If your carefully laid out life plan has been ruined by spending time on the couch, bingeing on fast food, drinking too much and not contributing to the company pension, it may be time to have a chat with your future self.
Without easy access to a time machine, researchers at the Massachusetts Institute of Technology (MIT) have built an AI-powered chatbot that simulates a user’s aging and offers observations and pearls of wisdom. The goal is to encourage people to think more today about the person they want to be tomorrow.
Using a profile image digitally aged to show young users as wrinkled, white-haired seniors, the chatbot generates plausible synthetic memories and draws on the user’s current aspirations to tell stories about their successful life.
“The goal is to promote long-term changes in thinking and behavior,” said Pat Pataranutaporn, who works on the Future You project at MIT’s Media Lab. “This could motivate people to make smarter decisions in the present that optimize their long-term well-being and life outcomes.”
In one conversation, a student hoping to be a biology teacher asked the chatbot, a simulated version of her 60-year-old self, about the most rewarding moment of her career. The chatbot said he was a retired biology teacher in Boston and recalled a special moment when he helped a struggling student improve her grade. “It was very gratifying to see the student’s face light up with pride and accomplishment,” the chatbot said.
To interact with the chatbot, users are first asked to answer a series of questions about themselves, their friends and family, the past experiences that shaped them, and the ideal life they envisioned for the future. They then upload a portrait image, which the program digitally ages to produce an image of the 60-year-old user.
The program then feeds information from the user’s responses into a large language model that generates rich synthetic memories for the simulated older self. This ensures that when the chatbot answers questions, it is based on a consistent backstory.
The final part of the system is the chatbot itself, powered by OpenAI’s GPT3.5, which presents itself as a possible older version of the user that can talk about their life experiences.
Pataranutaporn has had several conversations with his “future self,” but he said the most profound was when the chatbot reminded him that his parents wouldn’t be around forever, so he should spend time with them while he could. “The session gave me a perspective that still impacts me to this day,” she said.
Users are told that the “future self” is not a prediction but rather a potential future self based on the information they provided. They are encouraged to explore different futures by changing their answers to the questionnaire.
According to a pre-printed scientific article about the project, which has not been peer-reviewed, trials involving 344 volunteers found that conversations with the chatbot made people feel less anxious and more connected to their future selves. This stronger connection should encourage better life decisions, Pataranutaporn said, from focusing on specific goals and exercising regularly to eating healthily and saving for the future.
Ivo Vlaev, professor of behavioral sciences at the University of Warwick, said people often struggle to imagine their future, but doing so could drive greater persistence in education, healthier lifestyles and more prudent financial planning.
He called the MIT project a “fascinating application” of behavioral science principles. “It embodies the idea of a nudge (subtle interventions designed to guide behavior in beneficial ways) by making the future self more salient and relevant to the present,” he said. “If implemented effectively, it has the potential to have a significant impact on the way people make decisions today with their future well-being in mind.”
“From a practical point of view, effectiveness will likely depend on how well it can simulate meaningful and relevant conversations,” he added. “If users perceive the chatbot as authentic and revealing, it could significantly influence their behavior. However, if the interactions seem superficial or gimmicky, the impact could be limited.”
