In 2018, unbeknownst to most employees, Spain’s social security system deployed an artificial intelligence algorithm to analyze the sick-leave claims among Spanish employees and flag unusual patterns as potential sick-leave fraud.
The algorithm itself could not pass a final judgment, but the flagged employees received extra scrutiny from a health inspector tasked with catching fraud. The AI system was flawed, giving many false positives and doing little to alleviate the strain on the social security system. The use of this algorithm wasn’t made public until an investigative piece in the Spanish newspaper El Confidencial brought it to the public’s attention this year.
In 2021, Spain passed a law to bring some transparency into the use of AI algorithms. The Spanish government announced the creation of an agency for the supervision of AI that will investigate dangers that may arise from algorithms, as well as monitor and minimize the significant risks that this technology may cause in terms of health and safety, said Gisella Rocío Alvarado Caycho, an attorney with Sagardoy Abogados, a member firm of Ius Laboris in Madrid. “The purpose was also to audit the AI platforms and algorithmic systems using an objective, impartial way,” she said.
AI Obligations in Spain
The decree, once implemented, will explain the ways employees must be alerted to the use of algorithms.
“The main objective is for the employee representatives to be aware of how algorithms or AI systems are used in part for making business decisions that have an impact on working conditions,” Caycho said.
The decree makes it clear that—similar to the process for the 2018 sick-leave algorithm—final decisions can’t be made by the algorithm and must be made by a person.
Benefits of Algorithms
When it comes to issues such as sick-leave fraud, the use of algorithms could potentially help determine which funds are being misspent from social security, which ultimately benefits all people who pay into Spain’s social security system.
AI also may help detect misuse and whether institutions are getting unfair access to allowances, said Juan José Hita Fernández, an attorney with Augusta Abogados in Barcelona and Madrid. “It’s relatively easy for the medical community to determine what is the average duration per illness,” he said. “At the end of the day, this is paid by public resources; it’s money that we all pay from our taxes.” So the use of algorithms could be beneficial for everyone, he said.
Disadvantages of Algorithms
As Spain weighs the advantages and disadvantages of using algorithms, it will be important to remember that an algorithm is only as good as the data and programming it is built upon, and AI cannot fully replace real human decision-making.
One of the flaws of the 2018 algorithm was the low-quality data it fed on. “The algorithms cannot be 100 percent used to make a decision, because in the end, algorithms are treated by feedback from, or by data provided by, an employer,” Caycho said. “So, we will have to be very careful, because maybe the information that will be provided to the algorithm to detect an analysis [could include] the personal bias from these individuals that will fit the data included in the algorithm. It would be highly advisable to have some kind of audit to these algorithms to avoid discrimination, because the power that they have is huge.”
Still, having some augmented system in place to determine things such as sick-leave fraud could benefit the social security system as a whole. “There is a relatively high level of fraud or false statements sometimes,” Fernández said. “In the whole system, something needs to be addressed in a different way. And that’s where artificial intelligence may help.”
Katie Nadworny is a freelance writer in Istanbul.