If that life plan you agonized over has rather spectacularly fallen by the wayside in favor of endless sofa time, binge-eating on fast food in front of the TV, knocking back one too many drinks, and neglecting to make any contributions to the company pension, I think you urgently need to have a word with your future self.

MIT's AI-Chatbot Encourages Long-Term Thinking and Behavior Change by Simulating Conversations with Your Future Self

Now, MIT researchers have designed an AI-chatbot that impersonates the future flourishes of a user, essentially talking about what it thinks or how it shall provide insights and advice, if a future exists. The goal is for people to think more about who they want to be tomorrow, today.

Sporting profile photos altered with age to make young users look like elderly fuddy-duddies, it mixes together faux memories and their present-day aspirations to weave tales of a rich life driven in part by its imagined success. “Our ultimate goal is to inspire long-term thinking, long-term behaviour change,” says Pat Pataranutaporn, on the Future You team at the MIT Media Lab. This may spur individuals to act more pragmatically in the present, based on the long-term sustainability of their health and life circumstances.

During one conversation, the simulated 60-year-old version of herself was asking about the single most fulfilling moment in her career to a student who had aspirations of becoming a biology teacher. The bot claimed to have been a retired biology teacher from Boston, recounting a moment of reaching out to a poorly performing student and turning their grades around. “Seeing the student’s face light up with pride and satisfaction was absolutely amazing,” says the chatbot.

Users who wish to chat with the bot are given a quiz and asked about themselves, their friends and family, key moments in their past that defined them, and the kind of life they dreamed of living. From there users upload a portrait image, which the program ages digitally to picture the user at age 60.

It then takes the user answers, feeds them into a large model that then creates vivid synthetic memories that the simulated older self can recall to answer questions. That way, whenever the chatbot answers questions, it answers with a believable backstory.

The last piece to the system is the chatbot (almost — other than how it is powered by the GPT3 model from OpenAI). 5 who is another possible previous you, born and evolved, and now able to speak on its compressed life.

Pataranutaporn has talked to his "future self" many times, but said the most meaningful was when the chatbot pointed out his parents would not be around forever, so he should go see them. The session “literally changed my perspective, and it still does to this day,” he said.

Users are told that this “future self” is not a prediction, but is instead an optional future self stemming from the data that they have given. Participants are invited to deviate from their predetermined future by altering responses to the questionnaire.

In trials with 344 volunteers messaging the chatbot, people reported slightly reduced anxiety and fatter future self-association in chats, the researchers found, according to a preprint scientific paper on the project that is not yet peer-reviewed. They should make people will more informed decisions, Pataranutaporn said, from setting concrete goals and making exercise part of your daily routine to having better dietary habits and saving for the future.

People had a hard time picturing the individual they would be years later, but were in fact capable of showing more perseverance when it came to stay in school, adopting healthier habits, and carrying out more serious financial planning if they merely imagined who they would be, according to Professor Ivo Vlaev of the University of Warwick, who specializes in behavioral science.

He said the MIT project was one of “a number of interesting implementations” of behavioral science principles. ‘It actually exemplifies this concept of a nudge – minor changes implemented to change behaviour in a positive way – by making the future self more real and meaningful, especially to the current self,' he said. It is also true that if executed well, it promises to create a seismic shift in the way that people are making decisions today around their tomorrows.

As a matter of practical application, its effectiveness will probably depend on how good a job it can do in simulating relevant and realistic conversations, DeYoung added. “Should the chatbot be seen as genuine and intelligent, this may have a big impact on behaviour. When the interactions themselves are perceived as superficial or gimmicky, that impact might be limited."