06.09.2021 Views

Mind, Body, World- Foundations of Cognitive Science, 2013a

Mind, Body, World- Foundations of Cognitive Science, 2013a

Mind, Body, World- Foundations of Cognitive Science, 2013a

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

were provided earlier in this chapter, in the discussion <strong>of</strong> carving pattern spaces into<br />

decision regions and the determination that output unit activities could be interpreted<br />

as being conditional probabilities.<br />

That computational analysis is possible for both connectionist and classical<br />

cognitive science highlights one similarity between these two approaches. The<br />

results <strong>of</strong> some computational analyses, though, reveal a more striking similarity.<br />

One debate in the literature has concerned whether the associationist nature<br />

<strong>of</strong> artificial neural networks limits their computational power, to the extent that<br />

they are not appropriate for cognitive science. For instance, there has been considerable<br />

debate about whether PDP networks demonstrate appropriate systematicity<br />

and componentiality (Fodor & McLaughlin, 1990; Fodor & Pylyshyn, 1988;<br />

Hadley, 1994a, 1994b, 1997; Hadley & Hayward, 1997), two characteristics important<br />

for the use <strong>of</strong> recursion in classical models. However, beginning with the mathematical<br />

analyses <strong>of</strong> Warren McCulloch (McCulloch & Pitts, 1943) and continuing<br />

with modern computational analyses (Girosi & Poggio, 1990; Hartman, Keeler, &<br />

Kowalski, 1989; Lippmann, 1989; McCulloch & Pitts, 1943; Moody & Darken, 1989;<br />

Poggio & Girosi, 1990; Renals, 1989; Siegelmann, 1999; Siegelmann & Sontag, 1991),<br />

we have seen that artificial neural networks belong to the class <strong>of</strong> universal machines.<br />

Classical and connectionist cognitive science are not distinguishable at the computational<br />

level <strong>of</strong> analysis (Dawson, 1998, 2009).<br />

Let us now turn to the next level <strong>of</strong> analysis, the algorithmic level. For classical<br />

cognitive science, the algorithmic level involves detailing the specific information<br />

processing steps that are involved in solving a problem. In general, this almost always<br />

involves analyzing behaving systems in order to determine how representations are<br />

being manipulated, an approach typified by examining human problem solving<br />

with the use <strong>of</strong> protocol analysis (Ericsson & Simon, 1984; Newell & Simon, 1972).<br />

Algorithmic-level analyses for connectionists also involve analyzing the internal<br />

structure <strong>of</strong> intact systems—trained networks—in order to determine how they<br />

mediate stimulus-response regularities. We have seen examples <strong>of</strong> a variety <strong>of</strong> techniques<br />

that can and have been used to uncover the representations that are hidden<br />

within network structures, and which permit networks to perform desired inputoutput<br />

mappings. Some <strong>of</strong> these representations, such as coarse codes, look like<br />

alternatives to classical representations. Thus one <strong>of</strong> classical cognitive science’s<br />

contributions may be to permit new kinds <strong>of</strong> representations to be discovered and<br />

explored.<br />

Nevertheless, algorithmic-level analyses also reveal further similarities between<br />

connectionist and classical cognitive science. While these two approaches may propose<br />

different kinds <strong>of</strong> representations, they still are both representational. There<br />

is no principled difference between the classical sandwich and the connectionist<br />

sandwich (Calvo & Gomila, 2008). Furthermore, it is not even guaranteed that the<br />

Elements <strong>of</strong> Connectionist <strong>Cognitive</strong> <strong>Science</strong> 201

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!