connectionism

Connectionism is the idea that our minds aren't like traditional computers with separate modules for different tasks. Instead, it says the brain works more like a huge network of connected units—like neurons and the links between them. In fields like artificial intelligence and cognitive science, it suggests that learning happens when those links get stronger or weaker with experience, rather than by following a set of pre-programmed rules. Imagine learning to ride a bike: you don't consciously tell your muscles what to do; your brain learns through repeated attempts and small adjustments—that's connectionism at work.

Connectionism contrasts with older approaches, like symbolic AI, which focus on logic and rules. Today, it's closely tied to artificial neural networks and deep learning, where networks of simple units learn patterns from data. You might hear it discussed in relation to how memories are formed or how we recognize patterns.

Connectionism is a school of thought within cognitive science that views mental processes as emerging from large-scale networks of interconnected nodes—often modeled after the biological structure of the brain. It's a bit like imagining your mind isn't a single, powerful processor, but rather a bustling city where information travels along countless roads and pathways.

Historically, connectionism arose as a response to "symbolic AI," or "Good Old-Fashioned Artificial Intelligence" (GOFAI), which dominated the field for decades. GOFAI proponents believed intelligence could be achieved by manipulating symbols according to pre-defined rules. Connectionists argued this approach was too rigid and didn't capture the flexibility and adaptability of human cognition.

Instead, connectionism proposes that knowledge isn't stored as discrete symbols but is distributed across a network of connections between simple processing units (often called "nodes" or "neurons"). Learning happens by adjusting the strength of these connections—strengthening those used frequently and weakening those rarely used. This process is known as "parallel distributed processing" because many nodes work simultaneously, rather than sequentially as they do in traditional computers.

You'll encounter connectionism most often when discussing artificial neural networks (ANNs), which are computational models inspired by the brain. These ANNs form the basis for deep learning, a powerful branch of machine learning responsible for breakthroughs in areas like image recognition, natural language processing, and game-playing.

However, connectionism isn't limited to AI. It also influences our understanding of human memory (how memories are stored as patterns of activation across networks), perception (how we recognize objects by identifying features and their relationships), and even consciousness itself. While it doesn't offer a complete explanation for all cognitive phenomena, connectionism provides a compelling framework for understanding how complex mental processes can arise from relatively simple interactions within a vast network.