Abstract
For two centuries, every major information technology widened who could speak and be heard. Print, the telegraph, the telephone, film, radio, television, the internet, social media: each one eventually pushed communicative and political capability outward, even when the channel it ran through passed through long stretches of monopoly. Artificial intelligence is the first in that line to run the pattern backward. Its core input is frontier compute, which is capital-intensive, physically scarce, and produced through an extremely concentrated supply chain. So AI concentrates the capability to create itself, not just the channel that carries it, and nothing guarantees it reopens on the schedule history has taught us to expect.
The paper makes two contributions. The first is the Diffusion-Concentration Diagnostic: a working instrument for judging whether an information technology distributes power or concentrates it, applied to nine technologies across the modern lineage. The second is the Inversion finding: that AI tracks the lineage on the dimensions of channel concentration but breaks from it on the dimension of generative capability, and that the break is of a kind history gives no guarantee of reversing. The implication for governance is plain: the countervailing powers that redistributed every earlier technology have to be organized before lock-in, not recovered after it.
What’s inside
- Prospectus. The argument in brief: the contribution, the research questions, the Diagnostic, the thesis, the three-method design, and the governance payoff.
- Part I · Literature Review. The genealogy from communications-and-development (Pye, Deutsch) through the cyclical view (Wu) to the political economy of technology (Acemoglu & Johnson), ending on the gap.
- Part II · The Coded Lineage (Method 1). The anchored rubric, the nine-technology dataset, the heat map and the inversion trajectory, with per-technology justifications.
- Part III · Paired Case Studies (Method 2). The mechanism: two structured comparisons, broadcast television against frontier AI, and the Arab Spring against frontier AI, tracing why capability concentrates and whether it reopens. The chapter that answers the sharpest objection to the coding.
- Part IV · Expert Elicitation (Method 3). A modified-Delphi instrument, specified as a protocol and not yet fielded, for testing the thesis against independent expert judgment.
The deposit also includes the full coded dataset: the codebook, the per-cell coding sheet, and the scores behind the figures, all under CC BY 4.0.
Keywords
artificial intelligence · compute governance · political economy of technology · communication and development · concentration of power · countervailing power · media literacy · anticipatory governance
How to cite
Adkins, J. (2026). The Inversion: Information technologies and the diffusion or concentration of power: why artificial intelligence breaks the pattern (Working paper, Version 1.4). Zenodo. https://doi.org/10.5281/zenodo.20481645
@techreport{adkins2026inversion,
author = {Adkins, Jeffrey},
title = {The Inversion: Information technologies and the
diffusion or concentration of power: why
artificial intelligence breaks the pattern},
year = {2026},
institution = {Independent working paper},
version = {1.4},
doi = {10.5281/zenodo.20481645},
url = {https://doi.org/10.5281/zenodo.20481645}
}