Stability, Short-term and Long-term memory

I got some positive comments on my previous post, so I decided to continue publishing in this format for a while. Now that I’m back in Tokyo, I might even post on the regular blog again: do not despair at this sudden change, French followers!

Today, I would like to talk about a theory of stability in the brain. In my previous post about the Chinese room experiment, I said that I have a problem with this: your brain is always changing, yet “something” seems to remain stable. This stability shows in the consistency of your behaviour, opinions,  tastes, ideas… These are part of your identity. Indeed, you probably feel as the same person waking up every morning and one year to the other, despite all the cellular changes in your brain and the rest of your body!

From here, I will give my own interpretation of existing research and explain my theory on the subject. Feel free to disagree and criticize it.

One hint about the stability of identity could be in recent research on behavioral economics. Several studies show that actually, our identity is not as well defined and fixed that we previously thought. In these experiments, they find that we take cues from our past behaviour to try to define what kind of person we are. When faced with a decision (“buying a drink”), you rely on what you did in the past (“I bought coffee from this shop last week”) to define your identity (“If I bought coffee there, it must mean that I like this shop”) and take a decision in the present (“I will go have a drink there, since I like the place so much”). Other example: you find yourself morally obligated to give to a charity, because all of your friends gave money to that same charity. You deduce that you’re a generous kind of guy, otherwise why would you give to charity? The following week you pass by a homeless person, and decide that giving them money is the obvious thing to do, because you’re a generous person. Note that the usual relation “I have this personality trait, therefore I act like this” is reversed into “I acted like this, therefore it must mean that I have this personality trait”.

Not convinced? There are lots of papers out there showing that this is actually what happens in our brains (I think the term is coherent arbitrariness, I recommend Dan Ariely’s books if you’re interested). The most compelling proof is that it works even when the decision in the past was *not* yours but you *believe* it was yours, and even if the past information has absolutely no link with the choice you must make in the present! I can actually think about some parents using this kind of tricks with their kids (“I know you will behave, because you’re a good boy”or”A nice boy like you does not hit his friends, does he?” can actually convince a rough kid to play nice), and some adults too. And it’s not just about the fear to disappoint.

We act as if we don’t know what kind of person we’re like, but instead have to guess it from our past choices. A series of psychology experiments also showed that if you can convince someone that they chose X over Y in whatever situation the past, and then ask them to justify that choice, they will come up with reasons why they absolutely prefer X and why X was obviously the rational choice for them and how they would totally choose X again if in the same situation… Even if in reality, what they had chosen was Y. We also use other hints from our environment to “guess who we are”, like the opinion that others have of us, or what objects we own, but ultimately  it seems to be reduced to choices we made.

So identity is not a rock hard concept embedded in you, but rather something that your brain tries to reconstruct when necessary. And, oh look! Unsurprisingly, the same goes for memory. A long time ago, science thought that memories where stored in the brain and retrieved as a block when necessary. But all recent science paints a very different picture. Memories are reconstructed from diverse bits every time you need them, not retrieved intact from some storage place in your brain. And this reconstruction itself tends to modify the memories. This is partly the reason why it is possible to induce false memories of the past in people just by influencing their present environment (yes, science did it… and it worked).

To summarize, your image of yourself (memories, preferences, personality traits) is not fixed, but gets reconstructed by your brain in real time, using whatever is found in your memory. This memory itself tends to change when accessed. You then base your present and future behaviour on this reconstructed image of yourself. Therefore, the stability I was looking for in the last post emerges from a loop:

– You make choices that get summarized and partially stored in your memory

– You rely on these memory bricks to re-build memories of what you have done in the past

– You use these memories to build an reconstruction of your own identity

– You make new choices based on this identity

– Start again from the beginning of the loop.

If this is true, it shines a new light on the effects of a disease like Alzheimer. People with this condition start by losing their ability to build new memories. They become unable to remember what just happened, then forget events further and further in the past. Family members often say things like “this is not the person I knew” or report feeling like talking to a stranger. Well imagine what happens if you become unable to complete the loop above because your ability to form new memories is impaired. You make choices that you forget immediately. You are forced to rely on older and older memories for identity-building, but the memory-bricks of these memories change when you access them and fail to be re-stored, making even that part of you more fragile. At some point you’re left with only some disparate fragments of old things to build your identity with. Then it’s not surprising that you’re a total stranger for people who thought they knew you.

This also works for some forms of amnesia. People remembering abilities (speaking their mother tongue, riding a bicycle), but with no memories from their previous life. If tastes and preferences were hard wired in the brain somehow, you would expect these persons to still like the same food and appreciate the same hobbies. But it is not what happens. Instead, they end up with a brand new set of tastes and a whole new personality, an new identity. Which would not be very surprising if my stability loop really is what happens in their brain, in mine and in yours.


I would like to finish this post by a part on artificial intelligence, of course. I think the following network-based system would be able to exhibit a form of stability, despite underlying instability:

– First we need a short term memory, to store recent events quickly. Add a necessarily slower long term memory with a “summarizing system”, to analyse the short term memory, store the bits that look important, and erase the rest. We also need a mechanism to re-build past events from bits found in the long term memory.

– Our system’s behaviour must be heavily based on rebuilt past events. By making choices, the system imprints it behaviour on the environment. These prints then go through the memory system.

Then my prediction is that even if the structure of the system changes drastically and completely over time, and even if storage errors tend to happen, and even if just about everything is unstable in the system, the behaviour of the system will be very stable. And if among stable systems this one is the minimally complicated, then the result of artificial evolution project I described in my previous post should look a lot like this one. This is important, because to be useful, a theory must be able to make predictions. Then if the prediction is correct, it is a good sign that there is some truth in the theory.  My theory is that the “stability loop” exists in humans and in just about every biological system in need of stable behaviour, and my prediction is that it emerges when there is selection pressure for stability applied on unstable systems.

This is the end of this post. I should definitely add the references to the papers I’m mentioning, but then writing these posts would take me as long as writing an actual paper!


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s