“Does this outfit look good on me? Should I buy it?”, “Should I order pizza or hamburger?”, “What do I say to my mom?” or even “should I end my relationship?”
Have you ever found yourself in doubt about something and asked an artificial intelligence (AI) chatbot for life advice? That it is increasingly present at work is nothing new, but research shows that technology has been used more (and more convincingly) to make decisions about our personal lives.
A 2024 study published in the scientific journal "Scientific Reports" and led by Shinnosuke Ikeda of Kanazawa University in Japan investigated the extent to which ChatGPT advice is capable of altering human decisions.
In two online experiments with 1,925 Japanese participants, researchers compared the effect of advice given to the chatbot with recommendations given to human experts.
The results showed that, in certain situations, AI advice was able to influence decisions in a similar way to that given to experts. The experiments ranged from moral dilemmas to economic choices. The research also showed that people with more difficulty and insecurity in making decisions were more susceptible to the recommendations.
Customized responses
In an interview with g1, Cleber Zanchettin, professor of Computer Engineering at the Federal University of Pernambuco (UFPE) and specialist in AI, explains that the more decisions are delegated to AI, the more the tool knows who is asking and the more susceptible to the chatbot's "advice" that person becomes.
"The more the user delegates these decisions, the more data the system will have about you. The more data the system has about you, the more convincing the personalization it will be able to build. And the more convincing this personalization, the greater its ability to shape preferences and behavior", he describes.
And within this cycle, the user may end up trapped in this "outsourcing" of decisions.
"As you don't make a decision, in the future you will wait for someone to make it for you. So, you become more dependent on technology", he says.
Shopping recommendation
When AI is used to choose products to buy, the risks increase. In addition to possible dependence, Zanchettin explains that the information may be biased.
This is because the models behind chatbots "learn" about products based on content about them on the internet, which includes reviews from other people who have purchased and even information from suppliers.
The AI does not test the product, but rather bases the recommendation on the content that exists about it on the internet.
How to find a balance
So is it better to stop asking AI for help? Not necessarily.
Zanchettin argues that the tools can be useful precisely for organizing large amounts of information and presenting alternatives. The difference is in using AI as support, and not handing over the final decision to it.
In practice, instead of simply asking the AI to choose which cell phone to buy, for example, the user can request a comparison between models, pros and cons and different alternatives — and decide based on this information.
"Use it to give you options and consider these options. And, this way, you make a faster and more assisted decision, and it doesn't necessarily have to be made for you."
