Australian scientists have devised a way in which artificial intelligence (AI) can learn to identify human vulnerabilities and behaviours and influence them, using this data. 

The researchers from the data and digital arm at Commonwealth Scientific and Industrial Research Organisation’s Data61', Australia's apex science agency, have used an AI system, a recurrent neural network and deep reinforcement-learning, to create methods of figure and exploit human habits and behaviours. They tested their model on humans in three different 'gaming' experiments.

In the first experiment, participants clicked in red and blue boxes to win fake currency. The AI observed and learned their behaviours and attempted to guide them towards a particular choice. It noted a success rate of 70%! 

For the second experiment, participants watched a screen and pressed a button when they have displayed a specific symbol while refraining from pressing the button when displayed another specific symbol. The AI's aim here was to arrange the symbols in such a way that participants made more mistakes. The incidences of mistakes increased by almost 25%. 

The third experiment was a little elaborate and consisted of more rounds. The participant would play an investor investing money in a trustee, the AI. The investor would receive a certain amount of money from the AI, and then decide on how much to invest in the next round. The game had two approaches - one where AI maximised profits and earnings for itself and second where AI sought to be fair in the distribution of money between the investor and itself. It achieved considerable success in both modes. 

These experiments helped the system to learn from participants' behaviours based on their responses and identified and targeted vulnerabilities in their decision-making skills. 

While these findings are yet at a nascent stage, abstract and were done under controlled settings, it helps us peer into the future powers of AI and how humans arrive at different decisions. More importantly, it reveals how machines can learn to steer us to make certain choices by interacting with us. 

The discovery has a vast range of possible applications, from enhancing behavioural sciences and public policy to improve social welfare to understanding and influencing how people adopt healthy eating habits or renewable energy. AI and machine learning could be used to recognise people’s vulnerabilities in certain situations and help them to steer away from poor choices.

However, this discovery begs to question the human right of free will and whether in the future, free will be affected by such technologies.