A new study by researchers from the Commonwealth Scientific and Industrial Research Organisation’s (CSIRO) Data61, in partnership with the Australian National University and researchers from Germany, has revealed that artificial intelligence (AI) can influence human decision-making.
Spearheaded by CSIRO scientist Amir Dezfouli, the study [PDF] involved running three experiments where participants played games against a computer.
The first two tests involved participants clicking on red or blue coloured boxes to win a fake currency. In the third experiment, participants were given two options as to how they could invest some fake currency. In the scenario, participants played the role of the investor while the AI played the role of the trustee.
As all three games went on, the AI learned the participants’ choice patterns that eventually saw it guide the players towards specific choices. For instance, by the third game, the AI was learning how to get participants to give it more money.
Dezfouli said the study highlighted that AI could influence human decision-making by exploiting the vulnerabilities of an individual’s habits and patterns.
“Although the research was theoretical, it advances our understanding of how people make choices. This knowledge is valuable because it allows us to mitigate our vulnerabilities so we can better detect and avoid flawed choice as a result of potential misuse of AI,” he said.
He added that the way future AI operates will be dependent on its creators.
“Ensuring AI and machine learning are used as a force for good — to improve outcomes for society — will ultimately come down to how responsibly we set them up in the first place,” Dezfouli said.
The study has been published in the Proceedings of the National Academy of Sciences journal.
At the end of last year, Data61 researchers developed an implantable artificial intelligence monitoring and seizure detection helmet system designed to prevent seizure disorders for patients who have undergone decompressive brain surgery.
The detection system was developed and trained using traumatic brain injury data from Monash University to monitor brain activity for seizures while in standby mode before it is reactivated when a seizure is detected.
“Monitoring brain activity post-surgery is especially critical to a patient’s recovery as seizures can regularly occur, often leading to patients developing epilepsy,” CSIRO’s Data61 researcher Dr Umut Guvenc said at the time.
“These seizures are often difficult to detect, with current monitoring techniques only able to be used in a hospital using bulky devices for less than 24 hours, providing a brief snapshot of brain activity during that time only.
“This new method can continuously monitor brain activity wirelessly, allowing the patient to be mobile, comfortable, and more socially active.”
