Theories of Agency

Key concepts present within "agency": the individual, action, will, intentionality, choice, freedom

Key concepts against which "agency" is commonly situated: structure, determinism, society, environment, inevitability



What is the individual, self or person? (e.g., what is the unit of 'agency'?) What, in contrast, is not-agent (environment, structure, inanimate)?

Postmodernist and poststructuralist skepticism about the individual or "the human subject".

How does the agent know about the difference between itself and the environment?

Cartesianism: the self is that which knows itself; existence is best understood by radical categorical divisions between mind-body, self-other, etcetera, for heuristic and ontological reasons.

What is an action?

Does the agent choose or will its action in the world?

Does agency exist even if the act changes nothing in the environment? Is there more agency if there is more change?

Does agency exist if the intentionality of the action and the change bear little or no resemblance to each other?


Social and Behavioral Science

Agency determines everything

Libertarianism and objectivism

Certain forms of Christian theology, both evangelical Protestantism and Deism (with the frequent proviso that God is the "uncaused cause" or prior determination of the individual struggle against sin)

Certain forms of 19th Century liberalism

Structure determines everything (macrostructures or microstructures)

Calvinist predetermination

Strong forms of structuralist anthropology, folklore and psychoanalysis (Levi-Strauss, Jung)

Strong forms of genetic determinism

Strong forms of developmental or evolutionary psychology (Skinner, Buss)


All practices and behaviors of agents are determined by logics which precede those practices, and which always make rational sense in objective terms outside the perception of human actors (which human actors may or may not be aware of) (Marvin Harris on human diet; Talcott Parsons on human institutions)

Certain forms of teleological Marxism, Hegelianism and other 19th Century social thought .

Structure-Agency feedback loop

Can be strongly determinist or indeterminist, depending on how closed the loop is represented as being. Malthusian thought, for example, is a structure-agency feedback loop, but it is intensely determinist.

Social contract theory

Individuals consent in some initial pre-social state to a foundational understanding of their social rules and institutions; those rules have binding force on individuals and exist outside of their agency until such time as sufficient numbers of individuals choose to withdraw their understood consent to the legitimacy of social structures.

Can have a "negative spin", as in Hobbes: social institutions as the only constraint which keeps individual agency from producing horrible suffering.

Anthony Giddens and structuration theory

Modernity not as "iron cage" (Weber) or "prelude to utopia" (Marx) but as a condition collectively chosen through the deliberate actions of many people; agency determines structure which determines the possibilities for the expression of agency and so on ad infinitum.

Neoclassical economic thought

Agents act out of self-interest, individually and differentially perceived and measured and achieved; the sum total of individual action is (or ought to be) a well-ordered political economy that maximizes the aggregate opportunities for self-interest even though the results for every individual will not be equally optimal (equal opportunity, non-equal results).

Historicist anti-functionalism and some forms of evolutionary theory

Practices, behaviors and institutions are 'structure', but explained largely by the fact of precedent and inertia, not by deeper 'preset' functionalism that precedes and trumps change over time; no teleological end to change. "One damn thing after another".

The "bounded circle" of agency

Agency exists within tight constraints, but is free within those constraints. this is a common way ever since the Enlightenment to describe the agency of individuals: absolutely constrained beyond a certain boundary, absolutely free or devolving upon the individual within it. Sometimes this is only an axiomatic assumption governing social institutions and sometimes it is an ontological assertion about agency. (e.g., you could argue that modern American criminal law assumes absolute individual responsibility for actions once constraints of circumstance and environment are considered, but does not require an ontological assertion about the reality of agency).

"Men make history, but they do not make it just as they please"--Karl Marx

Marx needs this in order to believe in the possibility of revolution, but it has long been debated among Marxists since Marx's time whether the "humanist" Marx who seems to believe in a limited but critical role for will and agency in choosing a revolutionary moment or the "scientific" Marx who believes in the structural inevitability of revolution.

"Methodological individualism"

Structure exists, and has determinant force, but a conscious heuristic decision that what individuals choose to do, or perceive themselves as choosing, is interesting as an object of study--not the individual as a "case study" of a larger whole, but the individual as exceptional or particular.


Computer simulation

Defining "autonomous agent" is a significant issue among people working with computer simulations.

Stan Franklin and Art Graesser propose the following definition: "An autonomous agent is a system situated within and a part of an environment that senses that environment and acts on it, over time, in pursuit of its own agenda and so as to effect what it senses in the future."

The sensory aspect of their definition is especially interesting. They argue, for example, that a robot with only visual sensors is no longer an agent at the moment that it is placed within an environment completely lacking in light.

In computing, they argue, "all agents are software programs, but not all programs are agents".

They also propose a really interesting classificatory scheme for software agents:

reactive (sensing and acting) responds in a timely fashion to changes in the environment
autonomous exercises control over its own actions
goal-oriented pro-active purposeful does not simply act in response to the environment
temporally continuous is a continuously running process
communicative socially able communicates with other agents, perhaps including people
learning adaptive changes its behavior based on its previous experience
mobile able to transport itself from one machine to another
flexible actions are not scripted
character believable "personality" and emotional state.


Emergent systems

Rely on autonomous agents (by definition?), whether in computers or other environments, meaning that they rely on individuated agents who act on environments according to rules: agents have rules associated with them individually and seek to fulfill these rules by changing the environment.

Agents do not choose their rules and cannot alter them individually or autonomously.

Environmental changes can alter the medium in which rules are expressed and thus alter the observable behavior of agents; the rules have not changed.

Emergent system simulations like NetLogo do not model the problem of imperfect information or agent perceptions of their environment; real-world emergence, in contrast, may depend upon it.

"Structure" is determined by the action of agents, but is in turn determinant of the actions of agents.

Structure has a temporal determination: it is cumulative over time in a single direction, non-reversibly, and determines the action of agents differently as it evolves; agents cannot undo the history of structure.

Unpredictability is strictly a by-product of agent-environment interactions and a probable observer effect, not of the variable expression of or interpretation of rulesets by agents.

Generally anti-functionalist: structures are not determined by harmonious correspondence with some deeper real or foundational logic or purpose, but only by the cumulative interaction of agents acting on set rules with environments.


Contrast with:

Genetic algorithim models of simulation where environmental structures built by agents alter the ruleset of the agents, and agents compete with one another to accomplish objectives and survive to next generation; with any evolutionary process whereby the changes to the environment in turn end up changing the actual rules attached to particular agents and where those rules dynamically shape which agents continue to be present within the environment. (e.g., where death occurs).

"Top-down" models where the rules are situated in the environment and there are no agents per se. (Where what may seem like agents are merely units of the environment).

Visions of agency that leave room for the variable expression of will, intention, desire; for instance-dependent and agent-dependent variable effectiveness of execution of rules; for "rule-breaking"; for imperfect information affecting agent knowledge of environment or expression of ruleset.


So, big questions and issues about agency and emergence:

In general, strong philosophical biasing towards will or choice runs quickly into the problem of simultaneity, pattern, organization.

Example: Baby names. Many couples think that they are choosing a baby name largely independently and autonomously, but strong 'coincidental' simultaneities that unmistakeably resemble the way emergence works are clearly affecting that choice.

The problem of observer bias: when do you know something is really a pattern? Example: evolutionary psychology depends crucially on the assertion of universalities, but is often willing to declare a universality based on what is ultimately a very small sampling (and usually based on survey sampling rather than ethnographic observation, but that's another issue) of the total possible range of observable human behaviors, most particularly when you include not just contemporary but past examples.

In general, strong biasing towards structure or determination poses the problem of novelty, change, unexpected or unpredictable results.

Example: Parsonian functionalism stressed equilibrium and the ideal purpose of every single institution to the point that it had absolutely no way to explain why anything ever changes.

In emergent systems, once a stable equilibrium emerges in a given system, perturbations must come from outside the system in order to produce certain kinds of observable change. The Game of Life won't do anything new after a certain point if someone doesn't come and make it do something new. (Is the "interesting" type of Wolfram's cellular automata at 'equilibrium' in its predictable unpredictability?) Where does novelty come from in emergent systems, especially in terms of applyiing them to human history and the human future?

Does emergence show us new ways to think about the relationship between agency and structure?

Yes; it is a a powerful new way to think about the "structure-agency loop", both philosophically and practically (what it means, and how it works).

Does it tell us anything about the perspective of agents in an emergent system, how they see and understand their environment and their place within it?

Not really, not yet, (not ever?)

Seeing things from the agent's point-of-view: the consciousness illusion, but how can you get an illusion that can perceive its own illusory nature? If I think that my thought is an illusion, why do I continue to think?

Do emergent systems tell us anything about the possibility of directed, intentional action in the world?

Not in "strong emergence": ultimately "strong emergence" adds up to an intensely determinist slanting of the agency-structure loop which is resigned to explaining change in the world as the consequence of the determinate interaction of many numerous complex systems rather than the result of willful effort by agents within the system. But in "weak emergence"? Yes, absolutely: in this case emergence has the possibility of being an applied technology of change over time. But this is only possible if there is a conception of an objective or goal which is seen as been intrinsically outside the loop of emergence, as something which is chosen by the agents who wish to pursue it.

Meme-starting experiments and the "tipping point"

Out of Context

How Weblogs Can Turn an Idea Into an Epidemic

Google Bombing

Seth Godin's "ideavirus" and marketing

Cheese-eating surrender monkeys

The Church of Virus

"Darwininan Processes and Memes in Architecture: A Memetic Theory of Modernism"

"Chess Moves and Their Memomics: A Framework for Evolutionary Processes of Chess Openings"

Best practices and cumulative precedent

The idea of "best practices" in social institutions

British common law as engineered emergence

Transparency and improving the flow of information to human agents

Smart mobs: communicative technologies that streamline and direct the flow of agent-environment interaction (the Internet) and possibly make it more susceptible to directed engineering