapes, despite their differences.
In 2021 the Scottish Government developed its AI strategy, taking an approach to AI that was founded on being trustworthy, ethical and inclusive. Under this strategy Scottish government have launched the Scottish AI register, a world first, which is a comprehensive public resource for understanding the various AI systems in use or development within the Scottish public sector. There is also a Living with AI course which has been designed to get the public engaged, interested and informed about what AI is and its impact. The course combines expert insights with interactive learning to improve AI literacy and help individuals navigate some of what is a very complex and often confusing space. The Scottish AI playbook helps small businesses to engage with AI and make ethical use of it, to help them succeed.
The current strategy period is coming to an end and the Scottish government is in the process of redeveloping its AI approach. Tom is confident that those same principles of fairness will be at the core of what Scotland continues to do around AI.
In Conclusion
AI remains a topic firmly in the forefront of citizens’ minds. There is optimism around this technologies potential to improve big societal issues that reaches far beyond tonight’s topic of AI in the City, but there are also concerns about the potential risk of widening the social divided through inequitable investment, unchecked bias within the data and the lack of firm regulation.
There is a clear appetite for more open discussions and public engagement about AI and how it should be used. It is important to note that this article is a brief overview of the evening’s discussions, pulling out the major talking points. We would welcome all those interested to listen to the full and unedited discussion here.
Acknowledgements
AI for Collective Intelligence would like to thank our event partners Architecture and Design Scotland and Edinburgh Futures Institute (EFI) for all their support in hosting this public forum and Marion Thain, Director of the EFI for her warm welcome.
We would like to thank our incredible panellists Alison Heppenstall, Caroline Parkinson, Jeremy Doherty, Raffaele Esposito and Tom Wilkinson for dedicating their time and expertise to answer some tough questions. Our thanks to Stephen Jardine for his masterful moderating of this discussion and finally, we would like to thank all those who braved the Edinburgh rain to join us, who added to the discussion and engaged with this topic.
