Making the Unaccountable Internet: The Changing Meaning of Accounting in the Early ARPANET

Accountability was actively designed out, not technically inevitable: The paper challenges the common narrative that the Internet's lack of accountability mechanisms was an unavoidable technical consequence of the End-to-End principle. Instead, it demonstrates that accountability was a deliberate design choice reflecting specific values and priorities of early Internet architects.

"Accounting" meant different things at different times: Early Internet designers conceptualized accounting in four distinct ways—as billing, measurement, management, and policy—and these shifting definitions directly shaped what kinds of accountability (or lack thereof) were built into the system's architecture.

Technical and administrative choices are inseparable: The paper reveals that decisions about how to technically manage shared resources in a distributed network were fundamentally intertwined with administrative and policy decisions, suggesting that "technical infeasibility" arguments often mask underlying value choices rather than genuine constraints.

Historical precedent matters for AI governance debates: By examining how early Internet designers grappled with accountability questions, the paper suggests that contemporary claims about the technical impossibility of designing accountable AI systems should be viewed skeptically—similar arguments were made about the Internet, yet alternatives existed.

Design choices have long-term consequences: The paper implies that decisions made during the Internet's early development created path dependencies that persist today, suggesting that choices made now about AI accountability mechanisms will similarly shape what's possible in the future.