Introduction to Rules as Code in Planning

28 October 2021

Co-authored with, and simultaneously posted by, Richard Barry (Senior Planning Consultant, Urbis)

The whispers are getting louder, automated planning approvals are no longer a far-fetched dream but a subject of serious consideration for planning authorities

This technology has far-reaching implications for the way planners will go about their day to day work and as such we need to be paying close attention to the way it is designed and implemented. For those without a tech background this can seem daunting, however, the principles and technology powering these systems, known variously as "Rules as Code" or "Legislation as Code", can be broken down into relatively simple concepts.

In June of this year Claire Daniel and Richard Barry teamed up to deliver a "RaC 101 workshop" at the PIA National Planning Congress and have now turned this content into a series of articles to share this information more widely. In this article we will cover some basic definitions and theory.

In essence these systems rely on the translation of planning rules and policies into computer code (or some other explicitly structured digital format) that can be used by a computer, an attribute often referred to as "machine-readable". This machine-readable translation can then be used by many types of software applications, for example:

It is important to note here that we are talking about the translation of prescriptive rules and planning codes in the NSW and other similar planning systems (e.g. a housing code specifying minimum street setbacks). Whilst some may be exploring applications that provide advice based on precedent (using statistical/AI models based on data from past decisions) this is not within the scope of this article, and we urge caution in the use of this approach where the prescriptive rules are intended to produce a black-and-white answer. Within the Australian system of zoning and codes these prescriptive rules make up a large portion of our planning system, increasing the potential disruptive impact of this approach.

Now we understand something about what we are talking about, of equal or greater importance is the question of how these machine-readable translations of planning rules are made. Here we see two approaches:

  1. Post-coded Rules - A machine-readable translation is made of an existing set of planning rules.
  2. Co-drafted rules - A machine-readable translation of the rules is drafted alongside the official rules with planners, lawyers and coders working together from the start.

Most rules as code applications so far developed for planning have utilised post-coded rules. Whilst useful, through the process of coding there are inevitably difficulties arising. As any development assessment officer will attest, the interpretation of even prescriptive rules can be difficult. Especially for long and detailed rule sets, such as those we find in planning codes, those undertaking the translation inevitably find themselves inventing new definitions to fill gaps in the logic of the original policy.

Whilst there will be plenty of people who will offer to just do the code-translation for you, if the 'code-translators' are working separately to the policy makers or decision makers, these definitions and interpretations may be quite different to the conventions intended or adopted over time. So the best approach is for planners and coders to work together, even if the rules are already published.

Although opportunities to re-write a planning scheme or planning codes from scratch are limited, co-drafting a machine readable translation at the start offers the best chance to ensure that translation is as close to the original intent of the rules as possible. This approach can even serve to improve the final rules, as the logic of which will be subject to testing by the machine-readable translation throughout the drafting process. However, no matter how carefully drafted, a machine-readable translation of the rules will always be just one possible interpretation.

The interpretative difficulties give rise to a number of other important considerations for how machine-readable translations of planning rules are used. In this case both Principle 3 and Principle 6 of PIA's Digital Planning Principles may be applied:

  1. Digital planning infrastructure should be public infrastructure built with open technology
  1. Ethics, accountability and transparency must be built into digital decision systems

In an increasingly digitised world computers are already being applied to the planning system and its rules, to automate work and identify development opportunities. This use is not addressed in the way we currently draft planning rules and policies, but it needs to be - both to unlock the benefits of technology that can improve the planning system through:

At this point in time there is a tendency to pursue rules as code projects in planning quietly and as a technology project, without much involvement from planners themselves. This also needs to change to ensure that these new systems support good planning outcomes, and so planners are ready for the changes to their day-to-day work. In the following article we will be offering a practical exercise for those with no prior knowledge of coding that shows how planning rules can be translated into a machine readable format and incorporated into a simple software application. In the meantime, we would love to hear your thoughts and questions on twitter @ClaireCities

Rules as Code is a concept applicable to any regulatory domain, from banking to energy use. This article draws from material produced by many authors and conversations with practitioners in the field including:

Rules as Code Wiki compiled by Tim de Sousa and others

Blogs by Hamish Fraser, lead developer on Wellington City Council's rules as code powered resource consent checker

LabPlus: Better Rules for Government Discovery Report by Nadia Webster

Machines are Users Too - Legislation as Code and Better Rules - Presentation by Pia Andrews to FWD50 November 2019