The distinction between informational and affective labor is decisive for understanding what AI does to work. AI can simulate the linguistic markers of empathy, warmth, and care — producing text that reads as emotionally engaged. But the engagement is simulated rather than genuine, and the people on the receiving end, though they may not articulate why, feel the difference. The nurse who performs identical words without emotional engagement does not produce the same care. The teacher who delivers identical material without enthusiasm does not produce the same education.
Judgment, taste, and vision — the capacities You On AI identifies as the irreducible human contribution after AI automates implementation — are not purely cognitive. The architect whose judgment tells her a design is wrong is experiencing an affect: a feeling of wrongness, a dissonance. Taste is an aesthetic feeling, a pre-cognitive orientation. Vision is an emotional relationship to a future that does not yet exist. These capacities consume emotional energy. A day of judgment calls is not merely thinking but continuous affective investment.
The Berkeley study's documentation of task seepage takes on new significance through this lens. The pauses AI eliminated — the debugging sessions, the dependency conflicts — served as buffers of low emotional expenditure. Their elimination transforms the workday into continuous affective labor, an unbroken sequence of judgments each requiring genuine emotional investment. The flat affect and diminished empathy the researchers documented are symptoms of affective depletion — the exhaustion of emotional reserves.
The gendered dimension demands attention. Affective labor has historically been disproportionately performed by women and by communities whose traditions emphasize relational responsibility — care workers, teachers, nurses, mediators. Feminized, devalued, rendered invisible. AI's concentration of human value in the affective dimension should, in principle, produce revaluation. In practice, affective labor resists measurement: the team leader who sustains morale has produced enormous value that appears in no productivity metric. When AI makes cognitive output abundant and affective contribution scarce, the scarce resource should command a premium — but the premium accrues only to resources the market can see. The concept of the affective commons names a shared reservoir of emotional capacity that the AI economy intensifies demand upon while providing no structure for replenishment.
The concept was developed by Michael Hardt and Antonio Negri in collaboration with Lazzarato and others in the post-autonomist tradition, drawing on feminist scholarship — particularly Arlie Hochschild's The Managed Heart (1983) — that had long insisted on the economic significance of emotional labor. The framework synthesized this feminist analysis with Marxist political economy, identifying affective labor as a central category of post-industrial production.
Production of affects, not information. Affective labor produces qualities of interpersonal experience that cannot be reduced to the words spoken or the information conveyed.
Structural impossibility for AI. Machines can simulate emotional engagement but cannot invest genuine feeling — a categorical difference, not a temporal gap.
Judgment and taste as affect. The human contribution that remains after AI automation is not purely cognitive but suffused with emotional investment.
Depletion without buffer. AI-augmented work eliminates the low-affect cognitive buffers that protected emotional capacity from continuous extraction.
The affective commons. Emotional capacity is a shared resource produced collectively and depleted by unlimited extraction, requiring institutional protection rather than individual self-care.