Non-experts could harness machine learning with new platform
Image credit: Ingram
A PhD student at Cornell University has led the development of a new platform which allows non-specialists to use machine learning efficiently and ethically. She hopes the platform could enable industries beyond computing to harness the opportunities of AI.
“We don’t know much about how non-experts in machine learning come to learn algorithmic tools,” said Swati Mishra, the PhD candidate who led the study. “The reason is that there’s a hype that’s developed that suggests machine learning is for the ordained.”
As machine learning has penetrated all sorts of industries populated by people without computing expertise, the need for effective tools to enable new users in leveraging AI is unprecedented, she argued. While existing research has mostly focused on understanding users and their challenges when navigating AI tools, this research approaches the challenge from the opposite direction: how can a system be designed such that users with limited algorithmic expertise but considerable domain expertise learn to integrate existing models into their work?
“When you do a task, you know what parts need manual fixing and what needs automation,” said Mishra. “If we design machine-learning tools correctly and give enough agency to people to use them, we can ensure their knowledge gets integrated into the machine-learning model.”
Mishra approached the problem by turning to a process called transfer learning as the starting point for introducing non-experts to machine learning. Transfer learning is a high-level approach to machine learning wherein users alter existing, pre-trained models to complete new tasks. This technique means that there is no need to build a model from scratch which requires colossal amounts of training data. Instead, a user can repurpose a model trained to, for instance, identify images of dogs, to a model that can identify images of cockatoos.
“By intentionally focusing on appropriating existing models into new tasks, Swati’s work helps novices not only use machine learning to solve complex tasks, but also take advantage of machine-learning experts’ continuing developments,” said Professor Jeff Rzeszotarski, senior author of the paper. “While our eventual goal is to help novices become advanced machine-learning users, providing some 'training wheels' through transfer learning can help novices immediately employ machine learning for their own tasks.”
Mishra’s research involved creating an interactive platform which exposes the workings involved in transfer learning, such that non-experts can understand how the model processes data and makes decisions. Through a lab study involving people with no background in machine learning, she was able to pinpoint when they lost their way, what their rationales were for making certain alterations to the model, and which approaches were most successful.
In the end, the researchers found participating non-experts were able to successfully use transfer learning and alter existing models for their own purposes. However, inaccurate perceptions about machine 'intelligence' – largely conflating human learning with machine learning – frequently slowed learning among non-experts.
“We’re used to a human-like learning style, and intuitively we tend to employ strategies that are familiar to us,” explained Mishra. “If the tools do not explicitly convey this difference, the machines may never really learn. We as researchers and designers have to mitigate user perceptions of what machine learning is. Any interactive tool must help us manage our expectations.”
Sign up to the E&T News e-mail to get great stories like this delivered to your inbox every day.