Symbols, Connections, and Optimization by Trial and Error

This chapter introduces the technical concepts that provide the foundation of synthesis of social and computer scientific theorizing. It discusses cognitive processes and the trial and error optimization of hard problems. The earliest approaches to artificial intelligence (AI) assumed that human intelligence is a matter of processing symbols. Symbol processing means that a problem is embedded in a universe of symbols, which are like algebraic variables; that is, a symbol is a discrete unit of knowledge that can be manipulated according to some rules of logic. Neural networks and the closely related fuzzy logic models are called “universal function approximators” because they can reproduce the outputs of any arbitrarily complex mathematical function. Their power as psychological models has grown because of their ability to simulate human processes such as perception, categorization, learning, memory, attention, and the errors that humans make. The social strategies of humans and evolutionary search are population-based, meaning that many solutions can be tested in parallel; in both cases, interactions among population members result in problem solving intensity greater than the sum of individuals' solitary efforts.