Artificial intelligence is increasingly framed in terms of efficiency and growth. But that framing sidelines harder questions about power, choice and democratic governance.

The policy conversation around artificial intelligence has become dominated by two metrics: productivity gains and innovation outputs. These framings are understandable. Governments seeking to justify AI investment want measurable returns. Businesses want efficiency improvements they can quantify. Researchers want evidence that their work delivers impact. The economic framing, for all its practical appeal, obscures questions that matter for democratic societies.

The conflation of productivity and innovation in policy discourse warrants careful attention. Productivity concerns efficiency. Innovation concerns novelty. AI contributes to both, but through different mechanisms and with different implications for how benefits and risks are distributed across the economy and society.

In the application of AI, there is a distinction between automation and augmentation. This is much more than a technical classification: it reflects competing theories of value creation with divergent distributional consequences.

Automation logic focuses on task substitution. It asks: which human activities can AI perform more cheaply, quickly or accurately? The economic case rests on labour cost reduction and throughput increases. In research contexts, automation enables high-throughput screening, automated literature review, routine data processing, and systematic replication of experimental protocols. The gains are real but bounded by the scope of tasks amenable to algorithmic execution.

Augmentation logic asks how AI can enable humans to do things they could not otherwise do, or to operate at levels of complexity and scale previously inaccessible. This framing positions AI as cognitive infrastructure that amplifies human judgment, creativity and insight rather than replacing it. In research contexts, augmentation........

© Pearls and Irritations