Wednesday, December 22, 2010

Michael Shermer - Stephen Hawking’s Radical Philosophy of Science

I did not know that Stephen Hawking had become a relativist in his perspective on science - or as Shermer calls it, belief-dependent realism: "None of us can ever be completely sure that the world really is as it appears, or if our minds have unconsciously imposed a misleading pattern on the data." Shermer refers to studies (and there are a lot of them) that show are beliefs, assumptions, biases shade the way we view the world, speak about it, and relate to it.

Hawking's new perspective is similar:
Hawking presents a philosophy of science he calls “model-dependent realism,” which is based on the assumption that our brains form models of the world from sensory input, that we use the model most successful at explaining events and assume that the models match reality (even if they do not), and that when more than one model makes accurate predictions “we are free to use whichever model is most convenient.”
Here is the whole article - very interesting reading from my perspective.

From Big Questions Online:

Stephen Hawking’s Radical Philosophy of Science

Is Hawking right to claim that reality is dependent on the model used to describe it?

Stephen Hawking in zero-G
photo: NASA
Tuesday, November 23, 2010

Do you think that there is a computer screen sitting in front of you right now?

It would certainly seem so if you are reading these words online, but in fact you are not actually “seeing” the computer screen in front of you. What you see are photons of light bouncing off the screen (and generated by the internal electronics of the screen itself), which pass through the hole in the iris of your eye, through the liquid medium inside your eye, wending their way through the bipolar and ganglion cells to strike the rods and cones at the back of your retina. These photons of light carry just enough energy to bend the molecules inside the rods and cones to change the electrochemical balance inside these cells, causing them to fire, or have what neuroscientists call an “action potential.”

From there the nerve impulse races along the neural pathway from the retina to the back of the brain, leaping from neuron to neuron across tiny gaps called synaptic clefts by means of neurotransmitter substances that flow across those gaps. Finally, they encounter the visual cortex, where other neurons record the signals that have been transduced from those photons of light, and reconstruct the image that is out there in the world.

Out of an incomprehensible number of data signals pouring in from the senses, the brain forms models of faces, tables, cars, trees, and every conceivable known (and even unknown — imagined) object and event. It does this through something called neural binding. A “red circle” would be an example of two neural network inputs (“red” and “circle”) bound into one percept of a red circle. Downstream neural inputs, such as those closer to muscles and sensory organs, converge as they move upstream through convergence zones, which are brain regions that integrate information coming from various neural inputs (eyes, ears, touch, etc.) You end up perceiving a whole object instead of countless fragments of an image. This is why you are seeing an entire computer screen with a meaningful block of text in front of you right now, and not just a jumble of data.

At any given moment there are, in fact, hundreds of percepts streaming into the brain from the various senses. All of them must be bound together for higher brain regions to make sense of it all. Large brain areas such as the cerebral cortex coordinate inputs from smaller brain areas such as the temporal lobes, which themselves collate neural events from still smaller brain modules such as the fusiform gyrus (for facial recognition). This reduction continues all the way down to the single neuron level, where highly selective neurons — sometimes described as “grandmother” neurons — fire only when subjects see someone familiar. Other neurons only fire when an object moves left to right across one’s visual field. Still other neurons only fire when an object moves right to left across the visual field. And so on, up the networks, goes the binding process. Caltech neuroscientists Christof Koch and Gabriel Kreiman, in conjunction with UCLA neurosurgeon Itzhak Fried, for example, have even found a single neuron that fires when the subject is shown a photograph of Bill Clinton (PDF) and no one else!

The models generated by biochemical processes in our brains constitute “reality.” None of us can ever be completely sure that the world really is as it appears, or if our minds have unconsciously imposed a misleading pattern on the data. I call this belief-dependent realism. In my forthcoming book, The Believing Brain, I demonstrate the myriad ways that our beliefs shape, influence, and even control everything we think, do, and say about the world. The power of belief is so strong that we typically form our beliefs first, then construct a rationale for holding those beliefs after the fact. I claim that the only escape from this epistemological trap is science. Flawed as it may be because it is conducted by scientists who have their own set of beliefs determining their reality, science itself has a set of methods to bypass the cognitive biases that so cripple our grasp of the reality that really does exist out there.

According to the University of Cambridge cosmologist Stephen Hawking, however, not even science can pull us out of such belief dependency. In his new book, The Grand Design, co-authored with the Caltech mathematician Leonard Mlodinow, Hawking presents a philosophy of science he calls “model-dependent realism,” which is based on the assumption that our brains form models of the world from sensory input, that we use the model most successful at explaining events and assume that the models match reality (even if they do not), and that when more than one model makes accurate predictions “we are free to use whichever model is most convenient.” Employing this method, Hawking and Mlodinow claim that “it is pointless to ask whether a model is real, only whether it agrees with observation.”

For example, in physics experiments sometimes light acts as a particle and sometimes it acts as a wave. Well, which is it, particle or wave? The answer depends on which model of light you use. In the famous double-slit experiment, light is passed through two slits and forms an interference pattern of waves on the back surface. When you send single photons of light one at a time through one slit, the light acts like individual particles. But when you shoot the single photons of light one at a time through two slits, they form an interference wave pattern as if they were interacting with other photons, even though they are not … at least not in this universe!

How is this possible? One solution to the mystery is that the photons are interacting with photons in other universes. Hawking and Mlodinow employ the model developed by Richard Feynman called “sum over histories,” in which every particle in the double-slit experiment takes every possible path that it can, and thus interacts with itself in its different histories.

So which model of light best matches reality? According to Hawking and Mlodinow, none of them do, or they all do. “There is no picture- or theory-independent concept of reality,” the scientists conclude. “If there are two models that both agree with observation, like the goldfish’s picture and ours, then one cannot say that one is more real than another. One can use whichever model is more convenient in the situation under consideration.”

Model-dependent realism argues that there is no privileged position in the universe — no Archimedean point outside of our brain that we can access to know what reality really is. There are just models. It is not possible to understand reality without having some model of reality, so we are really talking about models, not reality. Is there a way around this apparent epistemological trap?

There is. It’s called science.

The tools and methods of science were designed to test whether or not a particular model or belief about reality matches observations made not just by ourselves but by others as well. When one scientific lab corroborates the findings of another lab, and those findings support of a tested model, then it strengthens our confidence that the model (or hypothesis, or theory) more closely corresponds to reality, even if we can never know with 100 percent certainty the true nature of that reality.

Even when two models appear to be equally supported by observations, over time we accumulate more precise observations that tell us which model more closely matches reality. Historians of science contend that in the 16th century, the newly introduced Copernican sun-centered model of the solar system was, in fact, no better at explaining the observations of the movement of the planets than was the Ptolemaic earth-centered model. As observations of the movement of planets increased in accuracy, the Copernican model won out.

If model-dependent realism were taken to its nth degree, we could never actually say that the Copernican model is better than, or superior to, or more closely matches reality than the Ptolemaic model. Hawking and Mlodinow would surely agree, because they argue that a model is good if it meets four criteria:

  • Is elegant
  • Contains few arbitrary or adjustable elements
  • Agrees with and explains all existing observations
  • Makes detailed predictions about future observations that can disprove or falsify the model if they are not borne out

As a historian of science, I conclude that, in fact, nearly all scientific models — indeed, belief models of all sorts — can be parsed in such a manner and, in time, found to be better or worse than other models. In the long run, we discard some models and keep others based on their validity, reliability, predictability, and perceived match to reality. Yes, even though there is no Archimedean point outside of our brains, I believe there is a real reality, and that we can come close to knowing it through the lens of science — despite the indelible imperfection of our brains, our models, and our theories.

Such is the nature of science, which is what sets it apart from all other knowledge traditions.

Michael Shermer is the publisher of Skeptic magazine, a monthly columnist for Scientific American, and an adjunct professor at Claremont Graduate University. His books include The Science of Good and Evil, Why Darwin Matters, and The Mind of the Market. He can be reached at mshermer@skeptic.com.

No comments: