Purpose Criterion
Purpose is the criterion that extends Krauss's formal analysis into the domain of ethics and social consequence. Krauss's method is, by design, formal rather than ethical—she analyzes structural positions and conditions of production, not moral implications of what is produced. But AI output has consequences that formal analysis can identify but cannot, by itself, evaluate. Generated text circulating as journalism, generated images functioning as evidence, generated code operating critical infrastructure—these outputs affect people, communities, and institutions, and the adequacy of the output to its purpose cannot be assessed through formal properties alone. The purpose criterion asks: What does this output serve? Whom does it serve? Does it serve them well—not merely efficiently (the performativity criterion Lyotard identified as the postmodern default) but adequately, in a sense that includes care for affected populations that performativity cannot capture? Segal's question "Are you worth amplifying?" is a purpose question. It asks not whether amplification is possible but whether what is being amplified deserves the amplification it receives.
In the AI Story
The criterion is necessary because the expanded field of AI production includes positions whose outputs have redistributive effects, power asymmetries, and long-term consequences that purely formal evaluation cannot address. When algorithmic systems make decisions affecting employment, credit, criminal justice—the formal adequacy of the algorithm (does it minimize classification error?) is insufficient. The ethical question (does it reproduce structural injustice?) and the political question (who benefits from its deployment?) cannot be answered by formal analysis alone. Krauss's method identifies what the system does; purpose evaluation asks whether it should.
The art-world precedent is limited but instructive. When Hans Haacke produced works investigating corporate patronage and real-estate speculation in the 1970s, the formal question (is this adequate as institutional critique?) could not be separated from the ethical question (does it serve the public interest by exposing concentrated power?). Krauss analyzed Haacke's work structurally—as institutional critique occupying a specific position in the expanded field—but acknowledged that the work's value included its political efficacy, not merely its formal precision. Purpose is the dimension where formal analysis meets political and ethical judgment, and the meeting is unavoidable when outputs have real-world consequences.
AI production forces the purpose question at scale because the democratization of capability distributes the power to produce outputs affecting others to populations who may lack the institutional support, professional training, or ethical frameworks that traditionally mediated between capability and deployment. The solo builder creating a healthcare decision-support system with Claude in a weekend occupies genuine technical capability without the regulatory oversight, liability insurance, or professional standards that hospital-deployed systems require. The output may be formally adequate (the code works, the model performs) while being ethically inadequate (it has not been tested on edge cases, it reproduces training-data biases, it serves the builder's interests without adequately considering patients').
Purpose evaluation is the hardest criterion to operationalize because it requires judgments about value that formal analysis alone cannot generate. Specificity, care, and structural awareness can be cultivated through practice and assessed through examination of formal properties. Purpose requires the builder to ask what the output is for, beyond the immediate functional goal, and whether that purpose is worth serving. The question cannot be answered by the AI, cannot be delegated to the tool, cannot be optimized away. It is the irreducibly human contribution, and it is the contribution most at risk of being skipped in the rush from prompt to deployment.
Origin
The criterion is introduced in this volume's tenth chapter as the necessary supplement to Krauss's formal method. Krauss herself did not theorize purpose as an evaluative category—her analytical project was structural, not ethical—but the application of her method to AI production reveals that formal adequacy is insufficient when outputs affect populations who did not consent to their production or deployment.
Key Ideas
Formal analysis insufficient alone. AI outputs have consequences that structure can identify but cannot evaluate—ethics and politics are unavoidable.
What, whom, how well. What does this serve, whom does it serve, does it serve them adequately—the three-part purpose interrogation.
Beyond performativity. Mere efficiency is insufficient—purpose demands care for affected populations that performativity metrics do not capture.
Capability without framework. Democratization distributes the power to produce outputs affecting others without distributing the ethical and institutional apparatus mediating deployment.
The irreducible human question. Purpose cannot be optimized, delegated, or answered by the AI—it is the contribution that remains after execution is automated.
Appears in the Orange Pill Cycle
Further reading
- Haacke, Hans, and Bourdieu, Pierre. Free Exchange. Stanford University Press, 1995.
- Deutsche, Rosalyn. Evictions: Art and Spatial Politics. MIT Press, 1996.
- Bishop, Claire. Artificial Hells: Participatory Art and the Politics of Spectatorship. Verso, 2012.