Agency

Agency is the ability and freedom for an individual to act in their immediate context or environment.

See also: Tools for Conviviality, agentic computing

Self-Determination Theory (SDT)

When our social environments, including the places where we receive health care, are more supportive of these psychological needs, the quality of our motivation is more autonomous. Alternatively, when our psychological needs are not well met or even thwarted through our social interactions, the quality of our motivation is more controlled

Self-determination theory suggests that all humans have three basic psychological needs, which when met, help us be more intrinsically motivated in our actions (more internal locus of control)

1. Autonomy: the feeling one has choice and agency in their own lives
2. Competence: experience of mastery and being effective in one’s activity
3. Relatedness: need to feel connected and belongingness with others

See also: In Over Our Heads, burnout, taste

In software

4 Principles, Brooklyn Zelenka at Causal Islands

1. Empower users to participate (entry)
2. Option to leave (exit)
3. Control access to your data (safety)
4. Provide capacity to others (serve)

Mechanistic vs Volitional Agency

Source

”When we say that something is political, we mean that it involves the contestation of competing (human) wills—that is, of possibly-conflicting goals and intentions, each of which ought to be respected.”

Definitions:

• An Agent as defined by List and Pettit [2011]
  1. Have representational states that depict how things are in the environment
  2. Have motivational states that specify how it requires things to be in the environment
  3. Have the capacity to process its representational and motivational states, leading it to intervene suitably in the environment whenever that environment fails to match a motivating specification.
• Functional agency: It is common in machine learning literature to refer to AI (systems) as agents with little need for additional characterization (e.g. “multi-agent reinforcement learning,” “dialogue agent,” etc.). In these cases, agent is primarily meant as shorthand to capture particular functionality of the algorithmic object. (term mine, definition attributed to author)
• Agency in the lens of how ones “moral character” might be evaluated
  • Mechanistic agency: agency is defined primarily by the ability to take action in the world given some information
    • It draws from the Platonic claim that virtue—that is, ethical behavior—is knowledge about what is good
    • Ethical missteps are then epistemic failings, and it follows that more information
    • However, an agent is a moral agent in the mechanistic sense if and only if it can be held responsible. In examining the backwards-looking question, we argue that AI systems fail this standard, meaning that they cannot be understood as moral agents, even mechanistically
  • Volitional agency: agency is defined by actively making decisions in accordance with an internal desires
    • Extends Aristotle’s argument that virtue is an active practice towards becoming a certain “kind of” person.
• To “hold responsible” implies the existence of some action, such as a penalty, that is taken with the goal of shaping the agent’s future behavior. Yet such an action may not exist
• Q: can decisions made by purely stochastic processes be considered normatively significant? If I flip a coin to make a decision for me to determine whether to commit some action, can I be held responsible?
  • See: intentionality