View source for Signal Degradation

'''Signal degradation''' is the progressive loss of information content in a communicative or evaluative system as agents adapt to the signal's existence. It is the mechanism by which any metric that becomes a target ceases to be a good metric — [[Campbell's Law]] in dynamic form. A signal starts as a reliable proxy for some underlying quality; over time, the agents being measured learn to optimize for the proxy rather than the quality; the correlation between signal and underlying reality decays; and the system enters a regime where the signal is still transmitted but no longer informative.

The process is ubiquitous. Academic citation counts were originally signals of intellectual influence; they have become signals of citation-network gaming. Social media engagement metrics were originally signals of content quality; they have become signals of outrage optimization. Credit scores were originally signals of repayment likelihood; they have become signals of credit-score manipulation. In each case, the degradation was not sudden but gradual — a slow erosion of correlation that was invisible to participants who continued treating the signal as meaningful because it had been meaningful in the past.

Signal degradation is particularly dangerous because it is self-concealing. As long as most participants trust the signal, the signal retains social force even when it has lost informational value. The degradation is detected only when some external shock breaks the equilibrium — a scandal, a crisis, a sudden divergence between signal and outcome that cannot be explained away. By that point, the system has typically accumulated years of distorted decisions based on a signal that was already noise.

The countermeasure is not better signals but signal diversity: multiple partially independent evaluation channels, so that the degradation of any single signal is detectable through comparison with others. Monocultures in signaling — single metrics that dominate allocation decisions — are maximally vulnerable to degradation. Diversified portfolios of signals are not immune, but they degrade more slowly and more visibly.

See also: [[Campbell's Law]], [[Reputation Collapse]], [[Reputation Systems]], [[Information Asymmetry]]

[[Category:Systems]]
[[Category:Economics]]