������

Systems using artificial intelligence, including conversational chatbots, have become more like humans than ever imagined, with their communication skills now better than most people’s.

In published in the Proceedings of the National Academy of Sciences, researchers at the ������ Business School and the University of Washington’s show how these systems become irresistible, letting users forget they’re interacting with machines.

“The general public is not prepared for what’s coming,” said lead author Associate Professor Sandra Peter from the ������.

“We always expected AI to be highly rational but lack humanity. Instead, AI developers built the opposite.”

The authors say a wide range of studies taken together show the large language models AI systems are built upon now outpace most humans in writing empathetically and persuasively. Popular chatbots including ChatGPT, Gemini, Llama and Claude all excel at role-play and consistently pass the , fooling people into thinking they are talking to a real human.

Professor Kai Riemer from the ������ said it is no longer a case of people ascribing human traits to machines – they are now anthropomorphic by nature, often indistinguishable in communication because they mimic humanness convincingly.

“This is a significant development,” Professor Riemer said. “For the first time, machines are anthropomorphic and convincing in human-like communication, adjusting to match conversational style and tone. And that makes resisting them increasingly difficult.”

The authors coined the term ‘anthropomorphic seduction’ to describe the allure of human-like qualities exhibited when AI chatbots interact with humans, despite the absence of any true human traits.

“We must not forget these machines do not possess empathy or human understanding,” Professor Riemer said. “They only appear to.”

Although chatbots provide opportunities for user interfaces that make complex information widely accessible, the authors warn these conversational agents open the door to manipulation at scale.

Recent high profile examples of the risks include concerns Meta’s new AI app is , and a that claims AI bots were up to six times more persuasive than human users at changing people’s point of view.

“They allow for persuasion without moral inhibitions,” said Professor , co-founder of the Center for an Informed Public.

Providers of popular AI systems like OpenAI are also making their creations more engaging, giving them personalities. While that will make them more seductive, it could make them far more “sticky,” making it easy for users to spend increasing amounts of time using them and potentially giving up more data about themselves.

“While AI companion apps might alleviate feelings of loneliness, systems that lean on anthropomorphic seduction have also been criticised for exploiting it,” said Professor West.The authors argue the speed of technological advancement requires urgent awareness and regulation.

In the article, the authors suggest policymakers consider implications across three areas: risk level, transparency and mitigation. They point to the potential to design safety ratings for chatbots, similar to ratings used in the entertainment industry for films, television and gaming.

They note the next two years will likely be crucial for regulating the technology, due to the increasing commercial pressure for companies to take full advantage of their ability to fine-tune large language models for increased engagement and economic gain.