observation

From quantum mechanics we know that observing something changes its position or momentum, and so alters the state the object is in, preventing us from knowing what was true before the observation, or what would be true without the observation having happened.

Similarly (or I want to argue for a similarity), to inquire into the emotions or consciousness of a subject, you have to ask them to report on it, thus altering their conscious state. Here too, we don't know what would have been true without the observation having happened. What McClelland was discussing in the CNBC retreat discussion on consciousness last year about Crick and Koch's work suggested the same thing: that there is some tricky and too intimate connection between conscious thought and reporting on conscious thought. That work seemed to suggest that the act of reporting on it might be involved in bringing it to consciousness. I'd like to draw out this similarity.

It might also relate to my paper on feminist standpoint epistemology, but on a more abstract level. there, one of the points was that the person doing the observing puts something into the observations from themself necessarily.

dream-wake

another component of this might have something to do with the difference between being asleep and being awake. Now (p. 136) they're talking about how the brain is active during sleep in similar patterns to when awake (during REM sleep anyway) but that it functions without outside inputs, just with inputs from itself. one thing that could be a problem in terms of objectivity would be if something like consistency with observations was a criterion, then you'd have to make sure somehow that the observations were coming from outside and not from yourself. but perhaps there is some other way of telling whether you're alsleep or not. I always think back to Hinton's Sleep-Wake algorithm, and how it worked by doing backpropagation from outside-inouts durint the Wake phase, and then did backpropagation from its own projections during the Sleep phase. And there are reasons why people have to sleep and dream to maintain their sanity. Whatever these reasons are, there must be some process that we undergo during dreaming, that may or may not have something to do with Hinton's algorithm, and so it might be possible to tell the difference between brains that have slept too little, brains that have slept enough, and brains that have not slept enough. So maybe whatever this criterion would be for distinguishing them could also be some kind of criterion of objectivity. too little dreaming would likely mean not being internally consistent. too much dreaming would likely mean not being connected with reality. just a thought.

Consciousness, cont.

now on p. 127 there's something more to do with objectivity...
they're talking about how to apply information theory to brains, and how this presents 2 problems: that there's no "neural code", and that there's no external intelligent observer, or at least not within the system itself. so they decide to talk about differences that make a difference to the brain itself.
- A simple approach is to consider the system as its own "observer."

I had been thinking that a possible comp paper topic might be something like "Practical Notions of Objectivity" and maybe this is all coming together a little bit more. I'd like to go further into some of the things I was saying in my paper for Laura's class, both the bits about the types of objectivity, and about strategies for getting objective results. maybe also the bits about how this applies to AI, but mostly the more theoretical part, about objectivity in evaluations of intelligence. This could be made a bit more general, and so apply more widely than just to AI, so that it could also have to do with neuroscience / philosophy of mind. an example there would be these confusing questions about consciousness and how objectivity / subjectivity is also a blocking point there. I'm still confused, obviously, but maybe this is the beginning of some kind of direction.

I'm still undecided whether it woulld be better to take the Cognitive Neuroscience course I'm signed up for, or the Metaphysics and Epistemology Core course in the Philosophy Dept. I think both are important things to have as foundations, but I don't know what should come first. I'm leaning towards M & E, because I feel less sure of myself in that area, and for the philosophy comp, being good at that angle might be more important. So I should maybe do the CNBC courses a little later, or only where I can fit them in without fucking up anything else. The main consideration against that is that I feel guilty about hot being as involved in the CNBC as perhaps I should be. I guess it doesn't matter that much, as long as I take all the courses eventually. The only possible one for next semester is the computational course, but I wanted to take that after doing the rest of them so that I'd know enough to make an informed choice about what sort of project to do for that course, since I have this idea of doing a good project that might be publishable. So I guess I could just not do a CNBC course this year.

anyway, I think I'm going to try to use this blog as a place to take some notes and record some thoughts about things that might end up in my comp paper.

Consciousness: How Matter Becomes Imagination

by Gerald M. Edelman and Giulio Tononi
Penguin Books: 2000

some notes on this book. The chapter i'm in is about the unity of conscious experience. for one thing, i'm a bit sceptical about how they can claim that one of the primary characteristics of consciousness is that it's unified and can contain only one thought or decision, while at the same time talk about things like split brain patients, who can be conscious of 2 things at once, or make 2 decisions at the same time. maybe each of those consciousnesses is unified, but it's not really clear whether it's impossible for others to be able to do similar things. maybe it just takes practice. not sure if this matters.

second, from the cases where after strokes people's consciousness re-unifies itself and the person becomes unaware that anything is missing, this makes it very clear how we could have gaping holes in our normal perceptions of reality and not even be able to conceive of the fact that this is so, so strong is the impression that our field of view is complete. ... consequences for epistemology and objectivity...

conclusions

the Rethinking Innateness and Developmental Cognitive Neuroscience notes ended up in the powerpoint presentation, and I don't feel like reproducing them here. not much point, as far as I can see.

I was just thinking of some things that I should say in the wrap-up report.

I need to emphasize more what the conclusions are from the research. something about how there are modules in the brain, clearly, but that looking for ones that are simultaneously innate and domain-specific isn't going to turn up much. the innate things in the brain seem to be lower level things that bubble up somehow to make more complex things.

I think what's wrong with the cobbled together collection of mental modules view is that it isn't a complex system. It's not hierarchical. It's just a collection of simple machines that together would look like something smart, but really would be more like a computer. I think what's interesting about the brain is that it's not like that. It's that there are complex relationships between the parts, and there's an unintuitive and mysterious recipe for how it all gets put together.

so maybe that's why the complexity part seemed to get lost a little in this project. the models I was dealing with aren't complex models, even though what they're trying to model is.

i think i shoudl also say something about how i'm not sure this was a good choice of project for me for this class. it's a topic I wanted to know more about, but in looking for what mI wanted to learn, I kept getting pulled away, either from the question mI was supposed to be answering, or from the themes of the course. Maybe the course and the topic could have worked better together, but it seemed like it wouldn't be in ways that were the sort of thing i like to do. not sure i can put my finger any more accurately on what i mean than that.

last minute

I actually have done a bunch of work since the last post. While I was in Boston, I read several chapters of a textbook called Developmental Cognitive Neuroscience, by Mark Johnson that is basically a first textbook in a newly defined field. This field seems to be what I was searching for, only I didn't realize that the search would turn up anything as coherent as that. I'd found a reference to the field of Developmental Cognitive NEuroscience in the Quartz paper, and went looking for a book called Brain Development and Cognition, A Reader in the library. Next to it on the shelf was the Johnson book, and they turned out to be companion volumes, one the textbook, the other a collection of seminal papers on topics discussed in the textbook.

I read the DCN text while I was in Boston, instead of attending the conference. The only session I went to was the one I presented at. I didn't take any notes, because I was reading in crowded hotel lobbies and cafes, but I folded over the corners of some pages, so hopefully it will all come flooding back to me when I open up the book again.

I assigned readings for the talk out of another book, Rethinking Innateness, that I'd been meaning to get into for a while. Figured assigning reading from it would force me to read some. Luckily the short reading turned out to be on what I thought it would be, and does seem like a relevant sort of introduction. Now I'm going to read the longer chapter... this is going to be a looooong day. After that, I'll try to remember what mI read from the DCN text, then if there's time, I'll read one of the papers from the BDC reader. that's the plan.

new bits from tomorrow's presentation handout

Evolutionary Psychology and its critics:
Samuels: Massive modular minds: evolutionary psychology and cognitive architecture

EP says modules are domain-specific and innate.
Massive Modularity Hypothesis: like Fodor’s modules, except everything is modular, even the executive control mechanisms that he said were not modular.
unclear if informational-encapsulation is required. maybe taken as part of domain-specificity? unclear whether EP modules must be implemented by physiologically discrete neural structures, but how could they be domain-specific otherwise? can global characteristics or distributed structures implement modules?

do autistic kids demonstrate that there is a Theory of Mind module, or are they missing domain knowledge about other people’s minds?

evidence against innate domain-specific microcircuitry as a viable account of cortical development. do cortical plasticity and transplant studies in developmental biology undermine the idea of innate, domain-specific modules? (Quartz, Elman et al.)

Samuels says this criticism depends on a false assumption, Principle of Invariance: the innately specified properties of a piece of cortical tissue are invariant under alterations of location in the brain and the afferent inputs to it

what does innate mean?
Tissue Nativism: specific tissues have innate computational properties FALSE
Organism Nativism: it’s innately specified that organisms possess certain domain-specific cognitive structures TRUE?
But who would seriously propose tissue nativism?
and how would/could organism nativism happen? neural nets?


Additions to bibliography

Churchland, Sejnowski (1992) The Computational Brain. Cambridge, MA: MIT Press

Elman et al. (1996) Rethinking Innateness: A Connectionist Perspective on Development. Cambridge, MA: MIT Press

Gazzaniga (1989) “Organization of the Human Brain” Science. Vol. 245, No. 4921, pp.947-952

Segal (1996) “The Modularity of Theory of Mind” in Carruthers and Smith eds. Theories of Theories of Mind. pp.141-157. Cambridge: Cambridge University Press

more on the quartz paper

Section 3:
- present a behavioral systems model that regards the brain as a hierarchical control structure, where this hierarchical
organization is evident both developmentally and evolutionarily. This behavioral systems model places a premium on the complex interaction between developmental mechanisms and a structured environment
- highly conserved nervous system developmental mechanisms suggests that nervous systems, despite their apparent diversity, share a deep structure, or common design principles, just as the fact that two million distinct species share only 35 major body plans suggest that body plans share many common design principles
- At virtually all levels of the human nervous system, for example, reward systems can be found that play a central role in
goal-directed behavior (Schultz, 2000)
- Rather than simply carrying information regarding reward, it appears that octopamine signals information regarding prediction errors. (bumblebee)
- A system that learns through prediction learning need not have the path from goal to reward specified, in contrast to fixed behavioral patterns, such as stimulus-response learning. Instead, the path from goals to rewards may be left open and discoverable via learning, resulting in flexible action.
- in monkeys, Schultz also found dopamine reward signals that code prediction errors

midbrain dopamine system (old) projects to (new) dorsolateral prefrontal, premotor, and parietal cortex, ... structures implicated in crucial components of human cognition, particularly social cognition and theory of mind (Stone et al., 1998), symbolic learning (Deacon, 1997), representations of self (Craik et al., 1999), and executive function and behavioral inhibition ... these structures constitute a hierarchically organized control structure, where additional layers of control have been added to the evolutionarily conserved dopamine system and where this hierarchical organization is evident developmentally as well.

dopamine connected to all kinds of abilities involving the frontal cortex, and into adolescence, behaviour becomes increasingly mediated by frontal regions... the brain keeps developing until quite late.

Although there is strong evidence that an intact dopamine system is necessary for the developmental emergence of prefrontal functions, a largely unresolved question concerns the specific nature of this developmental link. One particularly intriguing possibility is that the dopamine signal serves as a learning signal that guides the construction of prefrontal structures during development.

The developmental link between the midbrain dopamine system and prefrontal structures suggests that an explicit account of the developmental trajectory of cognitive skills is necessary; an account based on innately-specified modules is inadequate.

neural constructivism whereby the functional properties of cortex are built from the dynamic interaction between neural growth mechanisms and environmentally-derived neural activity, acting hierarchically at the regional level and with high specificity at the cellular level (Quartz & Sejnowski, 1997; Quartz, 1999)

neural development during the acquisition of major cognitive skills is best characterized as a progressive construction of neural structures, in which environmentally-derived activity plays a role in the construction of neural circuits.

There is now good evidence to indicate that one component of social competence, theory of mind, depends at least in part on the appropriate social exposure for its development, as many deaf children show delays on theory of mind tasks

LOP = eeeek

I picked up a couple of the books that the Samuels paper pointed me to, but still haven't grabbed the Elman paper. The books are in various locations around campus, hidden in buildings I've never heard of, some of them at the top of annoyingly steep hills. and then the day i was doing all this, it started raining, so i fled home before making it to all the libraries. i have no good excuse for why i still haven't made it there. nor why i still haven't transferred my notes on the samuels paper in here.

nevertheless, i've got a book on brain development and connectionism, and another one on autism.

I've also finally looked at one of the papers from the Modularity book Sandy reviewed.
this paper critiques domain-specificity. I think the developmental cognitive neuroscience people are more into critiquing innateness, whatever that means, but maybe also informational encapsulation. all of these things seem to be quite related anyway...

Velichkovsky
Modularity and Cognitive Organization: Why It Is So Appealing and Why It Is So Wrong
the title seems somewhat misleading, but i just skimmed most of the article, so maybe i missed some parts. seemed like it was mostly about a few kinds of levels of processing (LOP) experiments and a bit about brain imaging of subjects doing the same sorts of tasks. When I took a class with Fergus Craik in undergrad, I thought the LOP idea was wishy-washy to the point of saying almost nothing useful, became indignant, and stopped going to class. we also had to read a bunch of ancient crap about working memory. maybe it was just that i was in 4th year and taking a 2nd year course, so it was a bit too dumbed down. i think i also participated in a study exactly like the ones discussed in the paper. Craik's voice was on the tape they played to me while i did the distracting task, listing words, but his strong Scottish accent made it difficult to tell what he was saying. The student running the experiment knew exactly when he was impossible to understand and automatically repeated some of the words without the accent. One of the formative experiences leading me to believe that the results of psychology experiments should be considered as highly dubious. but i digress...

LOP effects on saccades:
with visual, auditory, or "relevant" stimuli, the time lag before the next saccade is the same? or maybe different? i didn't read it that closely. oops. does this mean that there are cross-modal effects? i wasn't sure.

LOP effects on word memory:
with word memory tasks given after the encoding happened at different levels, differences in how much was remembered. deeper encoding means better memory, as far as i could tell/remember. not sure what the significance of this is supposed to be for domain-specificity either.

EEG results for the same sort of word memory tasks:
the deeper the level of processing (perceptual encoding, semantic emcoding, self-referential encoding), activation was seen farther towards the anterior regions of the brain.
this one does seem kind of relevant to modularity. recalling words seems like a specific task, but it doesn't happen in a well-defined region. several types of processes affect it, from perceptual to self-referential, whatever that means. they say something in the results about "neuroanatomical change". I'm not sure what they mean by that. but they think it's significant that the direction it happens in is in the main direction of evolutionary growth of the cortex. they had a bit near the beginning about what the older vs. newer regions of the brain are responsible for. laid out a hierarchy of levels of processing, basically that correspond to the brain regions and when they emerged in our evolution.

I still don't think it's all that clear what these results say about modularity or domain-specificity. memory isn't the sort of thing that most people call a module, for just these sorts of reasons. But maybe I missed the point because i only skimmed most of it. but i can't bring myself to carefully read a paper about LOP.

misc. notes

from tonight's e-mail to Sandy:
"quick update on my progress: I've found the secret cache of papers that talk about exactly what I'm looking for. I haven't grabbed them all from the library yet, but have been reading the ones available on-line. So it looks like a few people have already done exactly the sort of critique of modularity using developmental evidence that I was thinking of doing. On the one hand, it's slightly annoying to have been scooped once again. On the other hand, it must be a good idea if other people are doing it. On the third hand, maybe they didn't do a good job of it, and I'll be able to find some ways of saying something new. It seems like it's a fairly small set of papers that are on this topic, so it's likely that there is more to be said."

the Elman paper mentioned below seems to be the main thing I need to find. The Samuels paper that led me there was an assigned reading for Edouard's class. The bibliography wasn't included in what we got, but I've asked him to bring the book in tomorrow, so that I can follow up on the references. then I'll head to the library and see of I can find them. I already looked around on-line for a bunch of them, but without the paper names, it was rough going, although i found some different papers that might be useful. Stupidly, I didn't grab their bibliographic info from JStor, so I'll have to type it all up later.

I've got some notes written up from the Samuels paper in my notebook that i'll transfer here later. now some notes on the papers i've been looking through...

P.S. Churchland & T.J. Sejnowski paper (the wrong one):
Neural Representation and Neural Computation
Philosophical Perspectives, Vol. 4, Action Theory and Philosophy of Mind (1990), 343-382
about the metaphor of mind-brain as computer, comparing symbolist and connectionist models.
thought as a language-like or logic-like procedures = beliefs and desires as propositions = folk psychology
for many well-known reasons (to anyone versed in cog sci), this view has major problems
the paper is about mental representation, so isn't really relevant. i think they have another one that's more about modules.


S.R. Quartz paper:
Toward a Developmental...
the main point: why evolutionary psychology should pay attention to development.
- human cognitive architecture as a hierarchically organized control structure, where this hierarchical organization is evident both evolutionarily and developmentally
Hirschfeld & Gelman, 1994 -> another modularist target
developmental cognitive neuroscience (Elman et al., 1996; Johnson, 1997; Quartz, 1999)
- human development is both more protracted and more sensitive to environmental signals than nativist
cognitive psychology supposed
- discovery of homeobox genes and their striking conservation (reviewed in Hirth & Reichert, 1999; Reichert
& Simeone, 1999). Given the enormous differences in neuroanatomy between vertebrates and
invertebrates, their brains were long thought to be unrelated with little obvious homology. However, at a
deeper, molecular level they are remarkably similar in that homologous regulatory genes have been
identified that control regionalization, patterning, and identity in embryonic brain development.
Raff 2000 for review of evo-devo
- although cortex size (relative and absolute) varies widely across mammals, its organization into 6 horizontal levels is the same across species. differences in behaviour and cognition relate to underlying differences in the interconnectedness between layers.
- for structural brain modules to develop, neurogenesis must be dissociable between different structures -> prediction that different species might have brain structures of different relative sizes. results say that over 131 species studied, brain structure sized are highly correlated, with the exception of the olfactory bulb. -> highly conserved homeotic starting point for all mammal brains. the order of neurogenesis is also highly conserved across species.
...
more later