Agency

This note is about agency but it also relates to the notions of voice, action, cognition, consciousness.

Agent
Voice
Autonomy
Biosemiotics
Meaning
Biological Individuality
Self
goals
teleonomy: directedness that self-reproducible systems acquire through evolution 1, the study of goal-directed processes 2

Definitions

"Agency is a matter of doing, trying, initiating. Agency is by-me-ness; it is being a source of action and its effects."

Godfrey-Smith, Peter. Metazoa: Animal Life and the Birth of the Mind. New York: Farrar, Straus and Giroux, 2020.

Agency is a dynamical property of a system. Capacities to respond to the conditions:

transduce
configure
respond

Agential systems can maintain their stability in the changing conditions (all life has agency in a sense of resisting entropy).

Cf. Subjectivity

On the close relationship of consciousness and action, see:

Hurley, Susan L. Consciousness in Action. Cambridge, MA: Harvard University Press, 1998.

Levin, Michael, and Richard Watson. “Machines All the Way up and Cognition All the Way down: Updating the Machine Metaphor in Biology.” Seminars in Cell & Developmental Biology 177–178 (2026): 103668. https://doi.org/10.1016/j.semcdb.2026.103668.

In Biology

Modern Synthesis describes dynamical properties of populations at the cost of distorting the nature of biological processes that contribute to evolution.

The Agential Perspective represents biological process with greater accuracy but at the cost of generality.

Walsh, Denis M., and Gregory Rupik. “The Agential Perspective: Countermapping the Modern Synthesis.” Evolution & Development, 2023, 1–18. https://doi.org/10.1111/ede.12448.

Much novelty/innovation in evolution is possible without the need to evolve novel genes or cell types.

Sultan, Sonia E., Armin P. Moczek, and Denis Walsh. “Bridging the Explanatory Gaps: What Can We Learn from a Biological Agency Perspective?” BioEssays 44, no. 1 (2022): 2100185. https://doi.org/10.1002/bies.202100185.

Cf. levels of agency, organisational levels, boundaries of self/individuality.

Francis, Robert A. “Ecosystem Prediction and Management.” In A Companion to Environmental Geography, edited by Noel Castree, David Demeritt, Diana Liverman, and Bruce Rhoads, 421–41. Chichester: Wiley-Blackwell, 2009.

From Regulation to Agency

An attractor requires no agent, so dynamical convergence turns into agency only at a price, and the price rises rung by rung:

Persistence: the pattern remains.
Attraction: trajectories converge after displacement.
Regulation: feedback counteracts deviations in particular variables.
Adaptivity: regulation adjusts in relation to viability boundaries before failure. Graded normativity arrives here.
Goal-directed agency: the system selects actions relative to an endogenously maintained condition across varying circumstances.
Collective agency: regulation closes at the collective level rather than existing as an analyst's aggregate.
Representational agency: internal models, memories, or anticipations do the selecting.

Persistence alone, the first rung, supports none of the later attributions. The middle rungs do real explanatory work, because organisms generate evolutionary novelty through plastic, goal-directed responses that population-level accounts miss (see In Biology above). The upper rungs remain contested, which makes the ladder a field of live hypotheses rather than a settled scale.1 2

On adaptivity as the source of graded normativity:

Di Paolo, Ezequiel A. “Autopoiesis, Adaptivity, Teleology, Agency.” Phenomenology and the Cognitive Sciences 4, no. 4 (2005): 429–52. https://doi.org/10.1007/s11097-005-9002-y.

Barandiaran, Xabier E., and Matthew D. Egbert. “Norm-Establishing and Norm-Following in Autonomous Agency.” Artificial Life 20, no. 1 (2014): 5–28. https://doi.org/10.1162/ARTL_a_00094.

Operational questions for any alleged agent:

Question	Evidence sought	Failure mode
What is the candidate boundary?	Recurrent material or organisational closure	The analyst draws a convenient region
Which variable is controlled?	Perturb-and-recover evidence	Persistence without correction
What counts as error?	Response scaled to deviation	Reflexive covariance with a stimulus
Where is memory?	History-dependent response	Retrospective storytelling
How are actions selected?	Equifinal or context-sensitive alternatives	One fixed causal pathway
At what scale does control close?	Collective feedback irreducible to independent responses	Aggregation fallacy
Whose viability is served?	Named beneficiaries and costs	Harmony rhetoric masking conflict

Cf. Phase Space on attractors and targets, and Transformation on why retargeting as a design method assumes a system on the fifth rung or above.

Agents as Aligned Goals

The ladder above ranks how much a system does and leaves open where the system ends. Levin answers both questions through goals. He defines a Self as "a coherent system emerging within a set of integrated parts that serves as the functional owner of associations, memories, and preferences, and acts to accomplish goals in specific problem spaces where those goals belong to the collective and not to any individual sub-component."3 Alignment therefore makes many parts count as one. A blastoderm of thousands of cells counts as one embryo because its cells work toward one target anatomy, and scratches that isolate islands of the same tissue yield twins or triplets.4 A boundary persists where parts keep the same condition in range.

The scale of those goals sets the size of the boundary. Levin calls this scale the cognitive light cone: the most distant events, in space and time, that a system can care about and work to change.5^,3 The cone is a schematic of goal space, widest at the present and fading backward into memory and forward into anticipation. The ranges of sensing and of effectors are separate matters, and Levin says so explicitly.3 The distinction matters wherever Design Potential (Private) and Personas use the concept.

Property	Claim	What a claim about agency then has to state
Compound	Every agent consists of sub-agents and helps to form super-agents, each with a light cone of its own.3	The level in question
Aligned	Parts form one agent while they work toward a shared target.4	The condition the parts keep in range
Bounded by reach	The scale of its goals in space and time defines a Self.3 In Levin's examples a dog remembers far more than it can anticipate.5	How far back, ahead and out its concern runs
Nested and overlapping	Selves overlap with other Selves at the same, higher and lower levels,3 and subsystems have goals that are "not always aligned".5	Every agent that contains the same body
Mobile	Borders are porous and shift within a lifetime. They extend when parts couple and shrink when parts decouple, as when cancer cells leave the tissue network.3^,5	The date and the circumstances
Bound by stress	Stress is a systemic response to distance from a setpoint. Stress that spreads recruits other parts to the same goal and binds them into a higher Self, and the range of states that can stress a system indicates its sophistication.3	What can stress the candidate, and whose stress it shares
Posited	Outside observers posit a Self, and systems posit one about themselves. Estimates of agency depend on the observer and on the scale of observation, and experiment decides which model predicts and controls best.3	Who observes, with what instrument, at what scale

Aligned Definitions

Other accounts reach the same criterion from different starting points.

Source	What makes a unit	Convergence
Levin3	Goals that belong to the collective, at the scale of the light cone	The same criterion decides agency and individuality
McMillen and Levin4	Alignment of parts toward one target	Observers count one individual where the parts align
Queller and Strassmann6	Shared purpose, with "high cooperation and very low conflict" among the parts	Organismality is social and graded, and it needs no continuous body
Barandiaran, Di Paolo and Rohde7	An autonomous organisation that regulates its coupling with the environment and sustains itself as a consequence	Agents define themselves as individuals "as an ongoing endeavor"
Krakauer and colleagues8	Aggregates that propagate information from their past into their future	Individuality is a matter of degree, can nest, and can arise at any level
Pradeu9	Whatever the immune system includes and tolerates	Individuality depends on the question and comes in degrees
Gilbert, Sapp and Tauber10	The holobiont: host and symbionts that immunity integrates	Host and symbionts negotiate membership, with no self given in advance

Individuality as a Running Test

Cells, groups and observers construct individuality, and they never finish. It is relational, circumstantial and negotiable, and it spans levels of organisation. The organism is therefore one candidate self among the others (cf. Agent, Biological Individuality and Self). Philosophers of biology now largely agree on the same points: individuality depends on the question asked, nests across levels and comes in degrees, and "the level of the organism is not the only one at which biological individuality can be realized".9

A self keeps testing what belongs to it, and the testing never stops. Immunity gives the clearest case. An older view has the immune system defend a self that exists in advance. The current view adds "passport control" to the armed forces: immunity admits the symbionts that complete the animal and polices the cells that defect, in "a continuous negotiation" with the living surroundings inside and outside the body.10 Parents and groups run tests of the same kind on eggs, offspring and members.

The tests misfire, and they misfire in both directions. A cell that escapes immune control becomes a pathogen or a cancer.10 A cuckoo's egg passes as the host's own, and hosts that face good mimicry "mistakenly reject their own eggs".11 A young animal imprints on what it meets early and keeps that social preference in later life. Bias lists these cases as biases of self and non-self discrimination and of imprinting. Each misfire shows where a boundary currently runs.

Test	Who runs it	Admits too much	Rejects too much
Immune tolerance	Cells and tissues	A parasite or a tumour passes as self	The body attacks its own tissue
Kin and group recognition	Parents, mates, colonies	Hosts raise a brood parasite as offspring	A parent ejects its own egg
Attribution of agency	Researchers, carers, designers	An aggregate counts as an agent	A competent collective counts as an object

Three requirements follow for any claim about agency. The claim needs a level, because one body belongs to several selves at once. It needs a date and a setting, because boundaries move with season, life stage, isolation and shared stress. It needs an observer, because each observer registers some selves and misses others, and the self in question is one of those observers.

In Novel Ecosystems

There is significant deterioration of information webs and information use in the current, rapidly changing world.

Information is a crucial aspect of organims' niches (all life uses information to act and make decisions) so the impact is far-reaching and undermines the functioning of ecosystems.

There is an urgent need to conserve, restore, and rebuilt information webs.

Szymkowiak, Jakub, and Kenneth A. Schmidt. “Ecology of Information Enters the Anthropocene.” Oikos 2022, no. 10 (2022): e09677. https://doi.org/10.1111/oik.09677.

In Design

Some benefits of agential perspective in design:

Better for specifying action. Concrete and actionable in the local circumstances with local stakeholders is more compatible with the practical circumstances of design because the responsibility of designers is often to take concrete, local action now. Thus, the agential perspective might be more actionable than the population-level view.

Opens the possibilities for rapid more-than-human innovation. The ability of agents to adjust in response to the environment in non-genetic ways is also relevant for design because design situations require innovative responses in periods that are much shorter than those typical for innovations achieved via random mutation.

Establishes the possibility for multiple ways to transmit relevant Information. There are many senders/sources and many receivers. What they perceive and how it matters to them differs from who they are and what they do. This opens possibilities for multiple simultaneous design actions.

Declaring the Agent

Nature supplies no default level, so a design brief has to declare its agent. Five declarations condense the operational questions above. Agentic Stair states them for a check between two agents, and this section applies them to a brief.

Declaration	Question	Draws on
Boundary	Which collective shares the goal, and who is inside?	Alignment, candidate boundary
Variable	Which condition does it keep in range?	Controlled variable, setpoint, stress
Horizon	How far back, ahead and out does its concern reach?	Cognitive light cone
Perspective	Who observes it, with what instrument, at what scale?	Observer dependence
Beneficiaries and costs	Who gains and who pays when design serves it?	Whose viability is served

The choice of agent changes the brief. In animal housing, a brief for one animal would include refuge and refusal, and a brief for a population that moves between institutions would include transport and transfer (cf. Captivity, Housing).

Misfires of the tests above are design problems as well. Quarantine manages what immunity admits. A hand-reared animal may take its carer for kin. Residents may drive out a newcomer at an introduction.

Coordination can extend a light cone (cf. Design Potential (Private), Communication). By extension, and beyond what Levin claims, housing that lets carers register what the animals choose is a coordination device of that kind: animals and carers together can then pursue goals that neither pursues alone.

Levin also sets the empirical task: to fill in light-cone diagrams "with behavioral data from real organisms".5

In Education

“will and ability to positively influence [one’s own] life and the world around them” (OECD, 2019, p. 16; Organisation for Economic Cooperation and Development. OECD future of education and skills 2030)

References

Sharov, Alexei, and Morten Tønnessen. Semiotic Agency: Science beyond Mechanism. Cham: Springer, 2021.

Sharov, Alexei A. “Biology of Translation: The Role of Agents.” In Translation Beyond Translation Studies, edited by Kobus Marais, 63–80. London: Bloomsbury Academic, 2022.

Footnotes

Corning, Peter A., Stuart A. Kauffman, Denis Noble, James Alan Shapiro, Richard Irwin Vane-Wright, and Addy Pross, eds. Evolution “on Purpose”: Teleonomy in Living Systems. Cambridge, MA: MIT Press, 2023.˄
Dresow, Max, and Alan C. Love. “Teleonomy: Revisiting a Proposed Conceptual Replacement for Teleology.” Biological Theory 18, no. 2 (2023): 101–13. https://doi.org/10.1007/s13752-022-00424-y.˄
Levin, Michael. “Technological Approach to Mind Everywhere: An Experimentally-Grounded Framework for Understanding Diverse Bodies and Minds.” Frontiers in Systems Neuroscience 16 (2022): 768201. https://doi.org/10.3389/fnsys.2022.768201.˄
McMillen, Patrick, and Michael Levin. “Collective Intelligence: A Unifying Concept for Integrating Biology across Scales and Substrates.” Communications Biology 7, no. 1 (2024): 378. https://doi.org/10.1038/s42003-024-06037-4.˄
Levin, Michael. “The Computational Boundary of a ‘Self’: Developmental Bioelectricity Drives Multicellularity and Scale-Free Cognition.” Frontiers in Psychology 10 (2019): 2688. https://doi.org/10.3389/fpsyg.2019.02688.˄
Queller, David C., and Joan E. Strassmann. “Beyond Society: The Evolution of Organismality.” Philosophical Transactions of the Royal Society B: Biological Sciences 364, no. 1533 (2009): 3143–55. https://doi.org/10.1098/rstb.2009.0095.˄
Barandiaran, Xabier E., Ezequiel Di Paolo, and Marieke Rohde. “Defining Agency: Individuality, Normativity, Asymmetry, and Spatio-Temporality in Action.” Adaptive Behavior 17, no. 5 (2009): 367–86. https://doi.org/10.1177/1059712309343819.˄
Krakauer, David, Nils Bertschinger, Eckehard Olbrich, Jessica C. Flack, and Nihat Ay. “The Information Theory of Individuality.” Theory in Biosciences 139, no. 2 (2020): 209–23. https://doi.org/10.1007/s12064-020-00313-7.˄
Pradeu, Thomas. “The Many Faces of Biological Individuality.” Biology & Philosophy 31, no. 6 (2016): 761–73. https://doi.org/10.1007/s10539-016-9553-z.˄
Gilbert, Scott F., Jan Sapp, and Alfred I. Tauber. “A Symbiotic View of Life: We Have Never Been Individuals.” The Quarterly Review of Biology 87, no. 4 (2012): 325–41. https://doi.org/10.1086/668166.˄
Kilner, Rebecca M., and Naomi E. Langmore. “Cuckoos versus Hosts in Insects and Birds: Adaptations, Counter-Adaptations and Outcomes.” Biological Reviews 86, no. 4 (2011): 836–52. https://doi.org/10.1111/j.1469-185x.2010.00173.x.˄

Backlinks