Source: University of Auckland (UoA)
Meanwhile, people who gave their political affiliation as Democrat, were more likely to trust AI-generated forecasts than Republicans.
Despite the growing sophistication of artificial intelligence, investors prefer human expertise when it comes to stock market predictions, according to a new study.
The study, which involved 3,600 US participants, examined responses to S&P 500 stock predictions made by human analysts, AI systems and a combination of both.
Researchers Dr Gertjan Verdickt (University of Auckland) and Francesco Stradi (KU Leuven), say the findings challenge the assumption that AI’s data-crunching prowess might automatically earn investor trust.
“We found that investors are more likely to believe human analysts first, followed by a combination of both human and AI,” says Verdickt, a finance lecturer at the University of Auckland Business School.
“AI-generated predictions are viewed with the most scepticism.”
He says this result was somewhat surprising in light of recent developments in AI technology.
“Previous studies have shown that AI can outperform human analysts, but it’s apparent that trust is a major issue.”
The results also showed notable differences between genders, with women showing more openness to AI-driven advice than men.
“Men tend to be overconfident in their financial abilities, which may explain why they are more sceptical of AI,” says Verdickt.
“Also, we have seen in other studies that women, on average, get different and often worse advice from financial advisers, such as recommendations for products with higher fees and less risk.”
The findings also show that investors with a deeper understanding of AI are more likely to trust its predictions.
Meanwhile, people who gave their political affiliation as Democrat, were more likely to trust AI-generated forecasts than Republicans.
To explore whether using more familiar AI tools could boost trust, the researchers also tested whether investors would view the well-known large language model ChatGPT more favourably.
“Contrary to recent research suggesting familiarity enhances trust in technology, our results indicate that replacing ‘AI’ with ‘ChatGPT’ does not improve investor trust. In fact, we find that investors distrust ChatGPT-generated advice, perhaps even more than the generic ‘AI model’ we reference in our study.”
Verdickt says the findings show that technical effectiveness alone can’t gain investor trust.
“We are the first to study investors’ reactions to AI forecasts from a perspective of credibility and beliefs. Our findings show that financial institutions should approach AI integration cautiously and consider tailored communication strategies for different demographics.”
