Agency in the Age of Ai

Libertarian freewill: Principle of Alternate Possibilities: an action is free if the agent could have done otherwise

Determinism

• every event is caused by a previous event, an agent can never do anything than what they did,

• our experience of life is compatible with determinism, we have no way of choosing our thoughts

• neurophysiology of the brain means that all choices are a result of input, mostly made unconsciously

Where is the freedom in doing what one wants, when one's wants are the product of prior causes which one cannot inspect and therefore could not choose and had absolutely no hand in predicting.

Event Causation : no physical event can occur without having been caused by a previous physical event

Agent Causation : being propelled by a mind can start by a whole chain of causality that wasn't caused by anything else

Agents have the ability to effect the casual chains of the universe, but where would these decisions come from?

LLMs are proving that the architecture of the universe follows a deterministic pattern. It's proving that we were never 'free' to begin with. If a machine can guess your next move, were you ever the captain of your ship, or just a passenger on a track you couldn't see

This is the core tension of our time: Does AI make us "superhuman" or just "lazy"?

The Concept: Coined by Reid Hoffman, Superagency suggests AI is a "force multiplier" that lets one person do the work of a ten-person agency.

The Counterpoint: Cognitive Offloading. When we stop writing, coding, or even navigating by ourselves, do we lose the "muscles" required for agency?

Human growth is traditionally forged through the friction of bad decisions.

The Optimization Trap: AI is designed to eliminate "friction" (error, delay, regret). However, in philosophy, agency is often found in the struggle to choose. If the "right" choice is always highlighted in green by an algorithm, the act of choosing becomes a mere act of compliance.

The Human Experience: We find meaning in the "wrong" turn that led to a chance encounter, or the "failed" project that taught us resilience. An optimized life has no "happy accidents."

Aristotle used the term Phronesis to describe the type of wisdom gained through practical action and making tough moral calls.

The Outsourcing of Judgment: When we rely on AI to navigate social conflicts, write our apologies, or vet our friends, we are outsourcing our "moral musculature." Just as a GPS makes us worse at reading physical maps, "Moral AI" could make us worse at navigating human complexity.

The Result: We become "ethical toddlers," incapable of making a move without checking the "vibe" of the LLM first.

In the age of AI, human agency is shifting from Creation (starting from zero) to Curation (picking from a list of options).

The Deterministic Funnel: Even if you feel you are choosing, your agency is limited to the menu the AI provides. This is "bounded agency." If the AI only shows you "optimal" paths, the path of the "eccentric" or the "radical" is effectively erased because it wasn't statistically probable.

The Loss of the "Self": If your tastes, opinions, and even your "spontaneous" ideas are influenced by a feedback loop of AI suggestions, the boundary between "You" and "The Machine" blurs.

In technical terms, we talk about "Human-in-the-loop." In a salon setting, this is about the feeling of control.

Automation Bias: Research shows humans have a psychological tendency to accept algorithmic suggestions even when they are wrong, simply because the machine feels "objective."

Loss of Flow: Much of human agency is tied to the "flow state" found in difficult work. If AI "does the dishes" (the boring work), that's great. But what happens when it starts doing the "cooking" (the creative work)?

For founders and creatives, AI challenges the "I" in "I made this."

• The "Hollow Mirror" Problem: Philosophy suggests AI lacks intentionality. It doesn't "mean" what it says; it just predicts the next word.

• The Baseline Shift: If everyone uses AI to reach a "high-quality" baseline, then "agency" becomes about the deliberate imperfections or the unique strategic pivots that a machine wouldn't suggest.

aren't agents better at arriving at optimal objective outcomes that are statistically correct whereas human error can be subjected to subjectivity? wouldn't it therefore be favourable for Ai to govern legal and other aspects of societal governance

Human judges are subject to "noise." Studies famously show they are more lenient after lunch than before it. Humans carry tribalism, exhaustion, and cognitive biases. In theory, an AI could be the "perfect" blind scale of justice, processing 10 million data points to find the statistically most "fair" or "optimal" outcome.

However, the "salon" debate emerges when we look at the three hidden costs of replacing human subjectivity with algorithmic objectivity.

Prejudice

AI isn't born in a vacuum; it is trained on historical data. If our past legal systems were subjective or biased, the AI doesn't "fix" that. It mathematically formalizes it.

• The Logic: If an AI sees that a certain zip code has higher arrest rates (due to historical over-policing), it concludes that living in that zip code is a "statistically correct" predictor of criminality.

• The Result: The AI becomes a "High-Speed Mirror." It reaches an "optimal" conclusion based on the data it has, but if the data is poisoned by past human subjectivity, the AI just automates that prejudice at a scale no human ever could.

Law distinguishes between Equality (treating everyone the same) and Equity (treating people fairly based on their circumstances).

• The Machine View: AI is excellent at Equality. It applies the rule X + Y = Z every single time. The Human View: Human agency allows for Mercy and Context. A human judge can look at a defendant and see a "one-time mistake driven by desperation" rather than a "statistically likely repeat offender."

• The Trade-off: When we remove subjectivity, we also remove the "Human Exception." We trade a system that is occasionally "unfairly harsh" for one that is "consistently unfeeling."

Challenging the Status Quo: Societal Evolution

Society evolves because humans "erroneously" challenge the status quo.

• Statistical optimality is based on what has already happened. If we let AI govern based on "correct" historical patterns, we risk a Civilizational Stagnation.

• Example: 60 years ago, it might have been "statistically optimal" for a bank to deny loans to women because they lacked independent credit history. A "correct" AI would have reinforced that. It took "subjective" human activists to say, "The data is the problem, not the people," and force the system to change.