Embedded Files
Evolution of Agency:
Michael Tomasello's Evolution of Agency identifies four main types of natural agency and describes them in evolutionary order of emergence. He concludes that agency is "the most generalized organizational framework within which individuals formulate and produce their actions."
1. Introduction
2. Feedback Control Model of Agency
3. Ancient Vertebrates as Goal-Directed Agents
4. Ancient Mammals as Intentional Agents
5. Ancient Apes as Rational Agents
6.1 Ancient Humans as Socially Normative Agents
6.2 Modern Human Collective Agency in Cultural Groups
7. Agency as Behavioral Organization
Excerpts from Evolution of Agency:
Excerpts from Evolution of Agency:
1. Introduction
The issue is not complexity but control.
Agency empowers individual “in some key subset of situations” to decide for itself what to do according to its own best judgment.
Types of psychological agency in evolutionary order of emergence:
goal-directed agency in ancient vertebrates,
intentional agency in ancient mammals,
rational agency in ancient great apes, and
socially normative agency in ancient human beings.
agency is the most generalized organizational framework within which individuals formulate and produce their actions
2. Feedback Control Model of Agency
Agent does not just respond to stimuli but actively directs (or even plans) its actions toward goals
main evidence for agency is behavioral flexibility of agent, especially in novel circumstances
The machine or organism acts so as to bring or keep its perceptions in line with its reference values, which in some cases means behaving and in other cases means doing nothing.
Basic structure of psychological agency is manifest in cybernetic models of goal-directed action based on feedback control (e.g., Miller, Galanter, & Pribram, 1960)
Plan-And-Structure-Of-Behavior, Miller, 1960
TOTE model - like feedback control; cybernetics
Intent and the execution of plans
Therein resides a crucial difference between a chain of actions and a Plan of action.
When a chain is initiated with no internal representation of the complete course of action, the later parts of the chain are not intended.
When a Plan is initiated, the intent to execute the later parts of it is clear
All autonomous, intelligent machines have circular causal organization: action causes change in perception, which is then compared to the goal to determine if further action is needed.
Heterarchy of interrelated feedback control systems (Bechtel & Bich, 2021
feedback control model components:
(i) goal,
(ii) a sensing device or perception, and
(iii) a device for comparing perception and goal
(iv) execute a behavioral decision.
Tomasello: Building a structure of psychological agency,
1. Cybernetics, 2. Decisions under uncertainty, 3. Executive control, 4. 2nd order executive using “computational rationality”, 5. Socially share agencies
Figure 2.3 Evolutionary tree locating the species on which I focus
3 Ancient Vertebrates as Goal-Directed Agents
First organisms on Earth were not agents.
They were unicellular organisms that simply moved around with open mouths.
An organism’s action capabilities determine its experiential world.
Attention is goal-directed perception.
Goals are perceptions of the world - perceptually imagined situations - that the organism desires or is motivated to bring about. It then behaves until it perceives the realization of those desired situations in the actual world (Powers, 1973).
Figure 3.4 The organization of feedback control systems for goal-directed agents.
4. Ancient Mammals as Intentional Agents
“Pursuance of future ends and the choice of means for their attainment are . . . the mark and criterion of the presence of mentality in phenomena.” - William James, The Principles of Psychology
Mammals direct their actions toward goals not just flexibly but intentionally, as they cognitively simulate possible action plans toward their goal before actually acting.
intentional agency: not just an operational tier of perception and action - as in goal-directed agents - but also an executive tier of decision-making and cognitive control.
executive tier: enables more flexible forms of planning and decision-making that output not an action but an intention to act,
executive function: also referred to as cognitive control; a widely accepted typology:
(i) inhibition (including inhibitory control, self-control, and behavioral inhibition, as well as interference control, selective attention, and cognitive inhibition);
(ii) working memory (i.e., holding information in mind and mentally working with it in various ways); and
(iii) cognitive flexibility (e.g., set shifting, mental flexibility, or mental set shifting, closely linked to creativity) (paraphrased from Diamond, 2013).
An executive tier of psychological monitoring and control
which is itself a feedback control system
with new forms of decision-making and behavior monitoring
that facilitate the organism in directing and controlling its actions.
intentional action (Bratman, 1987) requires individuals to cognitively simulate in an organized way their own potential actions, the potential obstacles and opportunities for those actions, and the probable outcomes of those actions.
Figure 4.2 The organization of mammal’s intentional agency.
Planning: process of cognitively simulating possible actions and their outcomes.
Tolman called this behavior observed in rats: Vicarious, Trial and Error.
Tolman, E. C. (1948) Cognitive maps in rats and men
VTE: hesitating, looking back-and-forth, behavior before choosing
squirrels and rats simultaneously consider two cognitively simulated behavioral options in acts of either-or decision-making.
A role for executive tier is curiosity and exploration.
Conscious experience exists only in creature with an executive tier of functioning
Consciousness is about executive-tier attention to the perception-action tier
essential points
(i) basic sentience in the sense of attention to, and experience of, the outside world is for agents a psychological primitive; and
(ii) basic consciousness involves the organism attending to its own goals, actions, and experience from its executive tier of functioning.
Mammals and other intentional agents are conscious in this sense.
5 Ancient Apes as Rational Agents
Reflective, Second-order executive tier of functioning,
individuals monitor and evaluate their first-order executive decision-making and cognitive control.
ability to attribute mental states to others; and ability to compare
Understanding causality and intentionality
expands the field of agentive action to include underlying causes of those events and
actions which can be manipulated to produce desired effects.
chimpanzees inferred gravity
that when one end of a balance beam tilted down, it meant that the opaque cup on its end contained a banana indicating an understanding of gravity
Causal learning: learning from causal and intentional relations among external entities.
Rational imitation: social learner is comparing its process of decision-making to another agent
not only executively monitor uncertainty but also reflectively monitor and control the decision-making process itself.
diagnosed that missing some information and determine how to alleviate ignorance.
computational rationality: decide if potentially available information is worth the effort needed to gather it.
Figure 5.2 Great ape’s rational decision-making via two tiers of executive control.
Piaget’s (1952): the bridge is none other than the use of tools
To use a tool flexibly and reliably there must be an integration of the movement of the tool, as caused by the agent, and the properties of the tool in relation to the substrate.
agentive organization characteristic of great apes
emergence of a second-order tier of executive decision-making and control –
led to the formation of an experiential niche structured by
the causes underlying physical events and
the intentions underlying agentive action,
both organized into similar logical-inferential paradigms,
enabling individuals to imagine causally and intentionally structured states of the world that are not directly perceived.
criteria for rationality:
(i) thinking about the external world using logically structured causal and intentional inferences, providing rational coherence to experience; and
(ii) adopting a reflective and self-critical stance to one’s own thinking and decision-making, including adjudicating between conflicting goals before acting by reflecting on their relative merits, providing rational coherence to one’s psychological functioning in general.
6.1 Ancient Humans as Socially Normative Agents
Early human individuals formed with other individuals a joint agency.
socially normative self-regulation
obliges individuals to direct and control their actions not just individually but also to comport with the normative standards of the shared agency in which they are participating.
Individuals acting in shared agencies are socially normative agents.
individuals came together to form socially shared agencies - socially constituted feedback control systems - that could pursue shared goals that no individual could attain on its own.
Forming a Joint Goal
cooperative cognition, or common ground, of the participants that they both know together that they both want to pursue the goal together:
they have formed a joint agency to pursue a joint goal.
individual’s ability and propensity to form with others a joint goal, thereby creating an evolutionarily unique, socially constituted feedback control system.
Coordinating Roles
Collaborative partners informed one another of things helpfully so as to facilitate their joint success
species-unique cognitive skills: including, most importantly, perspective taking and cooperative communication;
to mentally plan and coordinate joint agencies working toward joint goals.
Collaboratively Self-Regulating the Collaboration
two main mechanisms by which individuals can procure better collaborative partners are partner choice and partner control.
early humans attempted to control their collaborative partners with communicative protest.
normative protest often works to shift the behavior of the transgressor in a more cooperative direction.
In the context of the collaboration, "we" always has the last word.
The result was a kind of we > me sociomoral self-regulation,
each partner internalized a responsibility to play her role in the joint agency -
the individual used the joint agency to self-regulate her individual behavior -
in a way that comported with their common-ground normative standards.
Cooperative Rationality and Its Experiential Niche
Joint Agency:
three tiers of agentive functioning - operational, executive, and reflective
cooperativized: with the word joint in front of each component:
joint goal, joint attention, joint decision, joint action
Figure 6.2 shows the mode of joint agency in the middle, in a box with borders.
Cooperative Rationality - they did what made sense in the context of their collaboratively structured agency - and this required them to juggle simultaneously the operations and co-operations of three distinct but interrelated modes of agency.
human individuals closest to those with whom they shared the most experiences; derived through joint attention and common ground, in collaborative activities.
Compare with feminist AI author’s first point
Shared worlds experienced via recursive perspectives among mutually respectful and responsible cooperative agents: this is the new experiential niche inhabited by early humans.
6.2 Modern Human Collective Agency in Cultural Groups
emergence of distinct cultural groups that distinguished themselves from one another - 150,000 years ago
humans scaled up beyond the foraging pair to the social group at large; new skills of collective intentionality.
new form of collective foraging known as central-place foraging.
humans can only cooperate effectively on a personal level in groups up to 150 individuals
Is there any limit for AI groups?
competition between human groups was main driver of ever greater in-group cooperation.
Forming Collective Goals
Collective goals: destination of group travel, location of a home base, preparations for group defense, division of resources, and division of labor
in-group/ out-group psychology.
In-group favoritism accompanied by out-group mistrust is
one of the most well-documented phenomena in all of social psychology,
Homophily: the tendency to affiliate, favor, and bond with similar others, as the psychological basis of human culture.
human’s collective agency made possible by individuals evolving a group-minded concern for the culture’s goals and welfare
societal-level roles
collective agent whose individual members survive and thrive only if everyone does his or her job.
cooperation becomes more difficult as group size increases.
social norms: collective expectations for individual behavior served as self-regulators.
humans enforce social norms because they implicitly understand them as the means by which the group regulates itself.
the force of social norms comes from the social agreements that created those norms.
We collectively self-regulate everyone, including myself.
modern human individuals were operating simultaneously with three modes of agency:
I: an individual pursuing her own self-interest,
We: a collective agent operating via the group’s collective practices and norms, and
Me: a role agent performing the duties that the cultural group oblige me to.
objective-normative world.
shared agencies comprise both shared experience on a common focus along with different perspectives.
adults are not giving to children their personal preferences, but rather informing them about objectively correct ways to do things; and
children are predisposed to understand social norms transmitted in this way to be about the objective world
Normative rationality thus means adapting one’s individual agency to “objective” facts and values as they inhere in collective cultural experiences.
Modern human agency thus operates in a world of objective facts and objective moral values.
individuals extend this objectivity to their social-institutional worlds to create what John Searle (1995) calls social facts or institutional reality.
person or object acquires status based on deontic powers given by the group
via objectified agreement and so becomes part of external reality.
deontic: relating to duty and obligation as ethical concepts.
a reality made of fictional entities, like nations, like gods, like money, like corporations.
Complexities of Human Agency
Individuals who self-regulate their thoughts and actions using “objective” normative standards are thereby normative agents,
very likely characterized by a new form of socially perspectivized consciousness, what we might call self-consciousness.
hree different feedback control systems,
individual self-regulation,
joint self-regulation via normative protest, and
collective self-regulation via the cultural group's social norms.
Modern humans are individual rational agents who sometimes (though not always) subordinate their individual agencies to various shared agencies when doing so is either instrumentally or normatively appropriate
human normative agency is both liberating and constraining
role of coming to maturity in the midst of other persons
with whom we are vitally interdependent, both cooperatively and culturally.
7 Agency as Behavioral Organization
The issue is not complexity but control.
organisms are able to flexibly direct and control their actions if, and only if, their underlying psychology is organized agentively, in the manner of a feedback control system.
1. The “backbone” of behavioral agency is feedback control organization
when humans attempt to build a machine that acts autonomously, intelligently, and flexibly in the face of unpredictable ecological challenges,
pretty much the only organizational architecture used is feedback control organization.
models in artificial life, robotics, computational modeling and the philosophy of action and agency, have this same basic architecture:
the agent has goals or values,
perceptually attends to situations relevant to those goals or values,
makes behavioral decisions (and so acts) in light of those goals or values and relevant situations, and
observes its actions and their results to make ongoing adjustments as needed.
2. The ecological challenges leading to the evolution of behavioral agency all involve one or another form of unpredictability in the environment
when Nature cannot predict important future contingencies in the environment, her solution is to equip the individual to pursue certain goals flexibly by assessing the immediate situations and then choosing the best thing to do (see Veissière et al., 2019).
the most important cause of decision-making uncertainties for agentive organisms is other creatures.
certain types of ecological challenges create certain types of uncertainties in the decision-making individual, which lead to certain types of agentive behavioral organization
3. basic types of psychological Bauplans for the agentive organization of behavior
goal-directed agents
intentional agents
rational agents
normative agents
Body plan - Wikipedia Bauplans
Table 7.1 – Typology of different types of agency and their key characteristics
4. behavioral organization involves “hierarchical modularity” and “trickle-down selection”
hierarchical structure is necessary to explain the evolution of the behavioral flexibility and underlying agentive organization involved.
5. types of experience it is capable of having (its experiential niche)
different ecological and experiential niches depending on what they need to do to survive and thrive.
Each organism lives in its own ecological and experiential niche, as determined by its behavioral capabilities.
Nature selects most directly for adaptive actions, and this drives everything in the evolution of an organism's psychology, including its experiential niche.
6. The decision-making agent is necessary, and it is not a homunculus, at least not in a bad way
models of cognitive science built on the computer metaphor or neuroscientific explanations often do not have room for a decision-making agent.
what lies behind the individual’s production of agentive actions is not a new entity such as a homunculus but a particular kind of psychological organization in which a living individual attends to goal-relevant situations, makes decisions, and self-regulates the process
to explain how organisms behave flexibly and efficiently in novel environmental circumstances, we need a new behavioral principle, and
that new behavioral principle is an underlying psychological organization of agency based on principles of feedback control.
It is one of the great discoveries of the twentieth century that machines made exclusively of nonliving components can produce actions that are, in many important respects, agent-like.
This raises the difficult question of whether machines are, or could be, actual agents.
My own view is that what most clearly differentiates agentive organisms from behaving machines, as they are currently configured, is the way that living agents flexibly attend to relevant situations—opportunities and obstacles to their goals and values
perception and cognition are integrally fused with their goal-directed actions.
Nature has constructed an underlying psychological organization of agency enabling the individual to make its own decisions and self-regulate its own actions in pursuit of goals that, ultimately, Nature has built in.
Through normal processes of evolution by means of natural selection, Nature has crafted forms of agentive organization that empower individuals to act autonomously.
Selected References
Bechtel, W., & Bich, L. (2021). Grounding cognition: Heterarchical control mechanisms in biology. Philosophical Transactions of the Royal Society B: Biological Sciences, 376, 20190751.
Bonner, J. T. (1988). The evolution of complexity by means of natural selection. Princeton, NJ: Princeton University Press.
Dewey, J. (1896). The reflex arc concept in psychology. Psychological Review, 3(4), 357–370.
Juechems, K., & Summerfield, C. (2019). Where does value come from? Trends in Cognitive Sciences, 23(10), 836–850.
Lea, S. E., Chow, P. K., Leaver, L. A., & McLaren, I. P. (2020). Behavioral flexibility: A review, a model, and some exploratory tests. Learning and Behavior, 48, 1–15.
Miller, G. A., Galanter, E., & Pribram, K. H. (1960). Plans and the structure of behavior. New York: Holt.
Powers, W. (1973). Behavior: The control of perception. Chicago: Aldine.
Searle, J. (1995). The construction of social reality. New York: Free Press.
Veissière, S., Constant, A., Ramstead, M., Friston, K., & Kirmayer, L. (2019). Thinking through other minds: A variational approach to cognition and culture. Behavioral and Brain Sciences, 43(e90), 1–75.
Page updated
Google Sites
Report abuse