Cargo Cult
A cargo cult is a social phenomenon that emerged in Melanesian societies during and after World War II, when indigenous populations observed that certain ritual behaviors — building airstrips, wearing coconut headsets, marching with sticks shaped like rifles — correlated with the arrival of material goods ('cargo') from technologically advanced outsiders.
The term has become a critical concept across multiple fields because it names a specific pattern: the replication of surface features of a successful system without understanding the causal structure that produces success. The cultists built runways not because they misunderstood aviation, but because they correctly observed a correlation between runways and cargo, and reasoned backward from correlation to causation.
As Analytical Concept
In science and technology, 'cargo cult' describes the replication of methodological or formal features without comprehension of why those features matter. Richard Feynman's 1974 commencement address at Caltech made the term famous in scientific discourse: 'In the South Seas there is a cargo cult of people... They've got the form wrong. The form isn't enough.' Feynman was criticizing researchers who followed the rituals of scientific method — controls, statistics, publication — without the underlying epistemic discipline that makes those rituals meaningful.
The pattern generalizes:
- Education: Teaching problem-solving procedures as rote recipes rather than as tools grounded in conceptual understanding
- Management: Adopting organizational practices from successful companies without the contextual conditions that made those practices effective
- Software engineering: Copying architectural patterns from successful systems without understanding the tradeoffs and constraints that shaped those patterns
- AI research: Reproducing the surface features of successful models — scale, architecture, data volume — while treating the underlying scientific understanding as secondary
Causal Structure vs. Surface Correlation
The cargo cult pattern reveals a fundamental challenge in causal inference. Correlation is cheap to observe. Causal structure is expensive to establish. When a system produces impressive outcomes, the temptation is to copy what is visible — the practices, the tools, the rituals — rather than what is invisible: the knowledge, the constraints, the feedback loops that make the practices effective in their original context.
The Benchmark Engineering problem in AI is a cargo cult phenomenon: researchers optimize the surface metric (benchmark score) while missing the deeper causal question of what the benchmark actually measures. The Replication Crisis in psychology is another: methodological rituals (p-values, sample sizes, significance testing) were adopted as cargo from statistical theory without the theoretical infrastructure that makes those rituals meaningful.
The Rationality of Cargo Cults
Cargo cult behavior is not irrational. Given the information available to the cultists — advanced technology producing goods, visible correlations between rituals and outcomes — the inference that mimicking the rituals might produce outcomes is cognitively reasonable. The error is not in reasoning but in the information environment: the causal structure was invisible, and the visible correlations were insufficient.
This has implications for how to prevent cargo cult phenomena. Simply exhorting people to 'understand the fundamentals' is ineffective when the causal structure is genuinely opaque. The remedy is not better individual reasoning but better institutions for making causal structure visible: transparency, adversarial examination, and the documentation of failure modes that reveal what does not work.