The rectification of names (zhengming) is Confucius's doctrine that accurate language is the precondition for right action. 'If names are not correct, then language does not accord with reality. If language does not accord with reality, then things cannot be accomplished properly.' The teaching is not linguistic but structural: a civilization built on incorrect names will produce deformed institutions, because the institutions will be designed to address things as they are named rather than as they are. In the AI discourse, zhengming exposes the hollow language the industry deploys — calling intensification 'productivity,' calling expanded capability within concentrated power 'democratization,' calling pattern-matching 'intelligence,' calling destruction 'disruption.' Each corrupted name conceals a political question the correct name would force.
'Productivity' as deployed in the AI discourse describes volume of output without reference to value, cost to the producer, or sustainability. The Berkeley study found that AI did not reduce work — it intensified it. If productivity means the ratio of valuable output to sustainable input, what the researchers documented was not productivity but intensification. The distinction is morally consequential: organizations celebrating 'productivity gains' while actually rewarding compulsive intensification build incentive structures that deplete the workers the structures purport to serve, and the workers inhabiting the deformed structures experience the deformation as personal failure.
'Democratization' names the expansion of capability — the developer in Lagos who can now prototype, the non-technical founder who can now build. But democratization, correctly used, means the distribution of power, not merely of capability. The capability has expanded; the power — over platforms, distribution, capital, terms of service — remains concentrated. Calling the process 'democratization' forecloses the political inquiry the correct name would demand: who controls the infrastructure through which the expanded capability operates?
'Intelligence,' applied to both human consciousness and machine computation, creates an implicit equivalence that structures the entire discourse around competition. Human intelligence involves moral discernment — the capacity to perceive right action in particular situations, the judgment emerging from cultivated character. Machine computation involves pattern-matching at scale. These are different phenomena producing superficially similar outputs. Using the same name for both generates the existential panic The Orange Pill documents: if the machine is intelligent and faster, what is the human for?
'Disruption' has acquired a positive valence it does not deserve. To disrupt is to break — and the celebration of breaking conceals the cost of the breaking: the communities disrupted, the skills rendered obsolete, the relationships severed. The correct name would demand accounting. 'We disrupted an industry' would become 'We destroyed an existing structure of livelihoods and replaced it with one that serves our interests. Whether it serves the interests of the people who inhabited the previous structure is a question we have not answered.'
The doctrine of rectifying names appears in Analects 13.3, where Confucius answers the Duke of Wei's question about what he would do first if entrusted with government. 'Certainly it would be the rectification of names.' The passage has been interpreted throughout Chinese intellectual history as a foundational claim about the relationship between language, ethics, and political order.
Contemporary AI ethics scholarship has drawn on zhengming implicitly in debates over the language of 'intelligence,' 'alignment,' and 'democratization' — with researchers including Emily Bender, Timnit Gebru, and Jathan Sadowski documenting how imprecise naming structurally produces imprecise policy.
Names carry obligations. The person who calls herself a builder is obligated to build well. The person who calls her product 'democratization' is obligated to ensure that power, not merely capability, has been distributed.
Corrupted names conceal political questions. Each misnomer in the AI discourse — productivity, democratization, intelligence, disruption — forecloses inquiry that the correct name would have demanded.
Structures built on incorrect names deform lives. The worker in the productivity-celebrating organization experiences her depletion as personal failure when it is structural consequence of a linguistic error at the foundation.
Rectification is continuous. The discipline is not a one-time correction but ongoing moral hygiene, practiced at every level where the comfortable wrong word threatens to conceal the uncomfortable truth.
The junzi rectifies names as a moral duty. Precision is not scholarly preference but the precondition for structures that serve life rather than extraction.
Some have argued that language evolves and that insisting on 'correct' usage is itself a form of ideological control. The Confucian response distinguishes descriptive evolution (legitimate) from strategic corruption (illegitimate) — the latter being the deliberate deployment of emotionally resonant terms to foreclose political analysis.