• View
• Edit
• History
• Print

CognitionAndBiology

Assigned Reading

• Winograd and Flores, Cognition as a Biological Phenomenon (Chapter 4, 38-53) [16]
• Andy Clark, Being There: Putting Brain, body, and World Together Again, Bradford/MIT Press, 1997, Chapter 4, Collective Wisdom, Slime-Mold-Style, 71-82. [12]

Additional Resources and Readings

• A new article by Andy Clark, Re-Inventing Ourselves: The Plasticity of Embodiment,Sensing, and Mind Journal of Medicine and Philosophy, 32:3, 263 - 282

• Andy Clark, Embodiment and Ecological Control, Slides from online talk on Friday April 9, 2004
  • Stelarc, Stelarc, Homepage of third hand creator
  • Rick Weiss, Washington Post, Monkeys Control Robotic Arm With Brain Implants
• Humberto Maturana, Neurophysiology of Cognition [21]
• Humberto Maturana and Bernhard Poerksen, From Being to Doing: The Origins of the Biology of Cognition, Carl-Auer, 2004
• On rules, biology, function, and representation, see Ruth G. Millikan, Truth Rules, Hoverflies, and the Kripke-Wittgenstein Paradox, Philosophical Review, vol. 99, Jul. 1990, pp. 323-353.
• On realism about patterns, in particular mental states as patterns: Daniel C. Dennett, Real Patterns, Journal of Philosophy, vol. 88, 1991, pp. 27-51.

Comments - Before Class

Live and Learn - Jeff Wear

Coupling

First we see that the notion of an objective meaning behind a word or idea is untenable, because a word in someones head is so tightly coupled with all the other words/thoughts/feelings in their head that talking about a word in isolation will only yield part of the picture. In the same way we see that talking about an organism without talking about its medium (environment) is equality naive. And, similarly, a program without its environment. Also, interestingly, as Lakoff argued, a soul without its body. My point is that its fascinating how tightly coupled everything is, and how counterintuitive tight coupling is to everyday thinking. I wonder if its because these naive assumptions were somehow more practical for everyday life.

--Nate

Self-organizing computers?

The phenomenon of self-organization (in non-living, but even more in living nature) is amazing. The cases described in Clark’s work (but also by many other authors, e.g. Fritjof Capra in The web of life) show, that objects in interaction start to organize without an external (or higher) plan, just by following some basic rules, like the water molecules following the basic physical laws while forming the whirlpool, which is a higher-level organization. Prigogine, Maturana, Varela and many others showed that life emerged by self-organization of non-living matter into higher-level organizations (organisms), where uncertainty and far-from-equilibrium situations played a significant role. Self-organization and nondeterminism are also the basis of intelligent behavior and creativity. The question is, whether machines can evolve into anything alike, considering they are all built to behave deterministic. So far I’m not aware of anyone who would build a computer that would “behave” in an unpredictable way. Although intelligence is what many expect of them. Another question, which arose while reading the Winograd and Flores chapter on Cognition as a Biological Phenomenon is, how does the vertical communication and understanding between the levels of organization in an organism work? What “languages” are used here? And, what is the role of »will« in an authopoietic system?

--Danijel

Collective Wisdom Achieved? Mark Schar

Anarchy, Indirect Requests and Creeping Slime Mold, Interesting

The speaker George Lakoff is not only an interesting individual with quite a story and several good case studies to validate framing but those surrounding him are equally interesting. Robin Lakoff is a professor of linguistics and her Wikipedia page lists ten aspects that define woman's speech differences from man's. One is the "indirect request" used by Professor Winograd to present an example of context, the master said to the butler, isn't it cold in here, where the butler replied, why yes, shouldn't your wife reach over an close the window she left open. I specifically refuse to respond to indirect requests at home and will use this to further my argument that they are not part of language but a misuse. At least I am correct for half the population.

Next is the invalidity of inferring meaning from speaking in italics. Moving on, Professor Lakoff's main point of departure, Professor Noam Chomsky, is one of our nations leading anarchist [atheist typo corrected] organizers (is that irony?). Leading me to the point of this pre-discussion. Is self-organization a form of Anarchy? If society was provided the knowledge (enlightenment), communication capacity, and universal knowledge sharing necessary for self-organizing then would we be promoting ideals of Anarchy. And, if so would this be consistent with the ideals of the triad system of checks and balances? Last, is there an application for self-organizing on the construction project, or does this already essentially function like a termite colony, and any input form the management authority only acts to disrupt the system. If software tools for construction focused on facilitating knowledge sharing and improving communication at the line of workers directly behind the shovel rather than the line of managers directly below the executive, more dramatic results may be realized.

-- Forest

Can Agent-Based Simulations Escape the Limits of GOFAI?

In reading Andy Clark’s chapter on collective wisdom and emergent behavior, I was struck by how powerful these effects can be. It seems that in computing we have recently witnessed the rise of collective wisdom, and the effects have been tremendous. Where an individual organization such as the Encyclopedia Brittanica once created the definitive encyclopedia, this paradigm has now been almost completely eclipsed by crowdsourcing encyclopedia authorship to collective wisdom systems such as Wikipedia. Like Clark’s example of a sailing crew’s performance exceeding what a team of any one individual, cloned to perform every role could do, this crowdsourcing takes advantage of the role-specialization that has been seen as such an advantage since Adam Smith’s The Wealth of Nations and perhaps before.

It is clear there is tremendous behavior in collective wisdom and collective action, and, as mentioned above, Wikipedia shows how much better we can do when relying on a large collective author. However, can computer applications capitalize on collective behavior? The area of agent-based simulation comes to mind. A recent PhD? dissertation at Stanford analyzed the use of agent-based simulation to understand egress patterns of crowds from buildings in emergency situations. Agents were configured to exhibit human-like behavior, including panicking, running to the nearest exit, and other actions. However, in a recent conversation I had with the professor who advised the dissertation, we discussed whether we could really ever know if the agents would behave as they were programmed to do in the simulation. We can certainly set up variability in the behavior, but do we ever know that the simulation space fully covers the range of human behaviors that would result in an emergency situation? How would they behave when their nervous systems are perturbed by an earthquake versus a fire versus an explosion? What if a different population was in the building—perhaps a convention for the vision or hearing impaired is going on in the building, and evacuation patterns are completely different than a simulation showed?

The questions posed above are interesting, but I wonder if they are the right ones to ask. It has been demonstrated that agent-based simulations often can accurately simulate emergent behaviors in nature. Andy Clark’s example of a traffic jam simulation showing a traffic jam as a moving entity all its own shows that agents that are significantly simplified representations of human-controlled entities can show behaviors that do arise in the real world. This would appear to make a good case that agent-based simulation can do a reasonable job of harnessing the power of collective wisdom and emergent behavior. However I fear that the technique also falls victim to many of the same problems we have discussed about GOFAI. A highway, for example, is a very constrained environment with clear “rules of the road” that can be translated into computational logic with little difficulty. Simulating emergent behavior on a highway versus the interactions of a sailboat racing team, for example, are not comparable in complexity, and I feel that accurately making any predictions about a sailing team’s performance when faced with a new situation would be extraordinarily difficult to model in an agent-based simulation. At this point, I wonder if the agents would need bodies, nervous systems, and all the other components that Lakoff says we need for human behavior in the world to get any meaningful results.

Understanding what agents can do, however, and whether their collective nature can bring their utility beyond what is available in traditional artificial intelligence and simulation is an interesting question I would like to learn more about.

I also wonder if there are ways to use computer tools to allow “human in the loop” agent based tools across a network that would combine the value of collective wisdom with the structure of a simulation and data collection capabilities of a computer, that could lead to interesting results to answer questions about human and organizational behavior and prescribe new ways of working together.

-- Matt Garr

Dean:

Live and Learn: The Role of Induction In Cognition - Jeff Wear

cognition: the mental action or process of acquiring knowledge and understanding through thought, experience, and the senses. -- New Oxford American Dictionary

I was initially baffled by Maturana's sweeping statement that cognitive activity is common to all life: "Living systems are cognitive systems, and living, as a process, is a process of cognition." ("Neurophysiology of cognition," 1970, p. 8). What basis, I wondered, did Maturana have for extending the property of thought to all life, from protozoa to pine trees? The relation of cognition to autopoiesis was not of much use; so life maintains its unity, and adapts to the world if it is to be viable long-term, but what of it? I don't think that a plant can be said to "understand" the world because it bends to face the sun - at least, not in the common sense of "understanding". But there are two facets of "cognition", as defined above, that helped me to place Maturana's use of "cognition" in a context that was relevant to me.

The first was that I needed not be so parsimonious (and perhaps traditional) in my use of "understanding". Rather than be an animal solipsist of sorts, hung up on the question of whether or not a vole ruminates on rooting, I can instead say as an observer of the vole's behavior that it does act as if it does have some sort of understanding of the world in which it operates - its actions are appropriate to the world in which it lives. Because the vole is structurally coupled to its medium, it behaves in ways relevant to its "goals" - survival, reproduction, what more generally is termed autopoiesis.

But the crucial realization for me in regards to "cognitive activity" was that cognition, even when used in the traditional sense, can refer to the process of acquiring an understanding such as is described above through experience. In this way, cognition need not refer to any conscious, thought-ful process of learning. Yet I don't think that Maturana means to describe cognition as mere conditioning and reflex. Rather, he describes the process of cognition as inductive behavior - a characterization which captures the historical nature of learning, not only in animals, but in humans as well. This distinction is highly relevant to me because we have just finished studying induction and its justification in my philosophy class.

Induction - ampliative reasoning from experience - grounds human reasoning in all respects. Even the vaunted scientific-deductive method is a non-demonstrative form of inference (as "confirming" observations do not entail the hypothesis which suggests the line of inquiry). But, of all the efforts to justify induction which we studied in my class, none succeeded. There is no rational justification for why we should so highly esteem our experience as a means of inference, but we do, on the historical assumption that nature is uniform and will continue to be so. The uneasiness, yet inescapable acceptance, I felt concerning this revelation paralleled my apprehension of Maturana's claims regarding the historicity and nature of our structurally coupled behavior.

A theory of mind - Chris Anderson

Chapter four presents nothing less than a theory of the origin and function of cognition – one that has deep philosophical roots and implications. I spent a while trying to place this theory in the greater scope of all recent theories of the mind, but had some difficulty finding analogues. Part of this is the pervasiveness of the rationalistic tradition in established philosophy. Scanning through the textbook I used in Philosophy 80, “Philosophy of Mind: a guide and anthology”, the majority of the arguments were about whether materialism is true and whether we could construct functional descriptions of the brain (i.e. interpret the brain as software). Almost all of these assumed that the brain operates on symbols with some form of explicit rules.

One section, though, does bear a striking resemblance to our theory of cognition. The editor calls it “interpretationism”. Authors like W.V. Quine and Donald Davidson are categorized as having this view, which holds roughly that the meaning of language depends on being embedded within it, and that in order to ascribe thought to an “actor”, it must have language. This sounds as if it mostly agrees with Maturana, Winograd, and Flores, but it differs in a few essential ways. First, Quine and Davidson are extremely interested in explaining the rationalistic tradition through their theories. They use phrases like “(t) ‘Es regnet’ is true-in-German when spoken by x at time t if and only if it is raining near x at t” and elaborate on how their theories can decide on their truth conditions. Second, not all philosophers in this category are comfortable with the idea that language in the consensual domain is prior to thought.

Shedding the notion that there is an objective reality that must be obeyed in philosophical discussions is difficult. Arguments in the later section “Eliminativism” concentrate on the self-defeating nature of any theory that denies the proposition “x believes that p”. They argue that in order to have a theory, you must assume certain things, such as that truth-conditions still exist. What if we leave behind the concept of truth-conditions all together? Then, as Winograd and Flores write: “reality is not objective”.

Terry:

Dean:

Does Maturana's Structural Coupling and Cognition account for Deliberation and Rationality? - Luke Dahl

As I understand Maturana's terminologies, we can say that a living system's behavior (on a short time scale) is determined by its structure. For living beings who possess a nervous system part of that structure is the "state" of the nervous system at the time of the situation we are considering. I suppose by state I mean not a particular emotional or attentional state, but rather the current set of reactions and behaviors that the nervous system tends to favor. For an artificial neural network this would be equivalent to the weights. We think of the nervous system as ahistorical, just as it is at this moment. Since behavior is determined by this structure, we have a kind of mechanistic determinism.

Maturana describes a second domain of description where the structure of the organism can change over time, and this he calls cognition. This is the domain of learning and habituation for a particular individual, or genetic adaptation for a species. We are familiar with the mechanism of mutation and selection for effecting adaptation over many generations. When we come to individual learning, or modification of the "state" of the nervous system, I am led to a question:

What is the proper time scale for distinguishing between structurally determined behavior, and cognition? For example, if I am learning a new skill, my reaction to a stimulus might be different than it was a minute prior before I learned some particular distinction or nuance. This seems appropriate for "real-time" physical skills such as bike-riding, or soldering.

But what about deliberation and rationality, when a significant amount of time passes between the stimulus and the behavior? Can we say that a decision that has been deliberated and considered is structurally determined? As I investigate the possible actions and try to predict their outcomes, do I not sometimes learn to make new distinctions, and therefore change my structure?

One guess about how Maturana might answer this is that in the act of deliberation we become observers making distinctions which we describe to an observer, in this case ourselves. Indeed, one might claim that all deliberation takes place through a linguistic conversation with ourselves in our head, and so Maturana's domain of observation is the proper level of descritpion.

Terry:

Dean:

Fascinating examples by Clark -Vala

- My post below is not very insightful and very rushed so its probably better to ignore it :)

Just quickly, I wanted to comment on Clark's writing. Clark's examples about the slime, termite, traffic jam and beer can grabbed my interest and I enjoyed his fluffier writing. It acts as quite a good counter to some of the other readings that are more dense such as Dreyfus. His writing also brought the Cognition and Biology topic into focus for me.

The big difference that I could deduce between direct and indirect emergence is the extent to which environment is involved. Similarly to many other subjects that we have delved into in this class, there is a the clear-cut and a grayer entity. Direct emergence is clear cut where two or more entities meet and a byproduct emerges (hydrogen and oxygen producing water is a good example) but indirect emergence is grayer, depending on the environment and evolution. My question here is how much environmental influence is needed for an emergence to go from direct to indirect? This is almost impossible to answer so the borderline between the two is also a gray-area.

I would prefer the distinction to be made by saying that indirect emergence is “self-organizing”. This is the notion that Clark puts forward but I like how Danijel formalizes it by using the “self-organizing” phrase.

After class comment - Jeff Wear

In class, we talked a little about how a frog's perceptual system is more "brittle" than a human's, in that a frog can be "fooled" into reacting to a black moving dot as food, regardless of whether it is a fly or not; whereas presumably humans are somewhat better at examining our perceptions and rooting out errors therein. Of course, there are limits to human perceptual systems as well - contradictions or "illusions" we cannot resolve through our judgment alone (founded as it is upon the "erroneous" perception), such as the arrows of seemingly different length Prof. Winograd drew on the board. And in considering our limits, as compared to the frog's, I came to realize the possible inappropriateness of such terms as I have put in quotes - "fooled", "illusion", "erroneous" - to describe the content of our perceptions. Can we say that the frog was "tricked" into believing food was approaching? Surely not; it has no "knowledge" of a fly; on the whole, its perceptual system was not in error, but rather extracted relevant features from the frog's surroundings, exactly as it should have. Likewise, we are structurally coupled to our medium, and thus our perceptual systems also; how can we ever say that our perceptions of the medium were not appropriate to the medium? In other words, perhaps we never see the "truth" or "illusions" - shadows of an objective reality - but rather the world, always exactly as it is to us.

After class comment - Forest

I enjoyed the concept of unorganized organizations presented by Clark. Some of the class discussion focused on how the termites are truly not following a plan and eventually are able to construct arches. It is true that construction, even non-building construction, commonly called heavy civil, where there are no architects, does not fit this model. During discussion Jeff introduced the example made by Clark of a ship and the crew. In the same way construction crews interact. Even the ship has a charted course determined by some meta-organization that the crew is fulfilling. This is the example I proposed with construction. The workers too have a meta-organization that has planned a course, but at the individual level at the day to day basis, like the ships crew the construction workers are likely unaware of this higher goal.