AI could be the tool to get our inefficient planning system on track
by Garry Miley , https://www.thejournal.ie/author/garry-miley-/ · TheJournal.ieGarry Miley Architect and lecturer
The Housing Essay is a weekly deep dive from a rotating variety of voices into issues impacting the property crisis in Ireland. Are there potential solutions that may be currently overlooked or traditionally ignored by policymakers?
IN THE DEBATE around artificial intelligence, we hear a lot about the need to keep a human in the loop. The technology is moving so quickly, we are told, that someone needs to have a hand on the tiller.
I share that instinct, but only up to a point. It’s true that there’s a lot about artificial intelligence we don’t fully understand, so there are limits to how far it can be trusted. But there are also plenty of exaggerated claims about its dangers: well-trained, open-source models can be quite reliable when used carefully.
On the other hand, we know from bitter experience that human involvement isn’t always as benign as the human-in-the-loopers would have us believe. In fact, some systems would probably benefit from having computers in the loop to keep an eye on the humans.
Let’s look at the problem through the lens of our planning system.
Some cities use strict mathematical rules to plan their urban areas. These tend to be
straightforward and easy to understand, but the resulting arrangements can feel geometrically dull – your average North American city, for example.
The Irish system isn’t very mathematical. Here, the Government writes the legislation and councils shape development plans. Decisions on individual applications are left largely to planning officials, whose judgements ultimately determine what gets built. The Irish approach allows for more discretion and, theoretically, more appealing outcomes.
But here’s the problem. Its success depends on planners exercising near-perfect judgement. Few are that talented. I’ve heard it argued that others aren’t particularly committed to impartiality in the first place. The result is a process that lacks consistency and, consequently, public confidence.
A chance to restore confidence
AI-assisted assessments might be well placed to restore some of that confidence. By reducing the idiosyncrasies inherent in individual judgement, we could ease the difficulties around requests for further information or hard-to-interpret planning conditions. And because computers follow clearly defined procedures, we could reduce the need for judicial review.
Advertisement
Computers are also immune to the media controversies and political spats that can send an application off in curious directions. They’re unlikely to be influenced by powerful voices.
With the technology available today, AI models could check applications for completeness, assess proposals against national policies and guidelines, establish whether the requirements of local development plans have been met, and even examine architects’ drawings for compliance with building regulations.
Much of this work could be done quickly and accurately, without relying on untested technology.
The tools we’re talking about here are relatively old (in AI terms), well tested and easy to control.
Here’s how it might work.
First, an AI model would be installed and operated locally, allowing planning applications and other sensitive information to be kept on servers in Ireland. It would then be given access to relevant development plans, legislation and guidelines, enabling it to consult the applicable rules when assessing a proposal. We could also teach it the difference between good and bad outcomes by training it on previous decisions.
Now, to get the most out of this, we’d need to do a bit of preparation, and this would be the hard part. Computers are better at interpreting precise instructions than vague aspirations, so we would have to replace much of the imprecise language we currently use in our documents with clearer language about what is and isn’t permitted.
To get a sense of how this might work in practice, consider the recent controversy surrounding the redevelopment of St Stephen’s Green Shopping Centre. Dublin’s development plan might be revised to specify a maximum height of seven storeys for buildings on the Green, rather than relying on the current requirement that “building heights respond to their surroundings”.
Considerable time and much negative energy could be saved if we established in advance what “appropriate” actually means, with AI ensuring that the agreed definition is applied consistently.
Related Reads
Dereliction destroys our built heritage - will we ever learn?
The cast of actors in the soap opera of Irish planning needs a review
Could a long-shelved housing report be the solution that never happened?
AI wouldn’t make the final decision on an application like St Stephen’s Green, but it could establish whether a building is on the Record of Protected Structures, check compliance with height and density requirements, and examine public submissions to distinguish relevant planning arguments from attempts to generate controversy.
The St Stephen’s Green proposal also includes an architecturally challenging new entrance which people on all sides of the debate agree isn’t quite what St Stephen’s Green needs.
Deciding how to address this issue is where the human could prove useful. Until AI develops the sensitivity required to respond to all the considerations involved in creating a pleasing public entrance, humans will have to do that work instead.
The key point, however, is that the potential for human drama would be limited to one area. If the council’s decision were appealed to An Coimisiún Pleanála, the arguments could be confined to the design of the entrance, rather than reopening matters already settled by the AI quickly assessing some clear planning rules. This would narrow the scope of disagreement, speed up the process and keep the tone of the debate professional.
A major change to the administration of the planning system would, of course, cost a lot of money. Without the benefit of statistics, I imagine those costs would be comfortably offset by the considerable costs saved on redundant revisions and expensive judicial reviews.
Nowhere outside Ireland seems to spend quite so much time being scandalised by planning proposals, or to be quite so baffled by how its own system works. We’ve become so accustomed to it that we forget we’re actually entitled to a process that is timely, understandable, intelligent, first-world and consistent. If AI can help deliver that, then let’s invite it to the meeting.
By all means keep a human in the loop, but let AI help us design a smoother loop.
Garry Miley is an architect and lecturer who has worked both in private practice and with Limerick City Council in major housing and urban design initiatives. He teaches architectural studio, history and theory at South East Technical University (SETU).