Thick Ethical Concepts
Thick ethical concepts carry evaluative and descriptive content simultaneously. To call an action courageous is not merely to approve of it but to characterize it as involving a specific quality — the relationship between fear and resolve — that looks different from other forms of goodness. Thin concepts (good, bad, right, wrong) carry evaluative content but almost no descriptive content; they issue verdicts without characterizing. Williams argued that thick concepts are the primary vocabulary of actual moral life, that they emerge from specific communities of practice, and that the morality system's preference for thin universal categories produces systematic blindness to the textured distinctions thick concepts preserve. The AI discourse, conducted almost entirely in thin categorical vocabulary, exemplifies the impoverishment.
In the AI Story
Williams developed the distinction in Ethics and the Limits of Philosophy (1985), though the idea draws on earlier work by Philippa Foot and Iris Murdoch. The argument was part of his broader critique of moral theorizing: systematic ethics demands vocabulary general enough to be universally applicable, and the demand drives it toward thin concepts. But thin concepts cannot register the textured distinctions that make one form of goodness different from another. A situation described only as 'good' or 'bad' has been flattened beyond the point where adequate response is possible.
The AI transition is conducted almost entirely in thin concepts. The triumphalist says AI is good, disruption is right, adaptation is required. The elegist says depth is being lost, the transition is wrong, something valuable must be preserved. These verdicts are categorical but empty of texture. Thick concepts would differentiate: is the engineer's embrace of AI courageous (engaging with risk after honest assessment), reckless (abandoning practices without reflection), or desperate (driven by fear of obsolescence)? Is the practitioner's refusal principled, stubborn, or terrified? The thin verdict collapses these distinctions; the thick concepts preserve them.
Williams's deepest point is that thick concepts are not invented in seminar rooms. They develop in communities of practice — groups sharing a form of life who need vocabulary adequate to their specific demands. The programming community's concept of 'elegant code' is a functioning thick concept: it carries both description of a specific quality (economy, clarity, inevitability) and evaluation of that quality as admirable. When AI disrupts the programming community, it disrupts not only the practice but the evaluative vocabulary the community developed — and the vocabulary's collapse is a moral loss the productivity discourse cannot register.
The practical consequence is that the AI conversation needs more words, not fewer. More specific, more textured vocabulary for the different kinds of engagement, loss, and adaptation the transition produces. The words will develop not because philosophers prescribe them but because practitioners need them. The risk is that the transition will move faster than thick-concept formation can keep up, and the moral processing will be completed in thin categories before the vocabulary adequate to describe what actually happened has time to emerge.
Origin
The thick/thin distinction appears most fully in chapters 7 and 8 of Ethics and the Limits of Philosophy (1985). Williams credited the concept partially to earlier work by Philippa Foot and Iris Murdoch, and to Gilbert Ryle's distinction between thick and thin description in ethnography. The concept has since been developed by Sabina Lovibond, Jonathan Dancy, and others, and has become one of the most productive pieces of metaethical apparatus in contemporary moral philosophy.
Key Ideas
Fused content. Thick concepts carry descriptive and evaluative content inseparably; the description cannot be disentangled from the evaluation without losing what makes the concept function.
Community-specific. Thick concepts emerge from particular forms of life and depend for their meaning on the practices embedded in those lives.
Resist universalization. Because thick concepts carry practice-specific descriptive content, they cannot be rendered universal without being thinned out into categorical verdicts.
Guide response. The practical value of thick concepts is that they tell an agent not only how to evaluate a situation but what kind of situation it is, which is prior to and more important than the evaluation.
Vulnerable to practice-disruption. When a technology transforms the practices in which thick concepts are embedded, the concepts themselves can lose their substrate and decay into thin verdicts.
Debates & Critiques
The status of thick concepts remains contested in metaethics. 'Disentangling' positions (Simon Blackburn, Allan Gibbard) argue that thick concepts can be analyzed into descriptive and evaluative components, challenging Williams's claim of inseparability. 'Non-disentangling' positions (Jonathan Dancy, John McDowell) defend the view that the fusion is philosophically fundamental. The AI application has so far been developed most systematically by Vallor, whose technomoral virtues framework extends Williams's concept.
Appears in the Orange Pill Cycle
Further reading
- Bernard Williams, Ethics and the Limits of Philosophy (Harvard, 1985), ch. 7–8
- Simon Kirchin (ed.), Thick Concepts (Oxford, 2013)
- Jonathan Dancy, 'In Defense of Thick Concepts' (Midwest Studies in Philosophy, 1995)
- Sabina Lovibond, Ethical Formation (Harvard, 2002)
- Debbie Roberts, 'Thick Concepts' (Philosophy Compass, 2013)