"The Proactive Brain: Using Analogies and Associations to Generate Predictions"
November 27, 2007, 4-6PM
Note-taker: E. I.
Last Updated: January 17, 2008
Q&A
[Q = Question, A = Answer, C = Comment, R = Response]
Q. [I agree with most everything you said. I was wondering however, whether it is better to substitute your notion of “prediction” with a broader term, “anticipation”?] So you might say that anticipation is almost a prediction but not quite. Could that be because anticipation is not always conscious?
A. I’m not sure I agree with you that prediction is always conscious.
Q. If prediction is unconscious, how can you say it is a prediction?
A. Well, even a reflex is some sort of prediction. And this is done at a very limbic, superficial level without one being aware of it. Thus, I do not have any evidence to argue that all prediction is conscious.
Q. I try to draw the distinction between prediction that is more or less conscious and anticipation that is not necessarily conscious. Even in the resting state there is a background activity [of the brain] that always makes our brain in a state to anticipate something. This anticipation is not necessarily conscious, and it is related to a kind of preparedness, where one is alert, prepared for something but not necessarily clear of what it is going to be. But once the input comes, which is something that fits into the spectrum of what was anticipated, then it will select [[relevance]] with the anticipation. My major point is trying to draw a distinction with your notion of prediction with an even broader notion of anticipation in terms of preparedness and in terms of the broader background activity [of the brain]. What do you think [about substituting your “prediction” with my “anticipation”]?
A. For us, many processes, as we do them more often, they become automatic/non-conscious. e.g. driving from home to work, your body starts to perform without…
C. You mean habituation.
A. Habituation, but your body starts to perform it without bothering you. There are psychological measures to determine whether or not something is automatic. This seems to indicate that not all predictions are necessarily conscious because there are predictions that we do often. For example, if you are about to bite into an apple after having done it 100000 times in your life so it is automatic. You don’t think consciously what the apple is going to taste like, your body anticipates it and prepares its state unbeknownst to you.
Anticipation can be more general, more vague, less specific than prediction. So, it may be possible to come up with a finer definition where anticipation is less specific and preparedness is more [specific]. For example, “Something big is going to happen tomorrow” would be anticipation, which is less specific than “I’m going to a party/movies tomorrow” (prediction). And in this regard, anticipation is less useful.
C. However, “anticipation” may be more appropriate to describe brain state which has no explicit cause, but is just background activity. We cannot stop making associations.
A. The brain doesn’t know what to do with a general non-specific anticipation, or arousal state (e.g. amygdala). Akin to general anxiety—where people are just feeling anxious in general, not from anything specific. I don’t see anything much a cerebral cortex can do with non-specific anticipation, other than just arousal, making the system more alert, making sure the system will respond, but that seems quite wasteful (since the whole system will be fired up). I agree with you that there seems to be a difference between anticipation and prediction, and that we would have to investigate more carefully the specifications of tasks performed by the brain.
Q. Different contexts lead to different associations. How do these context dependent differences in associations show up in the fMRI?
A. The associations will likely look more like a switchboard than the map that was shown in the slide. What are consistently activated are the three regions that I’ve shown. Regarding context, the more contextual cues the experimenter provides, the more [strongly] the associations will be made.
Q. I will have particular associations when I am shown a cat, and will have other associations when I am shown a hairdryer. What is the difference in the fMRI?
A. So your question is, how is the activation different when I show you a chair vs when I show you a slide projector. One way to investigate it would be to show you hundreds of slides of a chair and find out what is consistent. In this case, the visual cortex will recall the specific features of these items. Regarding the association areas—for certain types of association, it will activate in a certain way. What we will see is an average of all types of activations of our cortex for all of our associations.
Q. What influence does genetics have on these associations vs regularity [environment]? Do you think the underlying brain formation has a direct impact on how we experience things? Would an infant have a very different experience than adults due to the lack of associations it (and the brain) has made in its life?
A. What we have found out so far clearly suggests that the more experience you have, the more associations you will potentially be able to make. If you see something for the first time, you may have no idea what it is. For example, my child did not at first understand that my gesture of pointing indicated that they were supposed to look at the direction I was pointing in, not look at my finger. With experience, they learned to associate pointing as a directive gesture.
Q. What about perception itself? I’ve heard a story that when Columbus came over to America, the Native Americans couldn’t see the ship because it was novel to them. Are children with less experience, likewise, unable to detect stimulus that they have no previous associations with?
A. I’m not sure about the accuracy of the ship story, as it would have to be the case that the ship was nothing whatsoever like what the Native Americans have seen (including figures, shadows, colors, contours). For the situation to work, the novel situation must be completely novel (like showing something in infrared when we’ve only been exposed to visible light). Some is due to development, some is due to genetic wiring. For humans, we also know that some of the cortex is pre-wired. We are born with some basic tools, including some associations.
C. I think the argument here is that if every experience depends on associations, no truly novel experience is possible.
A. I think this again comes down to the problem of definition. In the absolute sense, no novel experience is possible because if you show me anything that I had never seen before (e.g. a 3-D movie), I will still be able to relate it (or elements of it) to something I have experienced previously.
Q. What is completely original about your results?
A. Just like a novel electronic devise is dependent on the production of multiple different parts, my results exist because of prior works/discoveries of multiple other scientists. However, this philosophical vibe makes me think of the famous question about the blind man who was suddenly able to see.
C. Have you read “Mary the Neuroscientist” by Jackson? What will Mary the neuroscientist see if she is only exposed to black and white, and is one day exposed to red?
A. This is the only think I can think of that is close to a truly novel experience. If we acquire a new sense (e.g. see electromagnetic waves [beyond visual]). The question about vision is really a philosophical question, because how can they suddenly “see”…It almost means that novel experience is experiencing something without the sensors.
C. Regarding that point, I have heard that the literature is controversial, but I have also read an article about a person who was blind most of his life, who had surgery to be able to see. In his case, after the surgery, he was able to see splotches of light, but was not able to associate them with anything. He thought an inanimate object was a dog, he couldn’t identify faces. I’m not sure it got any better, but my guess would be that his brain may have been too rigid to learn too many new associations.
A. Most people like this who make this kind of decision get majorly depressed. It may be the case that you may not have plasticity to learn these things to associate, so in some ways this person is like an infant who can’t learn fast enough to benefit from the new kinds of stimuli.
Q. Making associations on one hand seems like making use of existing pathways of associations and activating these certain pathways, in other ways, it seems like constructing. Do you see these as two independent functions?
A. Activation of pathways vs making new associations: These are two different but obviously related functions. Making new associations involves the following kinds of scenarios: I realize that appearance of A is always followed by B. Or I realize that the appearance of A in this context is always followed by B. It is a formation of a new/stronger associative (Hebbian) connection from a connection that didn’t exist. The use of this connection refers to the situation when one thing is activated and consequently the most related things are also activated. One mediates the other, so they are related, but they are not the same thing. When we make associations, we’re primarily activating existing associations in memory. When you think about thinking process, it’s much like a conversation, where you can think of a certain thing for a second, and after some time, we jump to something else, and again to something else…I think that every one of us can think of an example where we are first thinking about our pants and end up thinking about the moon. It happens all the time.
Q. What are the natures of the things that are associated with one another? Are associations between what pictures, concepts, symbols?
A. That is a wonderful and complex question that I don’t think will be solved any time soon. We have some working definitions and working assumptions about what we’re activating when we’re making associations. If associations being made were between representations of this object and this object, it’s visual, so it relates to Carolyn’s question before about why it isn’t in the parietal cortex because if it were directing between this visual image and this visual image, the whole thing would take place in the visual cortex. And if it were between this sound and this sound, it would be in the auditory cortex. If it were between touch, it would be in the somatosensory cortex. This is a very interesting question that we’re thinking about almost every day, but I can’t say that experiments can be done to directly reveal the content of this.
Q. Are they black boxes for you in this research?
A. I wouldn’t say black boxes. More of “put it aside for now until we solve something else”.
Q. Coming back to summarizing your thesis: We have associative structures, which is past experience. Once the input comes, we have the present. The interaction between the present and the past will predict the future, right? There are many association structures that exists, and once there is an input, we immediately jump to the future. Thus, we’re actually always living in the future because of the interactions of the present and the past. Does this accurately summarize your thesis? One cannot avoid living in the future because one has such a rich associative structure. And once an input comes, we immediately have some anticipation for the future.
A. The only reservation I have in saying that we’re always living in the future is that (even though I love this idea), sometimes we can fantasize about seeing ourselves about the past.
R. The past is the mirror image of the future.
A. You have quite accurately summarized my thesis, but I would qualify it that it’s not necessary that these predictions are triggered by external inputs. They can be generated by internal processes—desires, thoughts, memories—can activate and generate predictions.
R. These are inside the body, but still external to the mental processes.
A. If I am thinking, while I drive, about writing an e-mail when I get home, this process is completely internal. Nothing in my immediate environment affected my prediction of the episode of writing the e-mail.
Q. Do you know what happens to the brain during sleep compared to what happens in your default state?
A. That is a difficult question. People are studying it now, and I don’t think there is an answer yet. We know that people do something that is related to memory during sleep, but there are no final answers regarding the extent it relates to predictions.
Q. I read that depressed patients and Alzheimer’s patients are less able to make predictions.
A. It has been suggested that depressed people are less able to make predictions, less able to activate associations. This brings in the disruptive, subversive part of thinking, e.g. rumination/dwelling on negative thoughts without being able to snap out of it. Also, for certain we know that anxiety brings about depression, and I think anxiety is in direct relation to the inability to create predictions. Many researchers are working directly on the relation between inability to create predictions, increasing uncertainty, and increasing anxiety and depression.
Q. You also mentioned a study that claimed “use it [brain function] and lose it”. I was very intrigued by that.
A. Yes, I was referring to another study that discovered that the Alzheimer network is the first cortical areas being atrophied by Alzheimer’s. So in a way, we expect the Alzheimer’s patients who have less associative thinking patterns and less predictions (from the little evidence I am familiar with) show less priming.
Q. How could associations/generations of predictions be related to consciousness?
A. You’re asking to speculate much further than my comfort zone. Consciousness and prediction/association is certainly connected in my mind. When I am fixating completely on a particular object, allocating all of my mental resources on this object, I may be no longer conscious, or less conscious, about other things in the periphery. And perhaps, if I am focused on a single object, and I lose this single point of attention, the result is that I fall asleep. Or in another way, the moment we stop allocating our attention, we lose attention. My intuition is that association/generation of prediction is tightly related to consciousness, but I am not sure exactly how as this is a big question.
Q. Nonspecific anticipation: I thought you said that it happens in the amygdala that is separate from specific anticipation. Is this correct?
A. Lymbic system is part of the reptilian brian that causes general alertness/arousal of the organism upon activation, and so has nothing to do with prediction. It does [have to do with prediction] in the sense that if you hear the lion roar again, you have both the higher arousal of wanting to run away, but you may also have explicit conscious predictions of what is happening in the environment. So these two systems, I would say, are non-overlapping, yet never-the-less highly interacting. Actually, it’s not even true that they are non-overlapping, because the prefrontal cortex, it seems, in both networks activates its prediction regions (though it’s too strong to say that they are prediction regions).
Q. We had a speaker in the past who mentioned that a hungry cat is aroused such that it attends to more highly to food stimuli. Would that be specific or nonspecific anticipation?
A. This reminds me of how I noticed once that people entering the cafeteria look at our food, while people exiting the cafeteria tend to look at our faces. When we are hungry, we don’t think about mating. In these ways, our current states/goals/needs shape our perception, in the same way a hungry shark seeing a person on a surfboard sees a seal (instead of a human). Our predictions shape our perceptions to such a large extent, that we will not notice any changes unless they violate our predictions to such a large extent. Our perceptions rely on our predictions to a huge extent, and the perception is counted delayed, counted a few milliseconds afterwards.
Q. How do you view consciousness (generally)? And how do you think it is related to your work? Specifically, I am interested in knowing what you think about a theory claiming that our perception produces biases that we are unable to change on the spot (e.g. our consciousness cannot control us to “not be racist”). Do you think this is true? Or do you think that consciousness can dictate how/what associations can be made?
A. We do know that people very quickly make associations, and these associations can last for a very long time. It is true that if someone appears very threatening to you in the first impression, it is very difficult to eliminate that association (even if that person turns out to be extremely nice). It’s likely that consciousness can interfere in a very limited capacity, only with things that do not threaten one’s survival. If it concerns survival, however, the subconscious takes over (which is not necessarily a bad thing).
Q. Is consciousness only an emergent feature that doesn’t cause anything? Or can consciousness tell our brain what associations to make?
A. I think both. I think that the subconscious is more stubborn about things to do with survival, but top-down causality by consciousness is also possible when decisions do not concern survival.
Reading: http://www.neuphi.com/images/readings/Bar-TICS_2007.pdf