The Constant Work of Social Connection
We tend to think of our relationships as static things—like entries in a digital address book. Once added, a friend is a friend. But that’s simply not how the human brain treats them. Our social landscapes are deeply dynamic, shifting with every interaction, every shared laugh, and every quiet disagreement.
To keep a handle on where we stand with the people we care about, we have to do constant cognitive heavy lifting. Think about it: the version of your friend you knew a year ago isn’t the version you’re dealing with today. Between then and now, jobs have changed, opinions have evolved, and life has happened. If we don’t actively update our mental models of these relationships, we’re essentially navigating our social lives with an outdated map.
Understanding our social standing requires that we constantly refresh our autobiographical memories. It’s what allowed us to stop asking "are we still actually friends?" in the same way we might have wondered about it in middle school. Instead, it’s a subtle, ongoing tracking of whether recent interactions have been satisfying or strained. This isn't just about social maintenance; it's about the fundamental way we keep our internal lives tethered to reality. Without the ability to accurately refresh these memories, we’d be adrift, unable to bridge the gap between who someone was and who they are right now.
How We Successfully Update Our Personal Histories
Human memory is a remarkable piece of machinery. It’s not just a passive repository; it’s an active, ongoing system designed to prioritize what's relevant. A key part of this is something called the "recency effect." We naturally weight our most recent experiences more heavily than those from the distant past.
For the most part, this is incredibly helpful. Take a practical example: finding your car in a packed parking garage. You don't need to recall every single spot you’ve ever parked in to find your vehicle. Your memory is effectively tuned to "now." You’re searching for the most recent instance of parking. If your brain tried to retrieve every historical parking location simultaneously, you’d be paralyzed by cognitive noise. By favoring the present, we stay functional (da Costa Pinto & Baddeley, 1991).
This same mechanism plays out when we face the mundane—but occasionally tricky—question of where to go for dinner. Should we return to the pizzeria we loved last month, or is it time for something different? We weigh those recent memories: Was the service last time actually as good as I remembered, or was it a bit rushed? Did the food really hit the spot, or are we just creatures of habit?
We’re constantly comparing the past against the present. When we meet friends we haven't seen in months, we naturally adjust how we talk about our lives based on that elapsed time. We share the major milestones rather than the daily grind, because our internal context tracker understands the difference between a daily check-in and a long-overdue catch-up. It's a sophisticated balancing act, one that we perform nearly every day without ever stopping to consciously think about the machinery behind it.
The Challenge of Overturning 'Zombie Ideas'
If updating our personal lives feels like a natural skill, updating our broader understanding of the world is surprisingly difficult. Scientists and researchers in every field struggle with what we might call "zombie ideas"—outdated concepts that are proven false, yet refuse to die.
You’ve likely come across one. It might be in a dusty textbook, a lingering piece of pop psychology, or a deeply held conviction that’s simply been superseded by better data. The challenge isn't just learning something new; it’s the active, uncomfortable work of discarding what we previously believed to be true.
This is the very heart of the scientific method. Progress happens not only when we make new discoveries but when we are willing to let go of old theories that no longer fit the facts. In my own teaching, I often find it necessary to present an old, discredited theory before walking through the newer evidence that challenges it. It’s a necessary process, but it rarely feels easy. We like our mental models to be stable; having to rebuild them based on new evidence feels, at best, like a persistent housekeeping chore and, at worst, an assault on our worldview. Yet, without this willingness to update, our knowledge doesn't just stagnate—it actively leads us astray. We end up clinging to theories that don't reflect the reality of the world as it currently exists.
AI and the Struggle for Current Knowledge
As I’ve been thinking about the human capacity for memory updating, I’ve also become curious about how AI systems grapple with these same challenges. After all, Large Language Models are built on staggering amounts of information. They are, in a sense, the ultimate curators of historical data. But that strength can become a limitation when it comes to adapting to new truths.
When an AI model is trained on a massive, static data set, it naturally prioritizes the sheer volume of information it has been fed. If that data set contains decades of popular (but discredited) literature on a topic, the model is likely to reflect those biases. Even when new, accurate research contradicts those ideas, the model may struggle to suppress that massive pile of older, "well-known" (but wrong) information.
For instance, ponder the old "carpentered-world hypothesis," which suggested that only people raised in environments full of right angles would perceive certain visual illusions. When I’ve tested AI on this, it often defaults to that outdated explanation because it is so prevalent in the historical record, even when the model contains the data about more recent, comprehensive studies showing the illusion is human-universal (Amir & Firestone, 2025). Similarly, myths like the "Mozart effect"—the idea that listening to classical music makes kids smarter—still linger in the answers you might get, despite extensive evidence debunking it (Mehr et al., 2013).
This is a critical distinction: AI is incredibly good at aggregating knowledge, but it is not necessarily designed for the active, vigilant updating that defines human cognitive resilience. If we rely on these systems to give us the "current state of the world," we risk becoming trapped in an infinite loop of zombie ideas that no one thought to explicitly update. AI might just make it harder for us to let go of the past.
Remaining Vigilant for Social and Intellectual Health
Ultimately, the ability to update our mental maps isn't just a quirky cognitive feature—it’s essential. It is what separates a flexible, responsive mind from one that is rigidly stuck in the past.
Whether we’re talking about mending a friendship, parking the car, or refining our scientific understanding of the world, the requirement for active, honest assessment remains the same. We have to be willing to look at the new evidence, even when it’s inconvenient. We have to be prepared to admit that our previous understanding was insufficient.
It takes work. It’s uncomfortable, and it often feels like we’re fighting against our own desire for stability. But it’s the only way to remain genuinely connected to the world around us. So, the next time you feel that urge to cling to an old idea or an outdated assumption about a friend, pause. Ask yourself if you’re operating on the best, most recent information, or just relying on an old, dusty map. The health of our relationships—and the accuracy of our knowledge—depends on us showing up and doing the updating.