Darwin launches interactive guide to jumpstart AI use

What’s the best way for local governments to innovate? Learn from each other.

This idea is central to the work we do at Govlaunch. By building off another city or county’s ideas, local governments can be more efficient, more effective, and drive innovation without reinventing the wheel. (This is why we’ve built a wiki of over 5,000 local government innovation projects to provide inspiration for agencies looking to enact change.)

So when we saw the work that Darwin’s doing, we got very excited. Not only is the Australian city pioneering new Smart initiatives like AI, they’re creating a set of best practices for others to learn from. And they got the idea by building off work done by another government.

Let’s take a closer look at $ Darwin’s AI Quick Start guide$ , and how other cities and counties can learn and build upon this collaborative piece of work.



Darwin goes digital

In recent years, Darwin has fully embraced modernisation, rolling out $ Australia’s largest Smart Cities project$  in order to drive more economic development and promote the city as a great place to live and work.

This effort was made up of many phases: $ Switching On Darwin$  introduced a series of technological advancements, such as a larger, more robust Wi-Fi network, sensor-driven climate and parking tech, and energy-saving LED lighting around the city.

These new devices and services are all monitored in a $ #SmartDarwin City Intelligence Platform$ , which is publicly available to allow for transparent reporting of the city’s functions. Anyone can log online and check how many available parking spaces there are at a given time, how much carbon dioxide emissions the city’s lights are producing, or what the average noise level is in the city core.

As the city grows more connected and primed for the adoption of new technology, it also needs to establish some ground rules. This is where a third component of the Smart City work comes into play: the $ launch of Darwin’s AI Quick Start Guide$ .

Why focus on AI?

$ AI, or artificial intelligence$ , is a growing asset in local governments’ toolkits. It drives greater operational efficiency and helps governments better serve their citizens.

Its application is broad — we’ve seen AI used for everything from $ fighting chronic homelessness$  to $ streamlining building safety inspections$  to $ reducing traffic congestion$ . But with so many ways to use the technology, many organisations don’t know where to start.

What’s more, AI is still new, so there aren’t universally held principles or practices to ensure it’s used responsibly. Because some uses of the technology involve personal or private data, it’s an especially sensitive tool to work with.

So, the opportunity is huge — but it needs to be done right. Darwin set out to make sure its businesses and organisations could take advantage of AI, while doing so responsibly.

Looking to another local government for inspiration

Darwin wanted to provide guidance to other groups planning on using AI. To do so, they looked for guidance themselves.

Through the $ Australia-Singapore Digital Economy Agreement$ , Darwin connected with Singapore’s Info-communications Media Development Authority (IMDA) and Personal Data Protection Commission (PDPC) groups. Singapore is a leader in innovation, and these two organizations have paved the way with best practices and $ case studies$  for the use of AI. To build its own guidelines, Darwin first looked at what Singapore had created.

Two of resources were especially helpful:

1. $ Singapore’s Model AI Framework$ : This framework is designed to create a set of transparent rules of the road to ensure businesses use AI responsibly, and to build public trust and confidence in the technology. $ It includes guidance such as$ :

• Decisions made by AI should be explainable, transparent & fair
• AI systems should be human-centric

2. $ The Implementation and Self-Assessment Guide for Organisations$ : If the Model AI Framework is the “what” of AI best practices, the ISAGO is the “how.” This companion resource outlines ways organizations can examine their AI readiness and opportunities, and provides case studies on how AI can be implemented.

Beyond its partnership with Singapore, Darwin also turned within Australia’s borders for assistance. The city leveraged experts from the Australian Computer Society (ACS) to consult and $ curate AI resources$  for the Quick Start Guide.

How the Darwin AI Quick Start Guide works

The ultimate goal of Darwin’s AI Quick Start Guide is to help organisations, from other local governments to businesses and other local contributors, utilize AI in their processes responsibly.

To do so, the city had to:

1. Consult existing experts, resources and materials

2. Synthesize the information into an easy to comprehend description of a complex opportunity

3. Create a resource that was easy to use, while still detailed enough to provide clear guidance

The result is a digital survey tool that’s hosted on the City of Darwin’s website. The survey contains more than 30 questions across two main categories.

$ Part one$  provides an organisational assessment:

“This section explores some of the project-related attributes of your AI implementation. This includes the characteristics of your organisation, policies and governance, stakeholder management and the role of AI in the project.”

Questions in this section focus on basic organisational structure, such as business type and number of employees, as well as goals for the use of AI, and the existence of policies that would affect the use of AI.

“Does your organisation have an existing risk management system that can be expanded to include AI-related risks?”

“Did your organisation conduct an impact assessment on individuals and organisations who are affected by the AI solution?”

“Did your organisation provide channels for feedback or queries about the AI solution?”

From there, the second part of the survey digs into the more technical aspects of AI. This includes validating the available dataset, identifying potential biases in the data, and more. Questions cover topics such as:

“Select the approximate age of the dataset(s) used”

“Did your organisation assess the degree to which the AI solution generalized well and failed gracefully?”

“Did your organisation document the process used to capture the relevant information such as datasets and processes that yield the AI model's decisions?”

Users have the option to fill out the full survey, or split Part One and Part Two between different subject matter experts. The entire thing takes less than ten minutes to complete; it’s mostly Yes or No or multiple choice questions.

AI Quick Start Guide Takeaways

Individual takeaways are provided instantly after a user takes the survey. Participants get an immediate assessment of their AI readiness, in the form of a percentage and a corresponding categorization:

• Aspiring: 0 to 30 percent
• Emerging: 31 to 60%
• Confident: 61 to 85%
• Leader: 86% and above

You can see where you’re stronger: Organizational assessment or the dataset and model assessment. You can also drill into each question and answer; if the answer you provided was not in line with Darwin’s AI best practices, the survey will provide you with guidance to be more compliant.

For example if your answer is “no/unsure” for this question, “Is the dataset used to produce the AI model fully representative of the actual data or environment the AI model may receive or function in?”

The survey will provide a recommendation:

To mitigate selection bias, consider:

-- Benchmarking data distributions against population statistics to identify and quantify how representative the data is

-- Whether it is useful to adopt a random assignment approach for the sample data

-- Whether it is useful to use quality metrics (e.g. completeness, freshness and context) to evaluate whether the dataset used for the AI model is fit for purpose and matches the population it is intended to represent

-- Whether it is necessary to use a heterogeneous dataset (i.e. data collected from different demographic groups or from a variety of reliable sources)

-- Whether it is necessary to use training data across different communities, events and attributes

At the end of the survey results, users can link off to more AI resources.

The tool is a simple, but powerful framework that can guide others toward responsible use of AI. Having these principles in place will make it easier for Darwin and its residents to continue to use new technology in innovative ways.

But moreover, Darwin’s creation of a streamlined, value-adding tool should be an inspiration beyond AI. At its core, the city saw a need for clearer guidelines in an area it hoped to grow, and it pulled in the right partners and information to make it happen.

Local governments should do as Darwin has done: embrace the work of another organisation, and customise it to create a resource that will benefit the local community. By doing so, they just might become an inspiration for yet another city to innovate.