Heavy metal, My Little Pony, and digital skills for government
Kent Aitken
Janelle Shane, a research scientist, recently experimented with using AI to sort through a list of names and figure out whether each was a My Little Pony or a heavy metal band. She found that, with minimal supervision, her AI improved, learned, and got to a 94% accuracy rate. It correctly predicted that Stormgarden and Death House were heavy metal bands - but also guessed heavy metal for the names Sparkle Cheer and Flutterbuns and got those ones wrong. The AI model had figured out that it could hit that accuracy level by always guessing “heavy metal band.” In fact, that simple heuristic was more effective than other more convoluted approaches the AI tried.
What Shane was doing was demonstrating the importance of solid training data. In this case, the corpus of data the AI was looking at was 94% heavy metal bands. So the AI took what it knew and applied that knowledge to every new problem.
This experiment serves as a cautionary tale for the use of AI in decision-making but it’s equally as valuable as a way to think about why public servants need to learn new skills and problem-solving approaches.
The Canada School of Public Service launched the Digital Academy in October 2018 as a beta. We’re often asked about the end goal. There are tempting answers like “build digital skills and mindsets” or “improve the digital and data literacy of the public sector” but those still fall short. Ultimately, our success lies in the success of our learners, in their many and multifaceted jobs across government.
“Digital thinking” won’t suddenly make everything better. But we have to make an intentional shift towards digital approaches precisely because of how human problem solving works. If all we know is heavy metal bands, that’s what we’ll try to apply to new problems. It’s the same if our experience is weighted towards analog-era approaches.
The Digital Academy is not about “more AI” or “more agile” as an end goal. It’s about expanding public servants’ toolkits so they can apply the right solution to the right problem at the right time. It’s not that 100% of products should be agile, just that it’s definitely not the case that 100% should be waterfall.
In some cases, we might find success stories where the program or service success can be traced back to the digital learning that preceded it. More often, however, the success will be in the absence of pitfalls, in smooth product launches, in conversations not had. It’ll be when you propose a solution that people had never heard of a few years ago and everyone already gets it: from leadership, to corporate enablers, to oversight and auditors. When you make a budget request for discovery and user research and no one blinks an eye.
That’s what an equipped public service means to us at the Digital Academy.
For more information, visit our research and development platform at busrides.ca, read our About page, join our newsletter, or email us. We’re currently helping public servants expand their data, AI, design, DevOps, and leadership toolkits and we’ll be growing to scale throughout 2019.
Page details
- Date modified: