Extreme amnesia cases, AI, and our imagined futures: in conversation with a Harvard memory researcher

Yahoo

17-06-2025

We tend to think of human memory as if it's one of those old steel filing cabinets: some information gets stashed inside, and when the time comes, we hope we can find it by flipping through the tabs on a few billion neuron-supported manila folders. But the truth, as science has learned—and continues to learn—is that memory is more than just the attic of our minds. We're realizing that it's a foundational part of how we interpret and imagine our futures. And few have done more work to unlock this powerful reframing of human memory than cognitive psychologist Daniel Schacter.
What he and his team at the Schacter Memory Lab at Harvard continue to tease out in their research is a picture (almost literally, thanks to fMRI) of how memory works in our brains—and that picture looks remarkably like how we imagine.
The systems largely overlap, which implies that memories—fallible, mutable, and spread across almost every region of the brain—are being accessed, consciously and unconsciously, while we do everything from engaging in creative pursuits to problem solving. He calls it 'constructive episodic simulation' (with 'episodic' memories being personal moments we recall, vs. 'semantic' memories, which are more facts and meanings.) The upshot: our memories aren't sepia-toned artifacts, but modular building blocks. And our brains are using them like a 5-year-old plays with Lego—no instructions, and plenty of experimentation.
And just to complicate matters more, Schacter's research has implied that the inherent messiness of memory—which he organized in his 1996 book The Seven Sins of Memory—may actually be a feature, not a bug. (Even if it doesn't feel that way when, say, you blank on the name of a person you've known for years.)
Intrigued by all of the above, I spoke to Schacter recently to better understand this expansive notion of memory and imagination and how they function on a 'common brain network.' He has the professorly mien of a man who's been teaching psychology for 34 years at Harvard: specific and circumspect, with an owl's suspicious glare but the patience of a saint when faced with a journalist trying to tease out the mysteries of memory that Schacter's devoted nearly a half-century to exploring.
In the process, Schacter touched on what A.I. and memory might actually have in common, the effects of being force-fed memories by our phones, and the acutely amnesic patient who helped inspire his research.
This interview has been edited and condensed for clarity.
What's your first memory of doing memory-focused research?
One of the first patients I tested. It was the first one. He seemed like a fairly bright guy—had a normal conversation, like we're having now. He didn't do very well in recalling, like, a word list, for example. That wasn't surprising. But what was surprising was when I had to go out of the room for some reason, and came back like two minutes later, he had no idea who I was. He had no idea what we'd been doing. And that is really what caught my attention. When you see a patient with a memory disorder that severe—who in other respects seems normal—carrying on a pleasant conversation, but is that impaired? Yeah.
What was your first eureka moment in doing this research?
That involved various observations related to what I later called 'implicit memory,' which was something that had been observed clinically in amnesic patients—they would show some effect of a prior experience without really having any recollection of [it]. That hadn't been named and really crystallized, but there were indications in literature from the '70s that even though these amnesic patients couldn't tell you what happened 10 minutes ago, they nonetheless could be impacted by it.
One of my first direct experiences with that was a patient who I tested, as a graduate student, who exhibited this related phenomenon that we call 'source amnesia.' So this is where I would tell the patient some obscure fact or a made-up fact; 'Bob Hope's father was a fireman' is one from some of the experiments we did. Time would pass and I'd say, "Do you know what job Bob Hope's father had?' And the patient would say, 'Oh, I think he's a fireman?'
[Here, Schacter pantomimes a conversation:]
'How do you know?'
'Oh, I heard it on the radio or a friend told me that a couple years ago.'
'Did I ever mention that?'
'No, no, you never mentioned that.'
'Well, actually, I just said it two minutes ago.'
So the first time you see that with your own eyes, that's pretty impactful.
What, for you, in these past 50 years has been the biggest change in how science understands memory?
During the time period that I've been working in our lab, we've been focusing less on memory just as a repository of information about past experiences or a retrieval of stored information, and looking at it more for the role it plays in thinking ahead to the future—simulating possible future experiences. For its predictive aspects. You're using information to think ahead: how you want to go about solving a problem or planning your day. Many of the same brain regions that support your ability to go back and remember past experiences [are the] very same brain regions involved in you imagining your future.
I get the feeling that memory is now understood to not be just a single discrete boxed-in function within our brains, and that in some ways it's maybe the foundational operating system of how our brains consciously and subconsciously work.
Maybe that's a little bit expansive, but what you've been studying lately around memory's link to creativity and imagination seems to point in that direction.
I think it does point in that direction. We're using the term 'memory,' here, but we have to keep in mind there's not just one memory. Certainly there was a distinction out there between short-term and long-term memory. As we got into the 1980s we focused on the implicit versus explicit memory distinction, where under 'explicit' you could group 'episodic' and 'semantic'—and under 'implicit,' a whole bunch of different things: priming procedural learning, some kinds of conditioning…
That goes back to what we were talking about before: that amnesic patients, for example, can show intact implicit memory without any corresponding explicit memory. So when we talk about memory, we always have to remember: it's not just one thing. That's something that I think is better understood as part of the way we think about memory since I've been involved in the field.
Once you start to dig into the concept of memory, does it start to feel slightly philosophical for you?
Well, I mean, there are always elements of philosophical perspective when we're talking about these kinds of distinctions, but we do try to root it in empirical observations. I would say probably for me, the first half of my career was more focused on different kinds of memory—trying to link up different kinds of explicit and implicit memory with different brain regions. And the second part has been more focused on looking specifically at how we use our episodic memories in a flexible way—to take bits and pieces of past experience and recombine them and construct simulations, and all of that stuff.
Take me down that path a little bit—the notion that memory underpins our imagination in a powerful way.
When I was in Toronto and we were at the unit for memory disorders, my interests were mainly focused on implicit versus explicit memory and source amnesia. There was one patient who we were very interested in back then. He was known in literature by the initials K.C.—that was short for Kent Cochran. Kent was a young man who, in the early 1980s, had a head injury in a motorcycle accident, and he happened to have brain damage that produced one of these very severe amnesic syndromes. It's fair to say that he could not remember a single specific episode from any time in his life.
There was a testing session [in] 1983 or 1984 where [psychologist and neuroscientist] Endel Tulving and myself were there, with K.C.seated on the other side of the table with us. Tulving asked him this seemingly innocent question: Tell me what you think you're going to be doing tomorrow.
Now, we know that when you ask K.C., tell me what you did yesterday, he'll say: 'I can't remember any one thing I did. Maybe I had breakfast, then I had lunch.' A script-like response. And the same thing happened when Tulving asked him, tell me what you're going to do tomorrow, K.C. just said, 'Well, I don't know.' If you pushed hard enough, he would eventually say, 'Well, maybe I'll have breakfast and lunch.' But he couldn't conjure up any one specific episode of something he might do in the future, just like he couldn't remember what he had done in the past. That was very striking, and suggested a role for episodic memory in imagining the future.
The question was, how do you study that? We as memory researchers knew how to study memory for past events—but how do you study imagination of future events? So I put it on the back burner. Nobody in the field was particularly interested in the issue of memory and future thinking [at that time].
Then, in the early 2000s, some people started publishing papers on the similarities between remembering the past and imagining the future. In 2005, I had a new postdoc in the lab by the name of Donna Rose Addis, and she was doing functional MRI studies of autobiographical memory. It'd been in the back of my mind for 20 years to look at the relationship between remembering the past and imagining the future, so the two of us talked and thought, hey, what if we do a standard autobiographical memory experiment where we've thrown in a future imagining condition? You give a keyword, and in some trials the subject—while being scanned—is asked to remember a past experience, and in other trials is asked to imagine a future experience, and in still other trials you give them control tasks that don't involve remembering or imagining.
And we found this really striking result of similar brain regions showing increased activity when people remember the past and imagine the future. We published our paper in 2007, and that was the year the field got interested in this question as a result of our study. That really then set the agenda for the 18 years since then, and we've been continuing to look at this issue in a variety of ways.
As you saw those fMRI scans come in and saw that overlap, was that like a real tap-dance moment for you?
I mean, you're not analyzing it subject-by-subject, but…yeah, that was very striking.
One of the theoretical ideas that has guided us, that we put forward back in that paper in 2007, was that maybe one of the reasons we experience certain kinds of memory errors is, it's a byproduct of a memory system that's set up to allow us to use our past in very flexible ways—to recombine bits and pieces of experiences so that we can simulate or imagine novel experiences in the future. Because the future is rarely identical to the past, so we want to think about how we're going to deal with new upcoming situations that we haven't experienced before.
A byproduct of an adaptive system that generally works well and allows us to flexibly use our past experience to think about the future may be that we're prone to certain kinds of memory errors, when elements of different experiences get miscombined.
What do you feel like the average layperson gets wrong about their concept of memory?
I think the general idea that memory is more or less like a data recorder or a photograph that fades. We all know we don't remember exactly what happened in every detail—that [our brain] more or less records what happens, but it fades over time. That's what I call transience in [my book]The Seven Sins of Memory. But I think people are less aware of some of the other influences that can change or distort memory. I think one of the questions you raised about self, one of the seven sins I refer to, is 'bias.' It takes various forms, but one is, for example, tendencies to remember the past in many situations as better than it actually was, or in ways that bolster our memory. I think we're not aware of a lot of the top-down influences on memory that exist.
One of the very recent studies from our lab was a collaboration with another lab here at Harvard, led by Jill Hooley, a clinical psychologist. A couple of our graduate students got together, and we looked at the impact of grandiose narcissism on remembering the past and imagining the future. And the upshot of this study was that people who score really high on grandiose narcissism, who think they're the greatest thing in the world, tend to remember the past and also imagine the future in a highly exaggerated positive way compared to people who were lower in narcissism.
It does sound like a blessing to feel so optimistic about the future.
Right, that's right.
And there's a lot of work along those lines that shows how our concepts of ourselves bias the way we remember past experiences. And then we and others have been showing that that same influence exists when we imagine into the future. But I think that's one of the things that probably is hard for us to just grasp intuitively, how our memories are changed or affected by our sense of [ourselves].
It's fascinating to think that memories can take on these different casts and tones based on our own feelings about ourselves, and create a cycle that feeds into itself psychologically.
Along these lines, one of the underappreciated aspects of memory concerns the potency of the act of retrieving a memory. That retrieval can do various things, but among them, change a memory. So retrieving memory is not just a neutral event. It's not like bringing up a file on your computer and then putting it back with no changes. Depending on the circumstances, retrieving a memory—talking about it—can introduce all kinds of interesting distortions potentially into the memory.
Related to that, I've noticed how so much of our technology force feeds us memories. My partner has that widget where a different photo shows up on her phone every time she turns it on. She'll say, Oh, remember this? And it's from a vacation we took years ago. Do you think that there's an effect on memory when we're being fed these moments, but outside their context? Is it deepening them? Is it warping them when a photo comes up and I'm thinking about the memory on a Tuesday during a Zoom meeting?
I think there are several things that can happen. One, there's potentially a strengthening effect for some aspect of the memory. You're reminded that this event took place and maybe the information that's in the photo that you're looking at becomes strengthened. It could potentially distort your memory because of what's not shown—you know, maybe there are other important things that took place in that event that aren't in the photo. So you start remembering it in a different way than you would have otherwise, had you just been thinking about it on your own.
And we know there's a really interesting phenomenon that's very well established, called retrieval-induced forgetting. And this is the idea that when you activate a memory, information related to that memory that you don't retrieve may become more difficult to retrieve later on.
Every time one of those photos comes up, it's reshaping that memory in a way I may or may not intend.
In interesting ways, yeah. Strengthening some aspects, weakening others.
I've got to ask about A.I., which seems to be everywhere right now. When it comes to large language models like ChatGPT, A.I. seems to mirror the concept of memory: it's this murky, probabilistic process based on really deep wells of information, though you have no clue what it's accessing and not. It hallucinates in ways that nobody really quite understands. It works off of both literal prompts and subtle, sometimes unintentional context cues in the same way that memory seems to. Do you look at these A.I.s, and think that they're close to the notion of memory the way we understand it?
Well, they're operating in a slightly different way than we are, I think. They're certainly capable of making some interesting memory errors. And, you know, that may be one of the telltale signs of relying on A.I. The field of cognitive psychology has gotten interested in the question of: To what extent does A.I. mirror human cognition?
We have one published study where we were looking at how people come up with creative stories in response to just a few word cues. As in: Try to write a creative story based on these three unrelated words. And we did it with ChatGPT-3 and ChatGPT-4. Could you tell the difference, really, between the creativity of these stories that humans came up with and what the two large language models came up with? The answer was no, they were about equally creative.
And there's the interesting question of, why do they hallucinate? What are they using as a criterion to say, yeah, that reference is what I'm looking for. I don't think we understand all the deep underpinnings of large language models enough to compare them in detail to human cognition and human memory. But they do make some of the errors that look like the kinds of mistakes people make.
If a genie granted you one wish, what's the one question about how memory functions that you'd want answered?
Wow. That's a tough one.
Or maybe to phrase it a little bit better, what's the one answer that would unlock something for you about our understanding of memory? That maybe, because of technology or the limits of science at the moment, you're stuck on.
You know, I think for me, it would be really having a deeper understanding of how we go about pulling together these different aspects of experience to turn them into, for example, future simulations.
As in, what route do we take?
Yeah. What exactly are the underlying neural pathways that are involved? How do we go from that mode of retrieving an experience and traditional sense of memory? What, at the level of underlying brain pathways, distinguishes that from using information to go back into the past? How does that differ—at the level of the relevant neural processes—from using information to think about the future or solve a problem? What is it that we do that allows us to shift those different modes of retrieval? I think that's one that really interests me.
This article is part of Your Memory, Rewired, a National Geographic exploration into the fuzzy, fascinating frontiers of memory science—including advice on how to make your own memory more powerful. Learn more.