AI will manipulate people to make the right decisions

Artificial intelligence (AI) tools can be used to manipulate online audiences to make decisions ranging from what to buy to who to vote for, according to a study by a team at the University of Cambridge on the new market of “digital intent signals” known as as the “economy of intentions.”

Image source: Christin Hume

According to researchers at the Leverhulme Center for the Future of Intelligence (LCFI) at the University of Cambridge, the “intention economy” is a successor to the “attention economy,” in which social networks keep users on their platforms and show them ads. As part of the “attention economy,” advertisers can buy access to user attention in the present through real-time bidding on advertising exchanges, or purchase it for future promotions, for example, by renting advertising space for a month in advance.

«For decades, attention has been the currency of the Internet, says Dr. Jonnie Penn of LCFI. “The exchange of attention with social media platforms such as Facebook✴ and Instagram✴ has led to the development of the online economy.”

The study claims that large language models (LLMs), used to power AI tools such as the ChatGPT chatbot, will be used to “anticipate and manage” users based on “intentional, behavioral and psychological data.”

In the “intention economy,” AI companies will sell information about user motivations, ranging from hotel stay plans to opinions on a political candidate, to the highest bidder.

«In the economy of intentions, LLM can cost-effectively exploit a user’s communication rhythm, political views, vocabulary, age, gender, preferences, and even propensity for flattery. This data, combined with intermediary rates, will maximize the likelihood of achieving a given goal (for example, selling a movie ticket),” the study says. It is also stated that in such a world, AI models will guide the discussion to the benefit of advertisers, businesses and other third parties.

The study claims that advertisers will be able to use generative AI tools to create personalized online advertisements. One example is an AI model called Cicero from Meta✴, which has achieved “human-level” ability to play the board game **Diplomacy**, where success depends on predicting your opponent’s intentions.

AI models will be able to adjust their output in response to “streams of user-generated input data,” the study indicates. They will be able to isolate personal information from everyday interactions and even “steer” the conversation in a way that elicits more personal information. The study cites a report from the Cicero research group that “an [AI] agent can learn to nudge its interlocutor to achieve a certain goal.”

The researchers also predict a scenario in which Meta✴ will auction user intentions, for example, to book a restaurant, flight or hotel. While there is already an industry dedicated to predicting and bidding based on human behavior, AI models will transform this process into a “high-quality, dynamic and personalized format,” the LCFI scientists emphasized.