Can Gen AI take us ‘Back to the Future?’

Can Gen AI connect us to our future selves and help support better decision making?

In the 1985 film, ‘Back to the Future,’ teenager Marty McFly accidentally travels back in time to 1955 where he meets his future parents and inadvertently disrupts their meeting, threatening his own existence. He must then ensure they fall in love while finding a way to return to his own time. Marty’s involvement in his parents’ past is a parable for the way current behaviours shape the future and how helpful it can be to have a strong connection to our future-self.

Even without the benefit of the film’s DeLorean time machine, it has long been known that people often feel disconnected from their future-selves, perceiving the older self as a ‘stranger’ as one psychologist put it. This disconnection can lead to poor long-term decisions, such as inadequate saving for retirement, unhealthy lifestyle choices, and a lack of planning for future needs. On this basis, finding ways to strengthen the connection with our future-self can help mitigate these issues, with the possibility of better decision-making and long-term benefits.

We might therefore welcome MIT Media Lab's Future You project which invited young adults to interact with AI simulations of their 60-year-old selves. These simulations, powered by personalized chatbots and featuring AI-generated images of their older-selves, answered questions, shared memories, and imparted life lessons.

According to a recent preprint paper of this project by Pat Pataranutaporn and colleagues, participants in the research found the experience emotionally rewarding, enhancing their connection to their future-selves, fostering a more positive outlook on the future, and increasing motivation to work towards long-term goals.

On the face of it, therefore, it sounds very promising – and an interesting use case for Gen AI that seems to have tangible value.  But just how does this actually work? Looking at the literature on temporal social science, there is a case to answer that perhaps our older-selves will always be strangers to some degree. And this tells us it is important to think carefully about the human experience when we design chatbots to fulfil increasingly intimate roles in our lives.

The risks of failing to face your future-self

Globally, the retirement landscape is changing due to increasing life expectancy and economic pressures. This prolonged period of retirement introduces risks such as outliving one's savings or experiencing a decrease in quality of life. And with that in mind, there is plenty of evidence that many people are not financially prepared for retirement. For example, in many countries, a significant proportion of the population fails to reach their target retirement income​.

The same considerations can also be made for health with a significant portion of older adults facing chronic health issues such as diabetes, osteoporosis, and heart disease, which can often be mitigated through preventive care and healthy lifestyle choices. But despite the well documented link between preventative steps and positive outcomes, many older adults do not engage in recommended preventive health behaviours. For instance, only about half of older adults meet the minimum guidelines for aerobic physical activity and muscle-strengthening activities​.

And in yet another area, we see that a significant portion of the workforce does not take full advantage of training programs to develop new skills, adapt to technological advancements, and improve their job performance – all important for long term employability. Notably, 59% of workers report never having taken any workplace training, relying entirely on self-taught skills​.

These findings could be explained in part by the ‘end-of-history’ bias which suggests we perceive our personal development and change as having reached a final stage, assuming that we will remain relatively unchanged in the future. In other words, we tend to believe that who we are now is essentially who we will always be, underestimating how much we will change in the future. Finding ways to correct this, connecting people with their future-self, would therefore appear to be helpful.

Ways to connect people with their future-self

Pataranutaporn and colleagues set out two types of intervention methods that have been used to date to connect people with their future-self:

• Reflective interventions: where participants write letters to or from their future selves;  these have typically been found to reduce negative emotions and improve future self-continuity, career planning, and delay of gratification.

• Presentational interventions: these use visual representations of a participant's future self, like age-processed images or avatars, to influence behaviour without needing active reflection. Studies have shown these can lead to increased saving behaviour, a preference for delayed monetary rewards, and improved financial knowledge.

However, the authors suggest there are a range of limitations to these approaches for increasing future self-continuity; reflective interventions need imaginative effort, whilst presentational interventions typically involve specialized software for age-processing images or hardware like VR headsets. Together these mean that the techniques are not necessarily very scalable.

To address this the Future You project aimed to create a realistic version of participants' future-selves by using a combination of AI-generated characters and age-progression technology, addressing the scalability challenges. Participants first provided personal information through a survey, including details about their demographics, life narrative, and future goals. This information was used to generate a ‘synthetic memory,’ a backstory that connects the participant's present self to their future self at age 60 (and in doing so addressing the imaginative effort challenge).

The system also asked participants to upload a portrait, which was then processed using a generative model that applies aging effects to the image to create a realistic representation of the participant at age 60. This age-progressed image, combined with the synthetic memory, allowed participants to interact with a personalized and believable future self in a conversational setting. The interactions were powered by a Large Language Model, making the conversations feel natural and personalized.

The study reported positive findings – participants interacting with the AI-generated future-self reported a decrease in negative emotions, alongside an increased sense of future self-continuity, feeling a stronger connection with their future selves, characterized by improved similarity, vividness, and positivity of their future self-image. The intervention also enhanced participants' sense of agency towards their goals.

Qualitatively, participants described the interaction as emotionally positive, with feelings of comfort, warmth, and solace. Despite recognizing the artificial nature of the chat, they found the interaction to be conversational and enjoyable. Overall, the brief, web-based intervention with an AI-generated future self significantly improved mental well-being and strengthened the participants' connection to their future selves.

So far, so positive. But does connecting people with an artificial version of their future selves work in the way we might assume?  Our analysis to date (such as here or here) suggests that things are not always as straightforward as they might first appear.

Can we go back to the future?

If we want to adequately represent the perspective of an older version of ourselves then it is surely sensible to look more carefully at the perspective of older people. The details of how the synthetic memory was developed for the Future Self project is not particularly clear, but from what information is given we can assume this is extrapolated forwards from our current self, augmented by information such as career ambitions, life goals, family plans.

Whilst this seems sensible in principle, it could perhaps be informed by work looking at the perspectives of older people. One such example is from Bronnie Ware, an Australian former nurse whose blog recording bedside conversations became the basis for a bestselling book. Ware suggests that older people at this life-stage seem to wish they’d been more authentic in their activities and prioritised friends and themselves, rather than work. While the Future You project did not seek to represent people on their deathbeds, but at the age of 60, nevertheless this may be more generalisable to our future versions of ourselves and as such should be part of the way we plan our futures. If we do not do so, then what future version of ourselves are we actually representing and who decides this?

Picking up the debate about the perspective of older people (and specifically those on their deathbeds) philosopher Neil Levy is sceptical of Ware’s reports for two reasons. He suggests that cultural pressures might lead people to report such regrets, whether they feel them or not. As Levy points out:

“The variously attributed saying ‘No one ever said on their deathbed I wish I’d spent more time at the office’ and close variations thereof return close to 40,000 hits on Google, indicating how deeply such sentiments resonate with us. That friendship and family and feelings are valuable is surely part of the reason why we value them. But the everyday banality of the advice gives us reason for suspicion.”

If cultural scripts are responsible for what’s reported or remembered, it may be that these they circulate so widely and are repeated so often, not because they come from older people, but that because they’re truisms that are attached to this group. On that basis, the notion that we should give these concerns special significance simply because they’re expressed by older people gets things back to front Levy argues, Perhaps we attribute these things to older people simply because we give them special weight.

This perhaps should give us warning not to fall into the trap of the cultural norms about older age groups, as baked into the assumption that certain life stages can be mapped out (you go to school, get a job, marry, bear and raise children, retire, and die) are all sorts of stereotypes that may well not be true for all groups.

Queer cultures, for example, have long lived according to “strange temporalities,” as Jack Halberstam has argued. His book, ‘In a Queer Time and Place’ argues that “queer uses of time and space develop… in opposition to the institutions of family, heterosexuality, and reproduction”.  On this basis, life milestones don’t necessarily follow those of heteronormative institutions; instead, time is transformed by time-bending experiences as varied as coming out, gender transitions, and generation-characterizing events such as the AIDS epidemic.

Quite how this, alongside the life experiences of other groups that sit outside conventional societal norms, can be referenced by synthetic memory is surely something that needs to be better described by the Future You project.

Hindsight bias

Levy goes on to give further consideration to why our future-self may not always quite be the source of good advice we might consider it to be. He cites Eric Schwitzgebel who suggests older people might be subject to hindsight bias, assuming that their current perspective looking back on the past is identical to the perspective they should have adopted at the time.

This means that if we take the advice (as set out above) that younger people should prioritise fulfilment over money then we can see that from the perspective of a financially comfortable older person this makes sense but how relevant is it to the anxious past-self, who may well struggling to make ends meet?

Of course, you might argue, why should we represent these sorts of views? After all, they are hardly going to encourage people to connect with their older self, save money, look after their health and take up training. Activities which prioritise fulfilment may not necessarily leave one in a good position in later years. It therefore seems reasonable to exclude them: but then how reasonable is it that we call the version of oneself that is being represented your Future Self?

It is a future version that seems to exclude the inconvenient parts that do not fit with the encouragement for taking sensible steps today.

Telic vs atelic

Levy offers the intriguing perspective that the view from people at later life stages might be the closest we can get to seeing life and its concerns from outside. From outside, we can’t grasp the significance of the day-to-day concerns of the younger-self: this is because most activities only have significance for the person who is confident they will live for some time to come. For example, saving money typically makes sense only if there is a future at which we might need to spend the money.

By contrast, the view from someone at a very late stage in life is the perspective of someone for whom simpler pleasures that can be realised relatively quickly are important. In a book on the midlife crisis, Kieran Setiya argues that these crises can arise because, as we realise our ambitions, they lose meaning for us. Setiya describes these ambitions as telic projects in that they have a goal, and it is our focus on that goal that makes them meaningful to us. But once we have met that goal, they then seem meaningless. Setiya suggests therefore need to find value in the atelic, where there is not a goal beyond the activity itself (such as going for a walk for the sake of it, rather than to get somewhere).

The perspective of the older person reflects what matters for those who have a shorter temporal horizon so perhaps they have less good advice for people at earlier life stages who are deeply involved in telic activities. Sitting outside the telic, only the atelic (which involve short-term goals) seems important. As Levy points out, things such as companionship, contemplation, beauty remain available to the older person and they take on an extra force. But their perspective is not necessarily relevant to the younger self as their commitments to telic commitments can only be fully understood from those at this younger point in their lives.

In conclusion

Generally this raises a host of questions for how we go about developing synthetic versions of people – that is all too easy not to see or consider. If people at a later stage of life may not be the most reliable sources of advice for people at earlier life stages, then what should we do?  It seems fairly unlikely that the sort of advice often attributed to people at very late stages of their life is something that will motivate younger people to save more, for instance.

Instead, drawing on motivational theories may offer broader perspectives to help inform a more dynamic and evolving nature of how synthetic version of people may offer motivational forces. For example, identity-based motivation (IBM) theory (see this paper for a recent review) outlines how action can be present or future-focused depending on how accessible and relevant a future-me seems to the current-me, and how difficulty might be interpreted in that moment. This process is a dynamic construction based on the context, which is recursively forms as action is taken. As such it evolves and shifts as behaviours unfold.

Drawing on IBM theory in this way moves away from a narrow view of the older self – and perhaps reduces the potential for hindsight bias or the way in which more risky advice could be given. But, as suggested, this is arguably not an older version of the self that is being created but a partial version that is designed with specific outcomes in mind.

Time travel is a theme that has long attracted writers, philosophers, psychologists and now technologists. And the Future You project seems to offer much promise to long standing challenges where we struggle to clearly imagine the future. But who and what is determining the way our future looks is something that deserves the kind of scrutiny that behavioural science can offer.