This article has been reviewed according to Science X's editorial process and policies. Editors have highlighted the following attributes while ensuring the content's credibility: To what extent can AI improve our financial decisions or steer us in the wrong direction? Researchers at the University of Bayreuth investigated this question in a new study involving 3,700 participants and fictional investment fund savings plans.
Their findings are presented in a CESifo working paper. Fabian Herweg, chair of international competition policy at the University of Bayreuth, the researchers presented participants with a simple yet realistic decision. Participants were asked to set up a savings plan based on a fictional stock market index and then choose between two fund providers.
Both funds tracked exactly the same index but differed in two separate fees. "This created an apparent trade-off in our study: Each provider was cheaper with respect to one of the two fees," Herweg says. "As a result, it was not immediately obvious which provider offered the better overall deal.
In reality, however, for every chosen contribution level, there was one provider that was clearly superior, meaning it was always the cheaper option overall." Participants either made this choice on their own or with support from one of three advisory options: an AI chatbot, a bank-affiliated AI chatbot or a human financial adviser. The researchers controlled these advisory options so they either provided the correct recommendation or a biased one. The online study was fully incentivized, meaning participants received higher payments when they made better investment decisions.
"Our findings show that AI can provide substantial support in decision-making, but only when its advice is correct. Equally, it can strongly steer people in the wrong direction if its recommendations are biased," says Joshua Greubel, a research associate in the University of Bayreuth's International Competition Policy research group. Without any advice, around 64% of participants selected the better provider.
When a neutrally presented AI chatbot recommended the better option, this figure rose to approximately 87%. However, when the same AI was instructed to recommend the inferior provider, the proportion of correct decisions fell to around 34%. The study also found that disclosing conflicts of interest offered little protection against poor decisions.
Extract — continue reading at the source.
