What is connectionism in cognitive science?

Connectionism is an approach to studying human mental processes and cognition using artificial neural networks. Unlike classical computationalism, which posits that minds operate by manipulating symbols according to formal rules, connectionism models the mind as a complex network of communicating nodes. It has evolved through multiple historical waves, moving from early perceptrons to modern deep learning and large language models.

What we know

• Connectionism is an approach in cognitive science that seeks to explain mental phenomena using artificial neural networks. [1, 2]
• Classical computationalism models cognitive processes as manipulating symbols according to formal rules, whereas connectionism models the mind as a network of nodes where information flows through communication. [3]
• The first wave of connectionism began in 1943 with McCulloch and Pitts and was later challenged by Minsky and Papert's 1969 book Perceptrons, leading to a period of reduced research activity. [2]
• The second wave of connectionism blossomed in the late 1980s following the introduction of parallel distributed processing models featuring hidden layers and sigmoid activation functions. [2]

Still open

• It remains difficult for connectionist models to decipher how artificial neural networks process information or account for the compositionality of mental representations.

Where to go next

• what caused itHow did the Perceptrons book cause an AI winter?
• compared withWhat are the core differences between computationalism and connectionism?
• how it workedHow do artificial neural networks process information?
• how we knowWhat role did the Parallel Distributed Processing book play in cognitive science?
• the bigger pictureWhat is cognition?
• an unexpected connectionHow do 1943 neural circuitry models connect to modern large language models?Connectionism traces a direct lineage from Warren McCulloch and Walter Pitts's 1943 formal models of neural circuitry all the way to contemporary deep learning and large language models.

Sources

• [1]connectionism (Wikidata Q203790) · CC0 1.0
• [2]Connectionism · CC BY-SA 4.0
• [3]Cognition · CC BY-SA 4.0
• [4]List of philosophies · CC BY-SA 4.0
• [5]Artificial intelligence · CC BY-SA 4.0