3 of 14

What is “Computational Social Psychology”?

Computational social psychology uses computational modeling and analysis to formalize, investigate, and test psychological mechanisms operating within complex social systems.

It is a subfield of social psychology that uses computational approaches across the full research cycle: theory building, hypothesis generation, research design, hypothesis testing, data analysis, prediction, explanation, and interventions.

4 of 14

What is “Computational Social Psychology”?

An integrative definition includes two key elements.

First, it has a formalizing orientation: it seeks to express social psychological theories in computational or mathematical-based form.

Second, it has a systems perspective: it examines how psychological processes operate within and across dyads, groups, networks, institutions, platforms, cultures, ecologies, time, and historical contexts.

5 of 14

Two Pillars

The current state of computational social psychology is shaped by two complementary approaches: computational modeling and computational analysis.

Computational modeling refers to the construction of formal representations of social psychological theories. It is especially important for theory building because it helps researchers formalize assumptions, specify mechanisms, simulate possible outcomes, generate predictions, and explore how psychological processes may produce social patterns over time.

Computational analysis refers to the use of computational tools to measure, detect, classify, predict, and validate patterns in data. It is especially important for empirical testing because it helps researchers study social psychological processes in large, complex, or naturally occurring datasets, including text, networks, digital traces, behavioral reports, and experimental or survey data.

6 of 14

Level of analysis	Role of the method in the research process
Level of analysis	Theory building/ hypothesis generation	Empirical testing/ measurement/validation
Interpersonal	Agent-based modeling; social simulation; dynamical systems models; opinion dynamics models; synthetic populations and digital twin; LLM-based simulations; models of polarization, norm change, residential segregation, intergroup contact, innovation diffusion, corporation, and collective action	Social network analysis; digital trace data analysis; large-scale text analysis; GIS and spatial analysis; platform data analysis; empirical tests of diffusion, contagion, clustering, and polarization; group dynamics, social interactions, and collective intelligence; network intervention
Intrapersonal	Computational cognitive modeling; Bayesian modeling; neural network and connectionist models; cognitive architectures; models of impression formation and stereotyping, models of attitude formation, change, and consistency	Machine learning models of individual judgment; NLP measures of emotion, morality, identity, prejudice, or social perception; attitude network analysis; behavioral prediction; model comparison using experimental or survey data

Several computational approaches are currently shaping the field of computational social psychology. One way to organize them is along two dimensions (Table 1). The first is the level of analysis. Some methods focus more on interpersonal-level processes, such as interaction, influence, networks, groups, culture, institutions, and collective behavior. Others focus more on intrapersonal-level processes, such as learning, judgment, belief updating, impression formation, and stereotype formation. The second dimension is the role of the method in the research process. Some methods are especially useful for theory building and hypothesis generation, whereas others are especially useful for empirical testing, measurement, and validation. These distinctions are not strict. Many methods can serve more than one purpose. Still, the two dimensions help clarify the current methodological landscape of computational social psychology.

7 of 14

Science of Complex Systems

It studies how complex patterns of a system emerge from simple interaction rules of components.
It builds a causal link between one level and the next.
It is interdisciplinary.
Agent-based modeling is one of the primary research methods.

Murray Gell-Mann (1987) Simplicity and Complexity in the Description of Nature for Coarse-Graining”

8 of 14

What is “Agent-Based Modeling”?

It is a scientific research method that studies the behaviors of a complex system by simulating an artificial society where multiple agents interact by following simple rules to generate the behaviors of the system.

9 of 14

“The whole is something besides the parts”

Complex systems are composed of parts or components which interact with each other at small scales.

Many components can interact and self-organize to exhibit global structures.

Complex systems change their state dynamically, are often unpredictable, can adapt and evolve.

Complex

Systems

Aristotle Metaphysics (Translated by W. D. Ross)

Murray Gell-Mann (1987) Simplicity and Complexity in the Description of Nature

10 of 14

A system

A forest

Parts that interact

Many different species of organisms (trees, birds, mammals, insects, fungi, bacteria, virus, etc)
Predator-prey, parasite-host, pollinator-pollinated, etc

Example 1: Let’s say we study forests – complex ecosystems

To study complex systems, we have to specify the level of detail that we are going to talk about and ignore everything else – “coarse-graining”

Let’s say we want to understand when forests are robust and/or fragile to fires.

We can “coarse-grain” by focusing on the density of trees and ignoring everything else.
Then, trees are “agents” in our forest fire model.

Murray Gell-Mann (1987) Simplicity and Complexity in the Description of Nature

11 of 14

A system

A human body

Parts that interact

Many different levels of subsystems (cells, tissues, organs, systems)
Different properties and arrangements
They are organized in a hierarchical structure.

Example 2: Let’s say we study human bodies – complex biosystems

To study complex systems, we have to simplify a complex system and specify the parts of interest and leave all other parts as black boxes – “near-decomposability”

Let’s say we want to understand how our body responds to virus.

You can “nearly-decompose” a human body into systems, and focus on the immune system and ignore the detail of activity in other systems.
Then, your immune system model includes immune-related components.

Herbert A. Simon (1968) The Sciences of The Artificial for “Near-Decomposability”

12 of 14

A system

A political institution

Parts that interact

Public, legislature, justice, executive branch
Federal, state, regional government
Constitutions, laws, rules, policies

Example 3: Let’s say we study a political institution – complex social systems

How would you “coarse-grain? What would you like to know?

Let’s say we want to understand why two party structure persists in some countries , and in other countries multiple parties emerge.

We can “coarse-grain” by focusing on parameters related to election (electorate, delegates, voting rules, representative hierarchy) and ignoring everything else.
Then, electorate, delegates, and candidates are “agents” in our political party model.

Murray Gell-Mann (1987) Simplicity and Complexity in the Description of Nature

13 of 14

Agents can be any level

Nations, political parties
Ants, birds, bees, people
Cognition, beliefs, attitude, reasons, evidence
Cells, particles

Boundedly rational

Agents have local, imperfect information.
Agents have limitations–biases, heuristics, memory, vision,

Heterogeneous

Agents have different roles, or different sets of elements.

Agents

In ABM, agents represent the parts/components of a complex system.

14 of 14

What are some research questions that �ABM is used for in social science?

ABM can establish the causality between micro-level, individual agent behaviors and macro-level system behaviors.

How much prejudice does it take to produce residential segregation?

If people tend to become more alike in their beliefs and behavior when they interact, why do not all such differences eventually disappear?

How much leniency do we need to spread minority’s voice to our society and maintain diversity?

Would our society still be polarized if we were all completely rational?