Internal and External Reasons

In the AI Story

The internal/external distinction threatens more of ethics than it initially appears to. If reasons must connect with existing motivational sets, then moral demands claiming universal rational force have a much harder argument to make than they typically assume. A person cannot be rationally required to act against everything she cares about on the grounds that the action would be 'objectively right' — the rightness, if it exists, must find purchase in her existing commitments or fail to bind her.

Williams qualified the thesis carefully. The motivational set is not static; it can be extended through reflection, imagination, and new experience. A reason can be internal without being immediately obvious to the agent — it may take work to discover that she has reason to act in a given way. But the work is discovery of something already latent in her motivational set, not introduction of something foreign to it.

The AI application is immediate. The triumphalist tells the senior practitioner she should embrace AI because it is more efficient, the market demands it, the aggregate good is served. These reasons are external to the practitioner's motivational set. Her motivations include craft, depth, the satisfaction of mastery, the community constituted by shared standards. 'More efficient' is not in her motivational set in the relevant sense. The argument fails to move her because it cannot.

The Orange Pill's concept of ascending friction functions as an internal-reason generator. It argues that AI does not eliminate difficulty but relocates it to a higher level — the level of judgment, taste, strategic vision. This connects with the practitioner's existing commitment to difficulty as a source of meaning. The argument is internal: it shows how the new landscape serves what she already cares about. The difference between 'you should embrace this because the market demands it' and 'the thing you've always wanted to build, you can build now' is the difference between an external reason that produces resentment and an internal reason that produces engagement.

Origin

The distinction was introduced in Williams's 1979 essay 'Internal and External Reasons' (collected in Moral Luck, 1981). Williams elaborated and defended the thesis in 'Internal Reasons and the Obscurity of Blame' (1989, in Making Sense of Humanity, 1995) and revisited it in later essays. The thesis sparked one of the central debates of late-twentieth-century metaethics, drawing responses from John McDowell, Christine Korsgaard, Derek Parfit, and T.M. Scanlon.

Key Ideas

Motivational grounding required. A reason binds an agent only if it connects with her existing motivational set — the network of desires, commitments, and evaluative dispositions that constitute her practical orientation.

External reasons are illusory. A 'reason' that fails to connect with any element of the motivational set is not a reason the agent has failed to recognize but not a reason for her at all.

Sets are extensible but not arbitrary. The motivational set can be extended through reflection and experience, but the extension must proceed from what is already there.

Internal reasons can be latent. Finding an agent's internal reason may require imaginative work; the reason can be genuine without being obvious.

Explains the failure of triumphalist persuasion. The AI transition's advocates rely on external reasons (efficiency, market demand, inevitability) that cannot move practitioners whose motivational sets are organized around different values.

Debates & Critiques

The internal/external reasons debate is one of the most developed in contemporary metaethics. John McDowell argued that sufficiently idealized reflection can yield reasons Williams would have to count as external. Christine Korsgaard developed a Kantian alternative grounding reasons in practical identity. Derek Parfit's On What Matters represents the most sustained recent defense of external reasons. The application to technology ethics — showing why top-down arguments for adoption fail to move practitioners — has been developed by Shannon Vallor and others.