Where does agency collide?

An agent is anything that perceives its environment through sensors and acts on it through actuators; a rational agent chooses the action expected to maximize its performance measure. Russell and Norvig made this the organizing idea of the standard textbook, and it deliberately includes thermostats. Wooldridge and Jennings gave the stronger notion: a system that is autonomous, reactive, proactive, and social, and that may be described with mentalistic terms (beliefs, desires, intentions) as a convenient stance rather than a claim about what is inside. Agency here is a design category, not a capacity anyone possesses.[1][2]

Unit

A system in an environment

Recovery means

Not applicable, an agent is specified, not restored

Measured by

Performance measure over an environment; task success; degree of autonomy in a taxonomy

An agency relationship is a contract under which one party (the principal) engages another (the agent) to act on their behalf, delegating some decision authority. Jensen and Meckling defined it in 1976 and named the problem: with diverging interests and costly monitoring, the agent will not always act in the principal's interest, and the sum of monitoring, bonding, and residual loss is the agency cost. Eisenhardt's review fixed the vocabulary for management. The word names a relationship and its frictions, never a capacity, an agent in this meaning is someone who works for someone else.[3][4]

Unit

A principal–agent pair under a contract

Recovery means

Reduced agency cost, through monitoring, incentive contracts, or ownership

Measured by

Agency costs inferred from ownership structure, pay–performance sensitivity, monitoring expenditure

The capacity of actors to shape an emerging technological path from inside it, mobilizing resources, meanings, and other actors rather than standing outside as a heroic founder. Garud and Karnøe contrasted Danish wind-turbine bricolage with the American breakthrough strategy and argued that agency was distributed across producers, users, evaluators, and regulators, and embedded in the very path it was creating. This is Giddens's structure–agency problem imported into innovation studies, with the entrepreneur as the case.[5]

Unit

A collective of actors over the life of a path

Recovery means

Not applicable, paths are created, not recovered

Measured by

Historical and process case studies; tracing who mobilized what, and when

The capacity of actors to act otherwise than they do (to make a difference) within structures that both constrain and enable them. Giddens's structuration theory made structure and agency a duality: rules and resources exist only as they are reproduced in action, and action draws on them. Emirbayer and Mische reconstructed agency as temporally oriented (a habitual past, a projective future, and a practical-evaluative present) precisely because the word had come to mean little more than "not structure." Nobody in this tradition has a scale for it.[6][7]

Unit

An actor situated in a structure, over time

Recovery means

Not applicable

Measured by

Not measured, interpreted from action, narrative, and institutional context

For Bandura, to be an agent is to intentionally make things happen by one's actions; its core features are intentionality, forethought, self-reactiveness, and self-reflectiveness, with self-efficacy beliefs as the foundation. Cognitive neuroscience narrowed the word to the sense of agency: the experience of controlling one's own actions and, through them, events in the world. Haggard reviewed its two sources (a prospective signal from motor preparation and a retrospective comparison of predicted and actual outcomes) and its breakdown in schizophrenia and under coercion. The first is a capacity; the second is a feeling that can be present when the capacity is not.[8][9]

Unit

A person, or a single action and its outcome

Recovery means

Restored sense of control after loss, clinically, after psychosis or learned helplessness

Measured by

Self-efficacy and sense of agency questionnaires; intentional binding; explicit judgments of authorship

Something is an action, and its author an agent, when it can be explained by the agent's reasons (a belief and a pro-attitude that, Davidson argued in 1963, also cause the action). Frankfurt raised the bar: a person, as opposed to a mere wanton, has second-order volitions (desires about which desires should move them) and acts freely when the will they have is the will they want. Neither view is about degrees or measurement. Agency is a status that an event and its author either have or lack.[11][12]

Unit

An action and its reason

Recovery means

Not applicable

Measured by

Not measured, argued from cases and counterexamples