Connect with us

Hi, what are you looking for?

Tech & Science

Algorithms 22 percent more accurate at predicting long-term benefit claimants

The artificial intelligence itself was able to process big data and it was found to be 22 percent more accurate compared with human experts.

AI: More than Human exhibition invites you to explore our relationship with artificial intelligence. - Tim Sandle
AI: More than Human exhibition invites you to explore our relationship with artificial intelligence. - Tim Sandle

A new study finds that artificial intelligence is more accurate with predicting if individuals within a given population are more or less likely to become long-term benefit claimants.

The technology has been developed by Dr. Dario Sansone (University of Exeter Business School, U.K.). It is hoped by Sansone  and his team that the technology could save governments around the world billions in terms of being able to pinpoint areas and to make earlier interventions in order to prevent long-term economic disadvantage and social exclusion.

The artificial intelligence itself was able to process big data and it was found to be 22 percent more accurate compared with human experts in assessing the more vulnerable members of society.

To test out the model, the researchers used the entire population of people enrolled in the Australian social security system across the period 2014 and 2018. Key variables included demographic and socio-economic data. Different measures of claims included unemployment, disability, having children, or being a carer.

The algorithm was developed by using a 1 percent sample of the five million people registered in the system. The sample was taken in 2014 and then used across the following three year period.

The data was next compared to the approach currently deployed for predicting welfare dependency. The current method is sociological, drawing on sex, age and education, income support history, migration status, marital status, and state of residence.

The greater accuracy for the AI program has been attributed to the growing computational power of machine learning algorithms. Such algorithms are capable of processing a far greater range of predictive factors (around 1,800 in total).

A further factor is, according to the researchers, is due to the algorithm being free from conscious and unconscious biases, with the human factor removed. However, care needs to be taken as there is evidence of the biases of the designers being incorporated into the operation of AI systems.

The researchers hope that the new approach will begin complementing existing models designed to identify prospective claimants early (those considered to be at the greatest risk) and to use these data to long-term welfare receipt.

Commenting on the application, Sansone summarizes the research: “ We found that the size and richness of the dataset on social security enrolees makes it ideal for a machine learning application, allowing the algorithms to achieve high performance by detecting subtle patterns in the data and by identifying new powerful predictors.” The new technology could one day allow caseworkers to focus their attention and time providing a personalised service to those in need.

Avatar photo
Written By

Dr. Tim Sandle is Digital Journal's Editor-at-Large for science news. Tim specializes in science, technology, environmental, business, and health journalism. He is additionally a practising microbiologist; and an author. He is also interested in history, politics and current affairs.

You may also like:

World

The decline in US life expectancy is well documented, perhaps to the point of being too well documented and very predictable.

Life

Many young people who have been incarcerated later struggle to achieve the basic milestones in adulthood, such as living on their own or maintaining...

Business

Hybrid work has fundamentally changed how IT contributes to sustainability initiatives.

Tech & Science

There are fewer than 50,000 Asian elephants in the wild, according to the World Wildlife Fund.