Comments on: Levels of analysis http://neurodudes.com/2007/10/15/levels-of-analysis/ at the intersection of neuroscience and AI. Thu, 20 Nov 2008 15:38:09 +0000 http://wordpress.org/?v=2.6 By: Eric Thomson http://neurodudes.com/2007/10/15/levels-of-analysis/#comment-588919 Eric Thomson Mon, 03 Dec 2007 21:02:08 +0000 http://neurodudes.com/2007/10/15/levels-of-analysis/#comment-588919 You can understand the computer at the software level just fine without knowing anything about how it works. The anti-reductionists think that the mind is the software level description of the brain. Presumably we could run the same software on vacuum tubes, integrated circuits, or using a complicated arrangement of used chewing gum. To focus on such detail, if you are interested in the program being run, would often be a mistake. They think the same applies in the explanation of human behavior. My problem with much of the antireductionist stuff is not their identification of software and mindware, but their claim that we can come to understand the brain at the software level without studying the hardware. The problem is that behavior underconstrains the set of possible computational mechanisms that underlie the behavior. Why should behavior be the one area of biology where the functional decomposition isn't aided by an understanding of the meat doing the functioning? Imagine people studying digestion saying that they didn't need to study enzymes, physiology, etc but that they could just study the inputs and outputs of the digestive system. This would be insane. Why should it be any different in the study of behavior? A guy in the 1800s actually did generate a complicated theory of digestion based on comparing the inputs and outputs of the digestive system (how pleasant). He made incorrect assumptions about the types of chemical reactions possible in the digestive system and his whole system was just wrong (I got this from a <a href="http://mechanism.ucsd.edu/~bill/research/bechtel_characterizingoperations.webversion.pdf" rel="nofollow">wonderful paper</a> by Bill Bechtel, a philosopher at UCSD). The upshot: when you ignore mechanisms, you are likely to end up with a load of shit. You can understand the computer at the software level just fine without knowing anything about how it works. The anti-reductionists think that the mind is the software level description of the brain. Presumably we could run the same software on vacuum tubes, integrated circuits, or using a complicated arrangement of used chewing gum. To focus on such detail, if you are interested in the program being run, would often be a mistake. They think the same applies in the explanation of human behavior.

My problem with much of the antireductionist stuff is not their identification of software and mindware, but their claim that we can come to understand the brain at the software level without studying the hardware.

The problem is that behavior underconstrains the set of possible computational mechanisms that underlie the behavior. Why should behavior be the one area of biology where the functional decomposition isn’t aided by an understanding of the meat doing the functioning? Imagine people studying digestion saying that they didn’t need to study enzymes, physiology, etc but that they could just study the inputs and outputs of the digestive system. This would be insane. Why should it be any different in the study of behavior?

A guy in the 1800s actually did generate a complicated theory of digestion based on comparing the inputs and outputs of the digestive system (how pleasant). He made incorrect assumptions about the types of chemical reactions possible in the digestive system and his whole system was just wrong (I got this from a wonderful paper by Bill Bechtel, a philosopher at UCSD).

The upshot: when you ignore mechanisms, you are likely to end up with a load of shit.

]]> By: Brainy http://neurodudes.com/2007/10/15/levels-of-analysis/#comment-546498 Brainy Fri, 09 Nov 2007 07:28:25 +0000 http://neurodudes.com/2007/10/15/levels-of-analysis/#comment-546498 I agree with you. If you don´t know the neural mechanisms underlying the mental processes you can´t sure about that phenomenon is true. Pinker is near conductism in these answer. For me his vision about the research of the mind is the reason for his opinion about the imposibility (now, and in the future) to know the mechanisms of counsciousness. Great web. I´m spanish. Sorry for the english language. I agree with you. If you don´t know the neural mechanisms underlying the mental processes you can´t sure about that phenomenon is true. Pinker is near conductism in these answer. For me his vision about the research of the mind is the reason for his opinion about the imposibility (now, and in the future) to know the mechanisms of counsciousness.

Great web. I´m spanish. Sorry for the english language.

]]> By: Ed http://neurodudes.com/2007/10/15/levels-of-analysis/#comment-514791 Ed Mon, 15 Oct 2007 20:04:57 +0000 http://neurodudes.com/2007/10/15/levels-of-analysis/#comment-514791 A better example is a laptop. By ignoring the architecture of the processor in an everyday laptop, you miss out on some of the most clever and powerful concepts in the history of human thought. The logic unit, the Harvard architecture, the idea of caching data, the idea of a bus for routing information -- just to stare at the black, plastic case of my laptop would yield no insight into how the computer computes. And yet by staring just a little harder, into the realm of circuits, how the entire device manifests computation into reality, rapidly becomes apparent. A better example is a laptop. By ignoring the architecture of the processor in an everyday laptop, you miss out on some of the most clever and powerful concepts in the history of human thought. The logic unit, the Harvard architecture, the idea of caching data, the idea of a bus for routing information — just to stare at the black, plastic case of my laptop would yield no insight into how the computer computes. And yet by staring just a little harder, into the realm of circuits, how the entire device manifests computation into reality, rapidly becomes apparent.

]]>