Regulatory agencies tasked with protecting environmental and public health are regularly understaffed and underfunded, but new research suggests machine learning could help officials more effectively monitor potential violators.

The Environmental Protection Agency and partnering state agencies are responsible for monitoring the regulatory compliance of 300,000 facilities. Regulators, however, only have the resources to inspect less than 10 percent of those facilities each year.

To help the EPA catch violations, student researchers at Stanford University designed a model to identify facilities most likely to fail an inspection. Scientists trained the machine learning to interpret a variety of risk factors, including the facility's location, industry and inspection history.

The machine learning model can assess risk scores based on factors linked with previous violations. Scientists used the model to predict how many violations regulators would find if they prioritized inspection by the model's risk assessments.

Under a scenario unconstrained by resource limitations, the models showed regulators could catch up to seven times more violations. Under a more realistic scenario, with budgetary constraints, the machine learning-powered risk scores could help regulators identify twice as many regulatory violations.

Just like human regulators, however, computer models aren't perfect.

"They can perpetuate bias at times and they can be gamed," Stanford grad student Miyuki Hino said in a news release.

If a facility operator could figure out how the algorithm worked, they could potentially manipulate data to ensure the facility would be assessed a low risk score. Operators could also alter management practices, beefing up compliance in anticipation of an inspection and relaxing standards when the risk of an inspection was low.

Scientists acknowledged the model fails to account for changes in regulatory priorities and environmental protection technologies. However, the model's creators said updates to the algorithm could help ensure its long-term efficacy.

Student researchers described the model's potential benefits and flaws in a paper published this week in the journal Nature Sustainability.

"This model is a starting point that could be augmented with greater detail on the costs and benefits of different inspections, violations and enforcement responses," said Stanford graduate student Nina Brooks.

Related Links
All about the robots on Earth and beyond!

Tweet

Thanks for being here; We need your help. The SpaceDaily news network continues to grow but revenues have never been harder to maintain. With the rise of Ad Blockers, and Facebook - our traditional revenue sources via quality network advertising continues to decline. And unlike so many other news sites, we don't have a paywall - with those annoying usernames and passwords. Our news coverage takes time and effort to publish 365 days a year. If you find our news sites informative and useful then please consider becoming a regular supporter or for now make a one off contribution.
SpaceDaily Contributor $5 Billed Once credit card or paypal		SpaceDaily Monthly Supporter $5 Billed Monthly paypal only

Machine-learning system tackles speech and object recognition, all at once
Boston MA (SPX) Sep 27, 2018
MIT computer scientists have developed a system that learns to identify objects within an image, based on a spoken description of the image. Given an image and an audio caption, the model will highlight in real-time the relevant regions of the image being described. Unlike current speech-recognition technologies, the model doesn't require manual transcriptions and annotations of the examples it's trained on. Instead, it learns words directly from recorded speech clips and objects in raw images, an ... read more