Social Connectionism

A Reader and Handbook for Simulations

Frank Van Overwalle

• Publisher: Psychology Press
• Year: 2013
• ISBN:
  9781134956135
• Edition: 1
• Topic: Psychology
• Subtopic: History & Theory in Psychology
• Index: Psychology

Many of our thoughts and decisions occur without us being conscious of them taking place; connectionism attempts to reveal the internal hidden dynamics that drive the thoughts and actions of both individuals and groups. Connectionist modeling is a radically innovative approach to theorising in psychology, and more recently in the field of social psychology. The connectionist perspective interprets human cognition as a dynamic and adaptive system that learns from its own direct experiences or through indirect communication from others.

Social Connectionism offers an overview of the most recent theoretical developments of connectionist models in social psychology. The volume is divided into four sections, beginning with an introduction and overview of social connectionism. This is followed by chapters on causal attribution, person and group impression formation, and attitudes. Each chapter is followed by simulation exercises that can be carried out using the FIT simulation program; these guided exercises allow the reader to reproduce published results.

Social Connectionism will be invaluable to graduate students and researchers primarily in the field of social psychology, but also in cognitive psychology and connectionist modeling.

Part 1
Basics

1 Introduction and overview

Frank Van Overwalle

What is this book about?

This book is a reader of recent original work on connectionist modeling of social phenomena, together with exercises and a computer program to run the connectionist simulations that are described in these articles. The original articles were all published in 1998 or later, so that the book gives an overview of the most recent developments and discoveries in the field. Connectionist modeling is a radically novel approach to theorizing in psychology and, more recently, also in the field of social psychology. In general, a connectionist perspective interprets human cognition as a dynamic, everchanging and adaptive system that learns from its own direct experiences or indirect communication by others. It assumes that cognitive processes are not immediately accessible, as it is becoming increasingly evident that many of our thoughts and decisions occur outside our awareness. Thus, the connectionist approach attempts to elucidate the internally generated cognitive dynamics that are hidden but that drive the thoughts and actions of individuals or groups.

Although many books have been written on connectionism, this is book is unique:

• The book covers the vast domain of social psychology and, more particularly, it includes practically all areas of social cognition (how people perceive and interpret their social world, that is, other people or groups; how they understand themselves and their attitudes; and how they causally interpret the actions and behaviors of themselves or others).

• It is a collection of (abridged) published articles covering the most recent theoretical developments of connectionist modeling in many domains of social psychology. Although in an edited format, it strives for coherence in its introduction of conceptual and theoretical underpinnings (it includes only feedforward or recurrent network models used predominantly in the field). It can thus be used as a handbook to introduce students in the field.

• The book includes exercises of all the simulations described in these original articles (with step-by-step instructions and simulation files), as well as the FIT simulation program, which allows you to run these original simulations on your own. A new version of the program, with a completely novel user interface, makes it even more accessible to users with little experience in the field (i.e. the typical researcher or student in social psychology). The new input now requires users to have no more than minimal computer expertise to specify the simulations in a sort of spreadsheet. The FIT program manual is available in the Appendix to this book.

• Web extras: the book provides supplementary web access to the FIT program (for PC only) and the files for running all simulation exercises. The web access also provides you with regular updates of the FIT program.

What is Connectionism?

To appreciate what connectionism is and what a radical shift in theorizing it brought to psychology, let us briefly take a look at how human thinking was portrayed in previous decades in social psychology.

What it is not: Traditional views on human cognition

Traditional views on human information processing were inspired by a computer analogy. Memory was seen as a sort of filing cabinet or storage bin, where each piece of information was a symbolic concept of some sort that was left in this memory storage so that it could be retrieved for later use. Retrieval consisted of a sequential search through this storage space until the correct piece of symbolic information was found. In more recent notions of activation-spreading networks, human memory was seen as a network of symbolic units connected to each other. (This approach is also termed associative networks but, to avoid confusion with adaptive learning models of an associate nature discussed later, this term is avoided in this book.) In these models, retrieval is easier because the activation flowing through the networks “wakes up”, or activates, some memory traces in parallel. In all these approaches, however, thinking was seen as accessing information from memory or from the outside environment, and modifying it by taking it through various processing steps, sometimes referred to as if-then production rules or semantic comprehension processes. The processed information was then ready for social judgment or action, and could be stored again in memory for later use.

However, many of the production and comprehension processes in traditional models were left unspecified, so that the essence of human thinking was still a mystery. Also unexplained was why this computer-inspired, rational view of human thinking and memory so often went astray, resulting in a plethora of judgment errors and biases. One popular solution to this paradox in social psychology was the idea that humans avoid excessive mental effort and use mental short-cuts or heuristics. To stress this point, researchers termed social perceivers either “cognitive misers” or “motivated tacticians” because they used whatever mental means—either heuristics or mental effort—to make their decisions. But this, again, failed to explain what was making these tactical decisions in the brain, another paradox known as the homunculus problem. Who or what in the brain decided that a given heuristic was the optimal tool from our mental heuristic “toolbox” to solve a given social question?

Connectionism: The essentials

The connectionist approach avoids many of the earlier traps that traditional models could not escape from. Crucially, it starts from a radically different view of human thinking and memory, based on an analogy with the architecture and working of a natural human brain. The basic feature of a connectionist network is that it consists of units (representing symbolic and non-symbolic concepts) and connections between the units along which activation is spread. In terms of the brain analogy, the units are analogous to neurons and the connections are analogous to synapses, which differ in strength or permeability for the spread of signals in the nervous system. The activation of each unit in the network is determined by the total activation received from other units along the connections. Connectionist models determine how activation spreads across the connections (in the short term) and also compute how the strength of these connections between units changes over time (in the long term). It is this latter feature—the ability to change the strength of the connections—that makes connectionist approaches so powerful as learning devices. Since the publication in 1986 of the two volumes on Parallel Distributed Processing by Rumelhart and McClelland (1986) and McClelland and Rumelhart (1986), this approach has emerged as one of the leading paradigms in the study of cognitive processes, including social cognition.

Many of the paradoxes and mysteries of older theories can immediately be solved by the connectionist approach, especially when it involves routine social thinking. If you plan on reading this book, I hope to convey to you two key messages that result from a connectionist perspective:

1. This book sees humans as “implicit cognizers”, who rely in most of their decisions and judgments on the automatic generation and reactivation of novel knowledge. Automatic processes play a prominent role in everyday social thinking, perhaps more than we are aware of because they are hidden from our conscious experience. What is the reason for believing that such unconscious processes exist? This is based on the logic that, in the evolutionary past, the human brain developed novel functions and, while doing so, it typically did not discard old structures that were still functioning but reused them for similar or slightly adjusted functions. Thus, from an evolutionary point of view, it is highly plausible that primitive associative or connectionist-like mechanisms are still of use. If most of human social behavior can be explained by a simple bunch of interconnected brain-like units, why assume anything more complicated?

2. Human social behavior involves adapting to a constantly changing social world. People move around, friends and colleagues change, new groups pop up or their members change, and, on top of this, people’s interests, goals and even their more deeply engrained traits change over time. Every one of us has to deal with these short- and long-term social changes in a fluent and efficient way. This is only possible if our capacity to deal with change is somehow built-in to our cognitive system. And this is exactly what connectionist systems are particularly good at. An inherent part of these systems is a learning algorithm that adjusts to flows of change in the external environment, so that the cognitive system is appropriately prepared to understand and anticipate a novel state of the external environment.

Connectionism: A deeper look inside

If you are convinced that these are important preconditions for a cognitive system to survive in a social environment, then you might wonder: What are the characteristics of a connectionist system that accomplish all that? Below I highlight the most important ones for theorizing in social cognition.

• In a connectionist network, units create representations of single individuals or groups of people, together with elements of their social context and their reactions and behaviors. The connections between these elements represent the associations in the brain, and thus may reflect memories, judgments about other people and groups, attitudes or approach or avoidance tendencies, causal inferences and (re)actions. Hence, a single concept immediately activates related concepts, such as where you know a person from, how much you like him or her, what he or she looks like and what dispositions or talents he or she possesses.

• Processing in a connectionist model consists of activation spreading along the units and modifying their connection strengths. Both these processes are controlled by the units directly involved, and are computed by them in parallel. Connectionist processes, therefore, can get rid of a central executive processor to make the necessary computations. Much of our routine social thinking and reasoning that occurs largely outside our awareness can be understood by this perspective. The fact that the computations in the network occur in parallel explains why routine social thinking is very fast. However, some of our cognitive activities lie outside the scope of connectionist theories, such as when we make deliberative choices that involve sequential steps, like making arithmetic calculations or listing all the pros and cons of a decision.

• Memory is not a huge cabinet where each individual piece of information is stored. Rather, memorizing involves a slow change of connection strengths so that new information about people, objects and their associated features slightly changes old memories about them. Hence, in a connectionist network, your impressions of familiar persons change slowly as you learn new information about them, and the old and new information are automatically and seamlessly integrated. Moreover, a connectionist memory is distributed among many features related to the target concept that is processed. Thus, your altered impression of a single member of a social group may generalize to the whole group and affect all members of it.

Although the connectionist approach has only recently been introduced to social psychology, tremendous progress has been made in the last decade, and many domains of social psychology have been covered by a connectionist approach. One of the leading motivations of this book is to demonstrate that the current state of affairs of theorizing in social psychology—a plethora of fragmented mini-theories, each designed to explain a limited set of findings—can be reappraised within the greater realm and framework of a connectionist approach. Connectionism provides novel insights to old research questions, and so contributes to progress in social psychology. To illustrate, several seemingly different types of thinking or biases that are traditionally interpreted as involving different processes may actually be understood by the same underlying connectionist processing principles. Because concepts and memories are distributed across the network and associated with a large pool of distinct features, you and I may access different parts of it. For instance, your attitude may be formed by paying attention to the ...